Dmitry Yu.

Murzin
Engineering Catalysis
Also of interest
Chemical Product Technology
Murzin, 
ISBN ----, e-ISBN ----

Product and Process Design

Driving Innovation
Harmsen, de Haan, Swinkels 
ISBN ----, e-ISBN ----

Chemical Process Synthesis.

Connecting Chemical with Systems Engineering Procedures
Bezerra, 
ISBN ----, e-ISBN ----

Process Synthesis and Process Intensification.

Methodological Approaches
Rong (Ed.), 
ISBN ----, e-ISBN ----

Chemical Reaction Technology

Murzin, 
ISBN ----, e-ISBN ----
Dmitry Yu. Murzin

Engineering
Catalysis

2nd Edition
Authors
Prof. Dmitry Yu. Murzin
Åbo Akademi University
Process Chemistry Centre
Biskopsgatan 8
20500 Turku/Åbo
Finland
[email protected]

ISBN 978-3-11-061442-8
e-ISBN (PDF) 978-3-11-061443-5
e-ISBN (EPUB) 978-3-11-061469-5

Library of Congress Control Number: 2019955320

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.

© 2020 Walter de Gruyter GmbH, Berlin/Boston

Cover image: zorazhuang / E+ / Getty Images
Typesetting: Integra Software Services Pvt. Ltd.
Printing and binding: CPI books GmbH, Leck

www.degruyter.com
To the memory of Elena Murzina
Preface to the first edition
If you do not stay the road will lead you
Andrei Platonov

I wrote this book – devoted to the engineering of heterogeneous catalysts and

reactions – in the reverse order of chapters. The last chapter on the technology of
industrial catalytic reactions was written first, with the intention that the needs of
technological implementations will define the reactions and reactors, the reactors set-
ting the requirements for catalytic materials, their preparation and characterization,
and so on. It is up to the readers to decide if this approach was successful. In any
case, the book is based on many literature sources, reflecting my experience as a re-
searcher, chemical engineer, manager and educator, first in a governmental research
institute, then in industry and finally in academia, while always maintaining very
close ties to industrial catalysis and catalytic engineering.

https://doi.org/10.1515/9783110614435-202
Preface to the second edition
Apparently, this book devoted to the engineering aspects of heterogeneous catalysis
found its readership, as the publisher suggested to prepare the second edition. One
of the reasons was that the author, with experience in both industry and academia,
tried to combine both perspectives on catalysis, which are sometimes antagonistic
as illustrated in a somewhat provocative slide

Industrial perspective: Academic perspective: publish

should work in practice in high-impact factor journals
Real feedstock Model feedstock
Continuous reactors Often batch reactors
Simple and reliable catalysts Fancy catalysts
Detailed mass balance and adequate anlaytics Do not spent too much time on analytics
Deactivation: an issue to consider Deactivation better not to discuss
Kinetic and reactor modeling Kinetic and reactor modeling: What is this?

presented by the author at the International Congress on Catalysis in Beijing, 2016.

The first edition of the book was corrected and expanded in scope with more cover-
age of industrial preparation of catalysts and their technological implementations.

https://doi.org/10.1515/9783110614435-203
Contents
Preface to the first edition VII

Preface to the second edition IX

About the author XI

1 The basics 1
1.1 Catalytic concepts 1
1.1.1 Definitions 1
1.1.2 Length and time scales in catalysis 5
1.1.3 Catalytic trinity: activity, selectivity, stability 7
1.1.4 Composition of catalysts 13
1.2 Reactivity of solids 16
1.2.1 Physisorption and chemisorption 17
1.2.2 Basics of chemisorption theory 20
1.2.3 Surface crystallography 21
1.2.4 Mechanisms of some catalytic reactions 24
1.3 Catalysis in industry and for environmental protection 30
1.4 Fuel cells and electrocatalysis 45
References 47

2 Engineering catalysts 49
2.1 Catalyst design 49
2.1.1 Being in shape 50
2.1.2 Catalysis informatics and high-throughput experimentation 57
2.2 Toolbox in catalysis 59
2.2.1 General overview of the characterization methods 59
2.2.2 Adsorption methods 62
2.2.3 Physisorption methods 63
2.2.4 Chemisorption 74
2.2.5 Temperature-programmed methods 77
2.2.6 Calorimetry 82
2.2.7 X-ray diffraction 83
2.2.8 X-ray photoelectron spectroscopy and X-ray fluorescence 84
2.2.9 Infrared and Raman spectroscopies 87
2.2.10 Catalyst particle size measurements 93
2.2.11 Electron paramagnetic/spin resonance 94
2.2.12 Mössbauer spectroscopy 96
2.2.13 X-ray absorption spectroscopy 98
2.2.14 Nuclear magnetic resonance 99
XIV Contents

2.2.15 Imaging of catalysts 101

2.2.16 Catalytic reactions: product analysis 110
2.2.17 Theory as a part of a toolbox 114
2.3 Preparation of catalytic materials 117
2.3.1 General overview 117
2.3.2 Unsupported metals 132
2.3.3 Preparation of bulk oxides by precipitation 135
2.3.4 Heteropoly acids 145
2.3.5 Catalyst supports 146
2.3.6 Supported catalysts 180
2.3.7 Catalyst forming operations 202
References 224

3 Engineering reactions 229

3.1 Introduction 229
3.2 Thermodynamics 229
3.3 Kinetics 232
3.3.1 Definitions 232
3.3.2 Reaction mechanism 237
3.4 Kinetics of complex reactions 247
3.4.1 Theory of complex reactions kinetics 247
3.4.2 Relationship between thermodynamics and kinetics 256
3.4.3 Non-ideal surfaces 257
3.4.4 Kinetic aspects of selectivity 261
3.4.5 Structure sensitivity 266
3.4.6 Mechanism-free kinetics – kinetic polynomial 268
3.4.7 What is behind a rate constant? 269
3.4.8 Apparent activation energy of complex reactions 272
3.4.9 Dynamic catalysis and deactivation 274
3.4.10 Mathematical treatment of experimental data 280
3.5 Mass transfer 284
3.5.1 Diffusion effects in heterogeneous catalysis 284
3.5.2 Reactor dependent external diffusion (interphase mass transfer,
film diffusion) 286
3.5.3 Calculation of diffusion coefficients 295
3.5.4 Size dependent internal (pore) diffusion 298
3.5.5 Non-isothermal conditions 307
3.5.6 Multiple reactions and diffusional limitations 311
3.5.7 Diffusion in micropores 314
3.5.8 Criteria for the absence of diffusional limitations 315
3.6 Catalytic reactors 320
3.6.1 Laboratory reactors 320
 Contents XV

3.6.2 Industrial reactors 328

3.6.3 Two-phase reactors 328
3.6.4 Three-phase catalytic reactors 336
3.6.5 Reactor modeling 365
3.6.6 Catalyst handing in a plant 377
References 386

4 Engineering technology 391

4.1 General structures of chemical processes 391
4.1.1 Safety in design 393
4.1.2 Conceptual process design: examples 394
4.1.3 Conceptual process design: general comments 404
4.1.4 Reactor selection 406
4.2 (Petro)chemical industry 423
4.3 Fluid catalytic cracking 423
4.3.1 Feedstock 424
4.3.2 Reactions/mechanism 426
4.3.3 Kinetics/process variables 431
4.3.4 Catalysts 433
4.3.5 Technology 437
4.4 Hydrocracking 441
4.4.1 Overview of hydrocracking 441
4.4.2 Hydrocracking catalysts 444
4.4.3 Hydrocracking technology 448
4.5 Steam reforming of natural gas 454
4.5.1 General 454
4.5.2 Process conditions for sulfur removal and primary reforming 457
4.5.3 Kinetics and mechanism 459
4.5.4 Technology 463
4.5.5 Catalysts 465
4.6 Ammonia synthesis 468
4.6.1 General 468
4.6.2 Thermodynamics 469
4.6.3 Kinetics 471
4.6.4 Catalysts 474
4.6.5 Reactors and process design 480
4.7 Oxidation 483
4.7.1 General 483
4.7.2 Epoxidation of olefins 485
4.7.3 Oxidation of alkanes to anhydrides 491
4.7.4 From alcohols to aldehydes: oxidative dehydrogenation of
methanol to formaldehyde 507
XVI Contents

4.7.5 Ammoxidation 512

4.8 Oxychlorination 518
4.8.1 Overview 518
4.8.2 Catalysts and reactors 520
4.8.3 Reaction network and kinetics 524
4.8.4 Technology 525
References 527

Acknowledgments 533

Recommended reading 535

Index 539
About the author
Dmitry Yu. Murzin studied Chemical Technology at the Mendeleev University of Chemical Technology
in Moscow, Russia (1980–1986) and graduated with honors. He obtained his PhD (advisor
Prof. M.I. Temkin) and DrSc degrees at Karpov Institute of Physical Chemistry, Moscow in 1989 and
1999, respectively. He worked at Universite Louis Pasteur, Strasbourg, France and Åbo Akademi
University, Turku, Finland as a post-doc (1992–1994). In 1995–2000 he was associated with BASF,
being involved in research, technical marketing and management. Since 2000 Prof. Murzin holds the
Chair of Chemical Technology at Åbo Akademi University. He serves on the editorial boards of several
journals in catalysis and chemical engineering field. He is an elected member of Academia Europaea
and the Finnish Academy of Science and Letters.
Prof. Murzin is the coauthor with Prof. T. Salmi of a monograph Catalytic Kinetics (Elsevier,
2005, second edition in 2016), and author of Chemical Reaction Technology (de Gruyter, 2015) and
Chemical Product Technology (de Gruyter, 2018). He holds several patents and is an author or coauthor
of approximately 750 journal articles and book chapters.

https://doi.org/10.1515/9783110614435-204
1 The basics

1.1 Catalytic concepts

1.1.1 Definitions

Catalysis is a phenomenon related to the acceleration of rates of chemical reactions.

Such acceleration was recognized long time ago and put into practice even before
the concept of catalysts was formulated. For example, a German chemist Döbereiner
discovered that oxidation of hydrogen can occur in the presence of Pt already in 1823.
Such invention resulted in substantial utilization of Döbereiner lamps just within few
years after the discovery.
The available data on the conversion of starch to sugars in the presence of
acids, combustion of hydrogen over platinum, decomposition of hydrogen peroxide
in alkaline and water solutions in the presence of metals, were summarized by a
Swedish scientist J. J. Berzelius in 1836 [1], who proposed the existence of a certain
body, which “effecting the (chemical) changes does not take part in the reaction
and remains unaltered through the reaction”.
Catalytic power or force, according to Berzelius, meant that “substances (cata-
lysts) are able to awaken affinities which are asleep at this temperature by their
mere presence and not by their own affinity”.
This concept was immediately criticized by Liebig, as this theory was placing
catalysis somewhat outside of other chemical disciplines [2]. A catalyst was later
defined by Ostwald as “a compound that increases the rate of a chemical reaction,
but which is not consumed by the reaction” [3]. This definition allows for the possi-
bility that small amounts of the catalyst are lost in the reaction or that the catalytic
activity slowly declines.
From these definitions, a direct link between chemical kinetics and catalysis is
apparently clear as, accordingly, catalysis is a kinetic process. There are, however,
many issues in catalysis that are not directly related to kinetics, such as mech-
anisms of catalytic reactions, elementary reactions, surface reactivity, adsorption of
reactants on solid surfaces, synthesis and structure of solid materials and catalytic
engineering.
An important issue in catalysis is selectivity towards a particular reaction. For
example, transformation of synthesis gas (a mixture of CO and hydrogen) can lead
either to methanol (on copper) or high alkanes (on cobalt). For consecutive reac-
tions it could be desirable to obtain an intermediate product. In the oxidation of
ethylene such target is ethylene oxide, but not CO2 and water.
+ O2 +O2
CH2=CH2 CH2– CH2 CO2 + H2O
O

https://doi.org/10.1515/9783110614435-001

Brought to you by | The University of Texas at Austin

Authenticated
Download Date | 2/15/20 2:18 PM
2 1 The basics

Although formaly catalysts remained unchanged during a reaction, they are in-
volved in chemical bonding with the reactants during the catalytic process in a
cyclic process: the reactants are bound to one form of the catalyst, and the prod-
ucts are released from another, regenerating the initial state (Fig. 1.1).

R (reactant)

Catalyst Catalyst-R

P (product) Fig. 1.1: Catalytic cycle.

It should be also noted that although it is usually assumed that a catalyst participates
in the process but remains unchanged at the end, there could be major changes in its
structure and composition.
Potential energy diagrams for the catalytic and non-catalytic reactions presented
in Fig. 1.2 indicate that in both cases the reactions should overcome a certain barrier,
which is lower in the presence of a catalyst.

G x‡

Uncatalyzed

Catalyzed

A+B

A+B C+D
C+D

Reaction coordinate

Fig. 1.2: Potential energy diagram for catalytic and non-catalytic reactions. Changes of Gibbs
energy along the reaction coordinate.

As follows from Fig. 1.2, the change in the Gibbs free energy between the reactants
and the products ΔG is the same, independent of the presence or absence of a cata-
lyst, but providing, however, an alternative reaction path.
 1.1 Catalytic concepts 3

A lower value of activation energy implies higher reaction rates, which could
be expressed through the rate constant k dependence on temperature:
− Ea
k = AT m e RT (1:1)

where A is the pre-exponential factor, Ea is the activation energy related to the po-
tential energy barrier and m is a constant. Equation (1.1) was proposed by Kooij and
van’t Hoff [4] to explain the temperature dependence of reaction rates and could be
derived from the transition state theory of Eyring and Polanyi [5]. Arrhenius [6] ap-
plied a slightly simplified form:
− Ea
k = ko e RT (1:2)

A decrease in activation energy and an increase in temperature lead to an increase

of the rate constant and thus of the reaction rate. Catalysts substantially accelerate
chemical reactions, enabling them to be carried out under the most favorable ther-
modynamic regime, and at much lower temperatures and pressures.
Equation (1.2) corresponds to an elementary reaction. In the case of complex re-
actions, it is appropriate to discuss the so-called apparent activation energy, which
can be expressed through the reaction rate r according to

∂ ln r ∂ ln r
Eact, apparent = −R = RT 2 (1:3)
∂1=T ∂T

Figure 1.3 also shows that if a catalyst is active in enhancing the rate of the forward
reaction it will do the same with a reverse reaction. As mentioned above, the catalyst

Gas-phase reaction

Ea
Energy

ΔH
Adsorption
E‡a
Product
Reaction Desorption

Reaction coordinate

Fig. 1.3: Potential energy diagram with catalyst binding.

4 1 The basics

affects only the rate of the reaction, but not the thermodynamics (for example,
the Gibbs energy of the reaction is the same), the approach to equilibrium or
the equilibrium composition. Thus, in the case of a thermodynamically unfa-
vorable process, there is no hope of finding an active catalyst that will beat
thermodynamics.
Instead, as a first step the reaction conditions (temperature, pressure and re-
actant composition) must be optimized to maximize the equilibrium concentration
of the desired product. Once suitable reaction conditions are identified the cata-
lyst screening is performed with the aim of finding a suitably active and selective
material.
The statement – that thermodynamics frequently limits the concentration of
a desired product and that a catalyst search will not help – sounds very straight-
forward and obvious. The author knows, however, a case where modern methods
of high throughput experimentation were used in industry to find an active cata-
lyst for a reaction with a maximum conversion allowed by thermodynamics of
only 1%.

Several years ago I started to be involved in a project related to the synthesis of dimethyl car-
bonate from methanol and CO2. The literature was full of papers describing various catalytic
systems and the mechanisms behind their catalytic action. Unfortunately, explicit data on
conversions and thermodynamics were absent from the literature. The PhD student doing the
experiments was initially very frustrated when, after a couple of months of catalyst screening,
very low conversions in the range of 0.1% were obtained. After repeated advises finally ther-
modynamics was calculated and it became clear that the reaction is heavily thermodynami-
cally limited. This switched the strategy from catalyst screening to finding a means of shifting
reaction equilibrium, in that case by in situ water removal.

Figure 1.2 represents a simplified situation, not taking into account the binding of
the reactant to the catalyst (called adsorption in heterogeneous catalysis), which is
illustrated in Fig. 1.3. The energy barrier between the catalyst-substrate and the
transition state should be lower than between the substrate and the transition state
in an uncatalyzed reaction (Fig. 1.3).
If the energy is lowered too much when the substrate is bound to the reactant
during adsorption, and the activation energy in the catalytic reaction is still rather
high compared to non-catalytic reactions, then the reaction rate is slow and catalysis
is not effective. In addition, if the binding of the products is too strong they will not
be desorbed (released) from the catalyst. At another extreme, if the binding is too
weak then the catalytic cycle could not proceed as effectively. It is thus intuitively
clear that bonding between a catalyst and a reactant should not be too strong. This
principle was first put forward by Sabatier [7], who proposed that a catalytic reaction
has an optimum (maximum) rate as a function of the heat of adsorption, and then
further developed by Balandin [8], who introduced volcano plots, relating catalytic
activity with adsorption energy (Fig. 1.4).
 1.1 Catalytic concepts 5

1,000 450 °C
100 bar
100 3:1 H2/N2

3
NH
%
10
01
0.
1 Ru
3
NH
Os
TOF (s–1)

1%

0.1
0.

0.01 Fe
3
NH

Co
1%

0.001

0.0001
3
NH

0.00001 Mo Ni
%
10

0.000001
–150 –100 –50 0 50 100
Relative nitrogen binding energy (kJ/mol)

Fig. 1.4: Activity (turnover frequency) in ammonia synthesis with different catalysts as a function of the
relative nitrogen binding energy, 450 °C, 100 bar, H2/N2 =3. (Reproduced with permission from [9]).

Sabatier is known for his experimental work on hydrogenation of organic com-

pounds, for which he was awarded the Nobel Prize in 1912. Together with Senderen,
he discovered that traces of nickel catalyze methanation of CO2:

400  C
CO2 + 4H2
! CH4 + 2H2 O
Pressure

This reaction is witnessing now a renaissance associated with a strong desire to di-
minish CO2 emissions. This reaction has also a special application, as it can be ap-
plied to produce water from exhaled CO2 in confined spaces (e.g., submarines or
the International Space Station). An important issue related to the latter applica-
tions is efficient and rather small-scale generation of hydrogen.

1.1.2 Length and time scales in catalysis

Heterogeneous catalysis covers a wide scale, from a molecular scale (nm) of an

active site (Fig. 1.5A) to a catalytic reactor (m) scale (Fig. 1.5B). Catalysis occurs
on the surface of solid materials, representing the chemistry in two dimensions.
Active sites of heterogeneous catalysts are related to a molecular or atomic level
arrangement of atoms, responsible for catalytic properties (Fig. 1.5A). At this sub-
nanometer level, chemical reactions proceed that involve the rupture and
6 1 The basics

1–2 m

H2 (g)

10–20 m
Attractive interaction H2 (adsorbed on Pt)
2H (adsorbed
on Pt)

Re-face adsorption

(A) (B)

Fig. 1.5: Length scales in catalysis ranging from (A) an active site to (B) a reactor.

formation of chemical bonds. The reactivity of solids and mechanisms of cata-

lytic reactions will be briefly reviewed here in Chapter 1.
Catalytically active particles are typically between 1 and 10 nm in size (Fig. 1.6A)
and are located inside the pores of a support material (μm level, Fig. 1.6B). Chemical
composition, texture of materials and pore structure are very important issues, re-
lated not only to catalysis, but also to the transport of molecules through the pores
to the active sites. Of particular interest are the size, shape, structure and composi-
tion of the active particles.

1 cm
(A) (B) (C)

Fig. 1.6: Different levels in catalysis. (A) active site level (subnanometer scale), (B) catalyst particle
powder (μm) level), (C) shaped catalyst (mm) level. Photos courtesy of Dr. E. Toukoniitty.

Catalysts in the form of powders can be used in industrial processes only in limited
cases. Shaped catalysts, in the form of extrudates, pellets and tablets on a mm scale
(Fig. 1.6C), are introduced into industrial reactors. Engineering of such materials
 1.1 Catalytic concepts 7

requires addressing porosity, mechanical strength, and attrition resistance in addi-

tion to activity, kinetics and mass transfer.
The engineering of catalytic materials through tailor-made preparation and the
understanding of kinetic and transport phenomenon will be discussed in Chapters 2
and 3. Chapter 3 also covers the modeling of catalytic reactors.
It should be noted that in industrial reality the mass and heat transport through
the catalyst bed can be as important as intrinsic kinetics.
Finally, the performance of a catalyst should be viewed as part of an overall
process that includes feed treatment, separation, etc., in addition to chemical re-
actions. For example, issues such as catalyst mechanical stability, sensitivity to
trace impurities in the reactant feed (as industrial feeds are rarely pure) and
leaching of the active phase can be very crucial to industrial operation. Basic
principles of designing catalytic processes (plant level, Fig. 1.5B) will be described
in Chapter 4, with examples for several catalytic processes, representing different
type of materials (supported metals, oxides, zeolites), catalyst shapes (powder, extru-
dates, tablets, irregular grains) and different type of reactors (with fixed and moving
catalysts).
The time scale of catalytic events varies from the picosecond level of bond
breaking and formation to the seconds/minutes level for completion of a catalytic
cycle. Pore diffusion may take seconds and minutes, while the residence times of
molecules in reactors may vary from seconds (fluidized catalytic cracking) upwards
(coke can stay for months or years on the catalyst before eventual regeneration or
disposal).

1.1.3 Catalytic trinity: activity, selectivity, stability

The main requirements of a catalyst for an industrial process depend on the trinity
of catalysis: activity, selectivity and stability. Activity refers to the ability to conduct
a process within a reasonable contact time, which influences the reactor dimensions
and process capacity. Insufficient activity in principle could be compensated by
higher catalyst amounts or some other means, such as higher temperature. Catalyst
selectivity is probably the most important characteristics of a catalyst, which should
be also sufficiently stable under operation conditions.
The reaction rate is calculated as the converted amount of the substance with
time relative to the amount of catalyst (mass or volume), specific surface area or to
the amount of active sites. In cases when the rate is defined per catalyst mass, the
unit of measurement is mol g–1 h–1.
The rate per amount of sites available for the reaction (exposed sites) is denoted
in heterogeneous catalysis as the turnover frequency (TOF) [or turnover number
(TON)], which is defined as the number of molecules reacted per site per unit time:
8 1 The basics

TOF = Amount of reacted substrate=ðactive site x timeÞ = mol mol − 1 s − 1 = s − 1 (1:3)

Even if the TOF concept is widely applied, its application is not straightforward,
because TOF depends on kinetics, i.e., the concentration of reagents and temper-
ature, and thus strictly speaking should be referred to particular conditions.
Typical values of TOF for industrially relevant reactions are in the range 10–2 ÷
102 s–1.

Exercise 1.1: Fig. E1.1 displays a fixed-bed reactor where there is a certain molar flow in the reac-
tor n_ Ain and out of the reactor n_ Aout . In such reactors, during catalyst screening conversions at
constant space velocity [volume flow rate (m3 s–1) relative to catalyst (kg)] are compared.
mcat

n•A n•A
in out

• •
n = cA V

Fig. E1.1. Fixed-bed reactor.

Calculate:

1. The conversion rates per catalyst volume

2. Weight
3. Surface
4. TOF (metal density 1,015 sites cm−2)

for a continuous reaction of benzene hydrogenation over 5 cm3 5 wt% Pd on a support with the
catalyst density of 0.9 g cm−3 and the catalyst surface area of 250 m3 g−1 (5% of which is cov-
ered by Pd) when the molar flow of benzene in the reactor is 4 mol h−1 and the outlet flow is
3.7 mol g−1.

Hints: The rates are defined in the following way:

– per volume rA = –dnA/(Vcatdt) = − dn_ A =Vcat

– per weight rA = –dnA/(mcatdt) = –dnA/Vcat ρcatdt)
– per surface rA = –dnA/(Scatdt) = –dnA/(Srel mcatdt)
– per atom of Pd rA = –dnA/(npddt)

The number of mols of Pd should be calculated using the following strategy. From the overall sur-
face, the surface of Pd atoms could be computed. This in turn gives the number of atoms of Pd if
the surface density is known. Finally the number of moles of the metal can be easily calculated.
 1.1 Catalytic concepts 9

Exercise 1.2: Lilja et al. [10] investigated esterification of propanoic acid with methanol over a hetero-
geneous fiber catalyst. The temperature dependence is presented in Fig. E1.2.
1

0.9

0.8

0.7
c/c0

0.6
55 °C
0.5
60 °C

0.4
63 °C

0.3
0 50 100 150
Time (min)

Fig. E1.2: Esterification kinetics of propanoic acid with methanol (the initial molar ratio 1:1) as a
function of temperature.

Calculate:
1. The activation energy, based on initial rates (10 min).
2. Gibbs energy of the reaction at 60 °C based on experimentally recorded equilibrium data.
3. Reaction enthalpy at 60 °C, assuming that the entropy change during the reaction is negligible.
Hints:

– Initial rates (ΔC/dt) should be calculated from the slopes for the initial period when the rates
are not changing with time.
– Composition at equilibrium defines equilibrium constant (Keq = Cproducts/Creactants) and subse-
quently Gibbs energy (ΔG = –RTlnK).
– Enthalpy of the reaction is defined through Gibbs energy: ΔG = ΔH – TΔS.

Exercise 1.3: Steam reforming of heptane was performed over a 6 wt% cobalt catalyst supported on
γ-alumina. The amount of catalyst was 15 cm3. Knowing conversion and specific activity calculates
the molar feed rate, the rate per catalyst volume and catalyst weight, and apparent activation
energy.

o
C Conversion of heptane (%) Specific catalytic activity (mol/(mol(Co) s))
  .

  .

  .
10 1 The basics

The turnover number is a useful concept, but is limited by the difficulty of determin-
ing the true number of active sites. The situation is somewhat easier for metals, as
chemisorption, which will be discussed further (Chapter 2, Section 2.2.4), could be
used to measure the exposed surface area.
For some reactions (structure insensitive) the rate is independent on size, shape
and other physical characteristics, while for structure-sensitive reactions the rate de-
pends on the detailed surface structure.
In order to increase turnover number it is obvious that the number of exposed
sites should be increased, which is possible with small sizes of active particles pro-
viding high surface area per unit volume (Fig. 1.7). This can be achieved with the
catalyst particles in the nanoscale range, thus placing catalysts in the domain of
nanomaterials.

1.0
NS/NT

0
1 1,000 Fig. 1.7: Ratio of surface (NS) to total (NT) atoms as a
dcluster (Å)
function of cluster size dcluster.

It should be noted that many granting agencies are somewhat reluctant to con-
sider heterogeneous catalysis as a part of nanoscience and nanotechnology, espe-
cially in comparison with some emerging applications. This is probably because
catalysis is a more established concept, with many dozens of decades of experi-
ence in synthesis of such materials as supported metals and zeolites.
Small particles of the active catalyst alone cannot provide thermostable highly
active catalysts because of sintering of them at conditions of catalyst preparation
and catalysis. Moreover, the separation of reaction products from nanosized cata-
lysts is far from being trivial, in many cases even impossible. Therefore an active
phase (responsible for activity and selectivity) is usually deposited on a thermosta-
ble support (Fig. 1.8), which also provides the required shape, mechanical strength
and pore structure.
Expensive active compounds (such as precious metals) are typically prepared
with the objective of maximizing the active surface area per unit weight of the ac-
tive compound, which is achieved at low loadings and rather small metal clusters.
 1.1 Catalytic concepts 11

Without support

With support,
no sintering

Fig. 1.8: Supported active phase. (Reproduced with permission from [11]).

For the less expensive active phase (supported base metal catalysts) the target is
maximum active surface area per unit volume, thus allowing higher loading of the
support with the active phase.
Fig. 1.8 depicts a simplified view on the catalyst nanocluster. In fact, the cluster
can contain different types of sites (Fig. 1.9A), with different binding to neighbor
atoms and thus differences in the ability to bind (adsorb) reacting molecules. As a
consequence, we can observe notable differences between the reactivity of different
surface atoms and active sites (edges, corners and faces as shown on Fig. 1.9B).

Face Corner Edge

atom atom atom

Adatom
Adatom
Corner
atom Edge
Corner

Edge
atom Corner
atom
Face
Corner Face
atom atom
Adatom Edge
(A) atom (B)

Fig. 1.9: Different types of sites in a nanosized catalyst.

By controlling the amount of different surface sites (atom) we can, in principle, con-
trol catalyst reactivity and selectivity. There are however practical challenges in the
preparation of a cluster with a desired size and a narrow cluster size distribution.
12 1 The basics

Relative amounts of surface atoms with different reactivity vary as a function of

metal particle size, thus leading to structure sensitivity (TOF dependence on the
cluster size, Fig. 1.10).

III 6

TOF (s–1)
TOF

I 4
II

Diameter 2 4 6 8
(A) (B) Cluster size (nm)

Fig. 1.10: (A) I, II correspond respectively to structure sensitivity, while III represents a structure
insensitive reaction and (B) ethene hydrogenation on Pd/TiO2 (data from [12]).

The temperature dependence of heterogeneous catalytic reactions was partially ad-

dressed above with the point that, in general, activity increases with temperature.
Increasing the temperature, however, shortens the life span of the catalyst and typi-
cally increases the rate of undesirable reactions. Moreover, a catalyst with an extremely
high reaction rate will certainly exhibit large mass- and heat-transfer limitations.
Although the catalyst must possess high activity and long-term stability, the most
important characteristic is selectivity, the ability to direct transformation of reactants
to the desired product along one specific pathway. For many reacting systems, various
reaction paths are possible, and the catalyst type often determines the path that will
be followed. A catalyst can increase the rate of one reaction without increasing the
rate of others. In general, selectivity depends on pressure, temperature, reactant com-
position, conversion and the nature of the catalyst. Therefore, selectivity should be re-
ferred to specific conditions.
There are two definitions of selectivity that are commonly used. Integral selec-
tivity is defined as the ratio of the desired product per consumed reactant, while
differential selectivity corresponds to the ratio of the desired product formation rate
to the rate of reactant consumption.
Although there is always a desire to have a stable catalyst for a particular reac-
tion, strong catalyst deactivation does not necessarily mean that in such a case a
particular catalyst cannot be applied at all. In fact, it is only in theory that the cata-
lyst remains unaltered during the reaction. Actual practice is far from this ideal, as
the progressive loss of activity could be associated with coke formation, attack of
 1.1 Catalytic concepts 13

poisons, loss of volatile agents and changes of crystalline structure, which cause a
loss of mechanical strength. There are industrial examples, discussed in Chapter 4,
that show successful industrial implementation of catalytic reactors in combination
with continuous regeneration, when the catalyst life is merely few seconds.
Because of the extreme importance of catalyst deactivation, the kinetic aspects of
this phenomenon will be treated in a separate section (Section 3.4.5).
In summary, it can be concluded that the target priorities in catalyst develop-
ment and applications are typically: selectivity > stability > activity.

1.1.4 Composition of catalysts

The catalytic properties of activity, selectivity and stability are closely related to the
composition of the catalyst. Typical catalytic materials are shown in Fig. 1.11. Most
catalysts are multicomponent and have a complex composition. Components of the
catalyst include the active agent itself and may also include a support, a promoter,
and an inhibitor.

Dispersed Pt/Al2O3, Ni/Al2O3, Pd/Al2O3

Metals Porous Raney nickel
Bulk Pt, Pd, Ag gauze

Single Al2O3, Cr2O3, V2O5

Oxides Dual, complex SiO2/Al2O3, CuCr2O4
Catalytic materials

Dispersed NiO/Al2O3, MoO3/Al2O3

Sulfides MoS2/Al2O3, WS2Al2O3

Acids SiO2–Al2O3; zeolites; natural clays

Base CaO, MgO, K2O, Na2O

Fig. 1.11: Catalytic materials.

For example, metal catalysts are not in their bulk form but they are generally dis-
persed on a high surface area insulator (support) such as Al2O3 or SiO2. Many het-
erogeneous catalysts need supports, which, from an economic viewpoint, are a
means of spreading expensive materials and providing the necessary mechanical
strength, heat sink/source, help in optimization of bulk density and in dilution
of an overactive phase. The supports also provide geometric functions (an
14 1 The basics

increase in surface area, optimization of porosity, crystal and particle size)

and chemical functions (improvement of activity, minimization of sintering,
and poisoning, as well as the beneficial effect of spillover, which will be dis-
cussed later).
Sintering of supports can start at a temperature substantially lower than the
melting point, thus supports with high melting points are applied, such as alumina
or silica. In particular, alumina is mainly preferred because of favorable bulk density,
thermal stability and price. It is applied in the form of γ-alumina in many reactions,
such as hydrotreating, hydrocracking, hydrogenation, reforming, oxidation, and so
on. Another phase of alumina (α-alumina) that stable at very high temperatures is
used for steam reforming of natural gas.
Silica has a lower bulk density, which means that more reactor volume is needed
for the same active phase loading on the support. Silica is more liable to sintering
above 900 K than alumina and is volatile in the presence of steam and elevated pres-
sures. There are still several applications (polymerization, oxidation, hydrogenation)
where silica is used, and a very specific example is the oxidation of sulfur dioxide to
trioxide (Fig. 1.12).

20 mm 12 mm 10 mm 9 mm 6 mm
rings daisy rings daisy cylinder

Fig. 1.12: Vanadium-based catalysts supported on silica for sulfur dioxide oxidation.

Although there were catalyst formulations using silica synthesized from water glass
(sodium silicate), the catalyst manufacturers predominantly use kieselguhr. More
details of the manufacturing are provided in Section 2.3.5.4. This diatomaceous
earth (diatomite) is a form of silica composed of the siliceous shells of unicellular
aquatic plants of microscopic size. Kieselguhr is heat resistant and, in addition to
catalytic applications, has been used as an insulator, a component in toothpaste
and an abrasive in metal polishes. The specific features of the diatomaceous earth
(surface area 20–40 m2) are related to small amounts of alumina and iron as part
of the skeletal structure and a broad range of pore sizes. The main components of
the sulfur dioxide oxidation catalyst include SiO2 as a support, vanadium, potas-
sium and/or caesium and various other additives. The reaction actually occurs
within a molten salt consisting of potassium/caesium sulfates and vanadium sul-
fates, coated on the solid silica support. Vanadium is present as a complex sulfated
salt mixture and not as vanadium pentoxide (V2O5).
 1.1 Catalytic concepts 15

Aluminosilicates, either amorphous or crystalline (zeolites), can be used as

supports and catalysts (as zeolites) because of their acidic properties: Active car-
bon is another support that is often applied, mainly for the synthesis of fine
chemicals in various hydrogenation reactions. Other supports include clays,
ceramic (cordierite), titania (for selective oxidations and selective catalytic re-
duction of NOx) as well as magnesium aluminate (steam reforming of natural
gas).
While selecting a support for a particular reaction, various physical and chemical
factors should be carefully considered. The physical factors are the desired surface
area and porosity, thermal conductivity, mechanical stability, resistance to coking,
and the possibility of applying a desired geometrical shape; while the chemical fac-
tors are the activity of the support, chemical interactions between the active catalyst
phase or the products/reactants and the support, and the resistance to poisoning.
Details on the various supports used in catalysts and their preparation will be given
in Chapter 2.
Sometimes both the metal and the support act as catalysts, which is called bi-
functional catalysis. An example of this is platinum dispersed on alumina used in the
gasoline reforming. Platinum is active in dehydrogenation reactions of linear alkanes
to linear alkenes which further undergo skeletal isomerization on alumina. Formed
branched olefins are hydrogenated on platinum. The active agent, discussed above is
the component, which causes the main catalytic action, while a promoter (being in-
active, when added into the catalyst in small amounts) improves activity, selectivity
or stability, prolonging the catalysts lifetime.
There are various types of promoters that improve the catalyst in different ways.
Textural promoters are inert substances, which inhibit sintering of the active catalyst
by being present in the form of very fine particles. Usually they have a smaller parti-
cle size than that of the active species, are well-dispersed, do not react or form a solid
solution with the active catalyst and have a relatively high melting point. Oxides
(with melting points in parentheses) such as Al2O3 (2,027 °C), SiO2 (1,700 °C), ZrO2
(2,687 °C), Cr2O3 (2,435 °C), CeO2 (2,600 °C), MgO (2,802 °C) and TiO2 (1,855 °C) can be
used as textural promoters.
Structural promoters change the chemical composition, produce lattice defects,
change the electronic structure and the strength of chemisorption. For example, Pb,
Ag, or Au can be used as structural promoters for palladium catalysts applied in se-
lective hydrogenation of acetylene to ethylene.
An inhibitor is the opposite of a promoter. When added in small amounts, it
diminishes activity, selectivity or stability. Inhibitors are useful for reducing the
activity of a catalyst for an undesirable side reaction. For example, silver sup-
ported on alumina is an excellent oxidation catalyst, widely used in the production
of ethylene oxide from ethylene. However, under the same conditions complete ox-
idation to carbon dioxide and water also occurs, resulting in mediocre selectivity
16 1 The basics

to C2H4O. It was found that the addition of halogen compounds to the catalyst in-
hibits this complete oxidation and results in satisfactory selectivity [13].
Metal oxides usually consist of bulk oxides. As semi-conductors, metal oxides
catalyze the same type of reactions as metals but are used in processes that require
higher temperatures. Often a multicomponent mixture of various oxides is used to
increase the catalytic activity.
For example: transition metals such as MoO3 and Cr2O3 are good catalysts for
the polymerization of olefins; a mixture of copper on chromium oxides, called cop-
per chromite, is used for hydrogenation; and a mixture of iron and molybdenum
oxide, called iron molybdate, is used for formaldehyde formation from methanol.

1.2 Reactivity of solids

Heterogeneous catalysts in industrial applications have a variety of different geo-

metrical shapes and porosity. Therefore, understanding of mass and heat transfer
in heterogeneous catalysis is essential for the description of the processes involved
in transporting the reactants to the catalyst surface and the transfer of energy to
and from catalyst particles.
The following steps are usually followed:
1. External diffusion: transfer of the reactants from the bulk fluid phase to the
fluid-solid interface (external surface of the catalyst particle).
2. Internal diffusion (if the particle is porous): transfer of the reactants into the
pores of a catalyst particle.
3. Adsorption: physisorption and chemisorption of the reactants at the surface sites
of the catalyst particle. During adsorption the reactants become attached to the
surface either by physical (van der Waals) or chemical forces, thus significantly
decreasing the entropy compared to the gas- or liquid-phase. This leads to nega-
tive ΔS, thus for the Gibbs energy (ΔG = ΔH – TΔS) to be negative, the heat of
adsorption (ΔH) should be negative and adsorption should be exothermic.
4. Surface reaction: the chemical reaction of adsorbed species to produce ad-
sorbed products. It can proceed by formation of precursors on the catalyst sur-
face, and surface diffusion of these as well as of other adsorbed species to the
reaction sites.

Langmuir advanced the first kinetic approach in catalysis based on physico-

chemical understanding of the underlying processes [14], by applying the mass ac-
tion law to reactions on solid surfaces. This was done by supposing that the rate of
an elementary reaction is proportional to the surface concentration (coverage) of
reactive species adsorbed on the surface. The Langmuir model of uniform surfaces
assumes that all the surface sites are identical, binding energies of the reactants are
the same independent of the surface coverage, and interactions between adsorbed
 1.2 Reactivity of solids 17

species may be neglected. This ideal adsorbed layer is considered to be similar to

ideal solutions with fast surface diffusion, thus allowing an application of the mass
action law.
Typically two main types of mechanisms are considered for a bimolecular reac-
tion. Either one reactant (A) is adsorbed and the other (B) approaches from the fluid
phase (the Eley-Rideal mechanism, see Fig. 1.13A) or a reaction occurs because of
the interaction of two adsorbed reactants A and B (the Langmuir-Hinshelwood
mechanism, see Fig. 1.13B). The products are then released either directly into the
fluid phase or first through adsorbed product species followed by desorption.

A* A* B*

(A) (B)

Fig. 1.13: (A) the Eley-Rideal and (B) the Langmuir-Hinshelwood mechanisms of catalytic reactions.

Detailed analysis of catalytic kinetics for various reaction mechanisms, including

Eley-Rideal and Langmuir-Hinshelwood ones, will be provided in Chapter 3.
Further steps in heterogeneous catalysis are:
5. Desorption: the release of adsorbed products by the catalyst.
6. Internal diffusion: the transfer of products to the outer surface of the catalyst
particle.
7. External diffusion: the transfer of products from the fluid-solid interface into
the bulk fluid stream.

1.2.1 Physisorption and chemisorption

Adsorption in heterogeneous catalysis takes place before surface reactions and thus
deserves special consideration.
In the case of chemisorption of a molecule (adsorbate), a chemical bond is
formed between the molecule and the surface (adsorbent). For chemisorption, the
adsorption energy is comparable to the energy of a chemical bond. The molecule
may chemisorb intact or it may dissociate with breaking chemical bonds. The chem-
isorption energy is 30–70 kJ mol–1 for molecules and 100–400 kJ mol–1 for atoms.
18 1 The basics

In physisorption, the bond is a van der Waals interaction and the adsorption
energy is typically 5–10 kJ mol–1, which is much weaker than a typical chemical
bond. As a consequence the chemical bonds in the adsorbing molecules remain in-
tact. As the van der Waals interactions between the adsorbed molecules are similar
to the van der Waals interaction between the molecules and the surface, many
layers of adsorbed molecules may be formed. Some other features of physisorption
and chemisorptions are presented in Tab. 1.1.

Tab. 1.1: Comparison between chemisorption and physisorption.

Features Chemisorption Physisorption

Adsorbent Sorbent and crystallographic plane Independent on the surface

specificity dependent structure
Type Can be dissociative and irreversible Non-dissociative and reversible
Temperature A wide range Near or below boiling point of
adsorbate
Heat of A wide range (– kJ mol–) Close to heat of liquefaction
adsorption (– kJ mol–)
Saturation Monolayer coverage Multilayers are formed
Kinetics Can be activated, for example, slow Fast, non-activated

Fig. 1.14 shows potential energy diagrams of chemisorption and physisorption.

When approaching the surface a molecule encounters the weak van der Waals inter-
actions. If the molecule loses energy upon interactions with the surface, it may be
trapped in the weak attractive potential at a distance corresponding to the sum of
the van der Waals radii of the molecule and the surface atom, and thus become phys-
isorbed on the surface. Moving even closer to the surface will result in repulsion.

E (d)

Physisorption

d (nm)

Chemisorption

Fig. 1.14: Potential energy curves for chemisorption and physisorption.

 1.2 Reactivity of solids 19

After rearranging its electronic configuration and interacting with the electron
clouds of the metal, the molecule, however, may become chemisorbed.
The equilibrium position of this new chemisorbed state is at a shorter distance
than for physisorption, Moreover, for chemisorption the potential energy curve is
dominated by a much deeper chemisorption minimum. The chemisorption curve in-
tersects with the physical adsorption curve at a value close to zero in the energy
axes. For a clean surface this barrier between physisorption and chemisorption is
below zero, thus chemisorption on clear surfaces is usually fast and non-activated.
Dissociative chemisorption can occur for a diatomic molecule.
The structure of solid surfaces is fairly complicated (Fig. 1.15) as they are usu-
ally complex and rough, and consisting of high Miller index (to be discussed later)
planes with different energetics.

Fig. 1.15: Surface structure.

Moreover, adsorbed species can form different types of complexes on the surfaces
(some examples are given in Fig. 1.16), which might interact with each other (lat-
eral interactions).
As a result, the heat of adsorption might decrease with coverage, with the mini-
mum of adsorption energy decreasing in magnitude. This leads to a situation when
such a decrease of the adsorption heat will lead to an activation energy of adsorp-
tion (Fig. 1.17).
Physisorption at low temperatures is a standard method for determination of
the surface area of a catalyst, pore size and volume, and pore size distributions;
this will be discussed in detail in Chapter 2 along with the application of chemi-
sorption measurements for active surface area, surface site energetics and deter-
mination of catalytic sites.
20 1 The basics

O
C C C C C C
H C C
M M M M M M M M M M M M

C O
M M
M C C
C C C M M M M H
M M M M M M M M M M

C C M

C C O O M
M
C C M M M
M M M

Fig. 1.16: Examples of the complexes that could be formed on the solid surfaces. M, metal site.

E (d)

d (nm)

Fig. 1.17: Chemisorption proceeding with activation.

1.2.2 Basics of chemisorption theory

Chemisorption of catalytically active metals that have 6–10 d electrons is usually

considered for analyzing the behavior of d electrons. There is an overlapping of or-
bitals, when, if large, gives a broad band and strong interactions. Such band is a
collection of a large number of molecular orbitals that form a continuum when they
are close to each other. In metals the bands are filled with valence electrons up to
the Fermi level or to the highest occupied state (Ef) analogously to the highest occu-
pied molecular orbital (HOMO). If an empty level exists below Ef it will be filled in-
stantaneously. A band behaves like a molecular orbital, in a sense that lower levels
are bonding, upper levels are antibonding and middle levels are non-bonding.
 1.2 Reactivity of solids 21

Behavior of d orbitals is important in understanding catalytic activity of the tran-

sition metals. Details on the reactivity of transition metals on the basis of d-band
analysis are provided in specialized textbooks [15].
When an atom is chemisorbed, bonding and antibonding chemisorption orbitals
should be considered. The strength of chemisorption depends on the position of the
bonding orbital. If interactions of an atomic orbital and a d-band are strong then
there is substantial splitting between the bonding and antibonding orbitals. The anti-
bonding orbitals are unoccupied, leading to strong chemisorption (Fig. 1.18A). In the
opposite case of weak interactions, small splitting can result in an antibonding or-
bital falling below E. Such an orbital is occupied, leading to repulsive interactions
and thus no chemisorption (Fig. 1.18B).

Evac Evac

Antibonding Antibonding

EF EF
d-Band d-Band

Bonding Bonding
s-Band s-Band No bonding

d-Metal Adsorbed Free d-Metal Adsorbed Free

(A) atom atom (B) atom atom

Fig. 1.18: Chemisorption of an atom. (A) strong chemisorption, (B) no bonding. (From [15]).

In the case when a molecule is chemisorbed, chemisorption orbitals are formed from
both bonding and antibonding molecular orbitals of a molecule (Fig. 1.19).
Filling of the orbital of the molecule, which was originally antibonding, makes
interactions with the surface stronger, while at the same time weakening the intra-
molecular bond of the adsorbed molecule and facilitating, for example, dissociation
of a molecule.
Filling of the bonding orbital is called donation, favoring “on-top” adsorption
(discussed in Section 1.2.3). Filling of the antibonding orbital, which binds a mole-
cule additionally to the surface, is called “back donation”.

1.2.3 Surface crystallography

The crystallographic structure of the metals that are most important for catalysis is
presented in Fig. 1.20(A).
22 1 The basics

σ-Orbitals σ*-Orbitals

Evac Antibonding

σ*
Antibonding
EF 1s 1s
Bonding

Bonding

d-Metal
Adsorbed molecule Free molecule

Fig. 1.19: Chemisorption of a molecule. (Reproduced with permission from [15]).

Fig. 1.20B shows three most important metal crystal structures: face-centered
cubic (fcc), hexagonally close-packed (hcp) and body-centered cubic (bcc). Other
structures can exist for more complex materials such as oxides, the bulk structures
of which can be represented by positive metal ions (cations) and negative O ions
(anions).
For catalysis, which is a surface phenomenon, it is not the bulk structure that is
important, but the exposed surface. As energy is required to cleave the surface, the
total energy is higher and the surface free energy is always positive.
Reactivity of the surface depends on the number of surrounding atoms, or in
other words, on packing. Dense regular packing (Fig. 1.21) has lower energy.
In order to discuss surface crystallography – the two-dimensional analogue of
bulk crystallography – the nomenclature of various metal planes (Fig. 1.20) should
be explained. The assignment of indices is done using the following approach. A
particular surface plane of a crystal is considered and is shown in Fig. 1.22.
Firstly the intercepts on the x-, y- and z-axes are identified. In Fig. 1.22 the inter-
cept on the x-axis is at x = a [at (a,0,0)], corresponding to the unit cell distance. The
surface is parallel to the y- and z-axes, thus there are no intercepts on these two axes
and they are considered to be at infinity (∞). The intercept coordinates on the x-, y-
and z-axes (a, ∞,∞) are converted to fractional coordinates by dividing by the respec-
tive cell-dimension (a) to give the fractional intercepts: (1, ∞,∞). The reciprocal values
of the fractional intercepts are 1, 0 and 0, yielding the Miller indices (100), which is
specified without being separated by any commas or other symbols (Fig. 1.23).
The surfaces in Fig. 1.23 are different in terms of their coordination or number of
neighbors. The most stable solid surfaces are those with a high surface atom density
 1.2 Reactivity of solids 23

bcc hcp fcc

(A)

bcc hcp fcc

(100) (110) (111)

(B)

Fig. 1.20: (A) Crystallographic structure of some catalytically active metals. (B) Metal crystal
structures.

that contains surface atoms with a high coordination number. For face-centered
cubic structures (representative of Ni, Cu, Pt, Pd, Ir, Au, Ag, etc.) the following order
of stability is valid: fcc (111) > fcc (100) > fcc (110). The most stable surface is then the
most closely packed. At the same time, the more open the surface is the more reactive
it is. It should be mentioned that surfaces of real catalysts are usually rough and con-
sist of high Miller index planes.
Figure 1.24 shows the most common adsorption sites: on-top, bridge (long or
short) and hollow (three-fold or four-fold).
From the discussion above it is apparently clear that the classical treatment of
adsorption and kinetics, which assumes the same adsorption strength for all sites, is
an oversimplification. In addition to different energetics of adsorption an adsorbed
molecule can bind to several metal sites and there could be interactions between
24 1 The basics

Energy
Typical neighbor
bond length

r
Typical neighbor
bond energy

(A)

Energy
Typical neighbor
bond length

r
Typical neighbor
bond energy

(B)

Fig. 1.21: (A) Dense, regular packing. (B) Non-dense, random packing.

(x,0,0)
x Fig. 1.22: Assignment of indices.

adsorbed species. This complicates the mathematical treatment of adsorption and ki-
netics, which will be addressed in detail in Chapter 3.

1.2.4 Mechanisms of some catalytic reactions

There are countless heterogeneous catalytic reactions with very many different
types of catalysts. It is thus impossible in a textbook focused on engineering rather
than on mechanisms of catalysis to describe even a small fraction of these, there-
fore only a few generic examples for some typical reactions will be mentioned here.
 1.2 Reactivity of solids 25

(100) (110) (111)

Fig. 1.23: Various metal surface planes.

(100) (110) (111)

1 2 1 2 1 2

6
5
4
3 4

1 On top 2 Bridge 3 Long bridge

Threefold Threefold
4 Fourfold hollow 5 6
hollow FCC hollow BCC

Fig. 1.24: Adsorption sites. (From [15]).

1.2.4.1 Oxidations
In industry, typical oxidation reactions over heterogeneous catalysts are usually
partial or selective oxidations, such as:

– ethylene epoxidation

Ag2O
C C C C
O
26 1 The basics

– oxidation of propene to acrolein

O
C C C C
Bi2O3–MoO3
C C C C

– oxidation of benzene or butane to maleic anhydride

O O

V2O5–MoO3 [cat.]
O + CO2 + H2O + 3.5 O2 O + 4 H2O
+ O2 400 °C

O O

In terms of carbon use, butane oxidation is much more efficient than benzene oxi-
dation, as according to stoichiometry in benzene oxidation two moles of CO2 are
formed per one mole of reacting substrate.
In the selective oxidation reactions, CO2, CO and even water are undesirable
products.
Partial oxidation reactions (and selective catalytic reduction, Fig. 1.25) are typi-
cally carried out using such oxides as, for example, V2O5, MoO3, or molybdates. They
proceed through changes in oxidation states following the Mars-van Krevelen redox
mechanism.

Reduction

V=O Standard SCR V – OH

Oxidation

½ H2O ¼ O2 Fig. 1.25: Mechanism of selective catalytic reduction.

In this mechanism (kinetics will be presented in Chapter 3) reduction of the oxide is

dependent on the organic reactant (NO and ammonia in the case of selective cata-
lytic reduction (SCR), or CO in Fig. 1.26) and takes place in the absence of gas-phase
oxidants (O2, H2O). Isotope labeling studies were used to confirm that oxygen in the
product comes from the lattice oxygen (Fig. 1.26A), but not directly from the gas-
phase species.
Re-oxidation of oxide happens independently of the reduction step (Fig. 1.26B).
Dissociative adsorption of oxygen often occurs at a site different from an oxidation
site, followed by diffusion through the bulk in order to refill the surface vacancy.
 1.2 Reactivity of solids 27

O
|
CO C CO2
|
O
M O M O M M M O M M M O M Interface

Oxide Oxide Oxide

(A)

O2

M M O M M O M O M Interface

Oxide Oxide
(B)

Fig. 1.26: The Mars-van Krevelen redox mechanism. (A) oxide reduction and (B) re-oxidation.

For oxidations not involving changes in oxidation state, such as oxidation of CO

on metals, a different mechanism could operate. For example, the substrates can co-
adsorb on the surface, which is followed by a reaction between the adsorbed species
according to the Langmuir-Hinshelwood mechanism (Fig. 1.27).

O
|
CO O2 C
— O— O CO2
| | |
Interface

Oxide Oxide Oxide

Fig. 1.27: Oxidation following a Langmuir-Hinshelwood mechanism.

Alternatively one reactant is adsorbed and the other reacts from the fluid phase.

1.2.4.2 Hydrogenation
From the fundamental as well as an industrial viewpoint, hydrogenations can be con-
sidered to be the most advanced field of catalysis. Various hydrogenation reactions
are employed in industry, such as hydrogenation of double, triple and aromatic
bonds, and C=O bonds to name a few. Different mechanisms were proposed for the
hydrogenation of olefinic double bonds [16]. The Horiuti-Polanyi mechanism as-
sumes that both the alkene and hydrogen must be activated by being adsorbed on
the surface (see Fig. 1.28).
28 1 The basics

1. H2 + 2 * 2H
*

2. C C + 2* C C
* *

3. C C + H C C + 2*
* * * * H

4. C C + H C C + 2*
* H * H H

Fig. 1.28: The Horiuti-Polanyi mechanism.

It should be noted that double-bond hydrogenation is stereo selective (Fig. 1.29),

corresponding to addition to the less crowded face of the double-bond (syn addition
of both H atoms).

H
CO2CH3 CO2CH3
H2, Pt

CO2CH3 CO2CH3
H
(100%)

Fig. 1.29: Syn addition to the double-bond.

The mechanism, which addresses stereo selectivity and is consistent with liquid-phase
hydrogenation kinetics, assumes the addition of hydrogen to the adsorbed organic
molecule, followed by isomerization in the adsorbed state of the formed olefin-
hydrogen complex (Fig. 1.30). The adsorbed product is displaced from the surface by
the incoming substrate. See [17] for a detailed discussion on the mechanism.

B
H2
ZA

ZB ZAH2

Fig. 1.30: Mechanism of catalytic hydrogenation. (From [17]).

 1.2 Reactivity of solids 29

1.2.4.3 Catalysis by solid acids

Crystalline aluminosilicates-zeolites can be used in catalysis because of the pres-
ence of acidic sites, which are generated by the substitution of silicon by aluminum
in the framework of silica. A detailed discussed on the properties of zeolites and
their preparation will be given in Chapter 2. Here we can mention that differences
in charges between Si+4 and Al+3 are balanced by protons, which is the reason for
strong Brønsted acidity of zeolites, caused by the presence of these protons.
The mechanism of acid-catalyzed reactions is thus connected to the presence of
protons and Lewis acid sites; the Lewis acid sites can act as solid acids by accepting
electrons. For example, double-bond isomerization in olefins such as butene starts
with the proton addition to the double-bond, forming a secondary butyl cation:

H2 C = CH − CH2 − CH3 + H +
!

H3 C − C H − CH2 − CH3
¯

This is followed by an internal hydride shift:

H3C–CH–CH2–CH3 H3C–CH2–CH–CH3
+ +

and the subsequent proton elimination:

H3 C − CH2 − C H − CH3
!

H3 C − CH = CH − CH3 + H
+
¯

Skeletal isomerization of olefins also proceeds with the addition of proton and fur-
ther rearrangement of carbenium ions because of their different stabilities:
CH3 CH3
H H2 H2 H2
R C C C R' R C C C R'
+ +
H
Thus, tertiary carbenium ions are more stable than secondary ones, with primary
ions the least stable. Because of the different mechanisms of proton addition, the
skeletal isomerization of alkanes is more difficult than that of olefins, requiring
stronger acid sites.
Catalytic cracking of heavy organic compounds is an industrially important re-
action in oil refining that will be covered in detail in Chapter 4. During cracking,
secondary carbenium ions are formed at random followed by cracking at the bond
at the beta position to the charged carbon atom, giving an olefin and a primary car-
benium ion:

RCH2CHCH2–CH2CH2R' RCH2CH=CH2 + CH2CH2R'

+ +

The primary ion undergoes a hydride shift to secondary ions, as discussed above. A
detailed account of the mechanism of catalytic cracking is provided in Section 4.3.
30 1 The basics

1.3 Catalysis in industry and for environmental protection

The immense importance of catalysis in the chemical industry is demonstrated by

the fact that roughly 85–90% of all chemical products have seen a catalyst during
the course of production. The number of catalytic processes in industry is very large,
with the catalysts coming in many different forms, such as heterogeneous catalysts
in the form of porous solids (Fig. 1.31), homogeneous catalysts dissolved in the liquid
reaction mixtures and biological catalysts in the form of enzymes. In fact, enzymes
could either be in the same phase as the liquid reactants or in their solid form (either
poor soluble enzymes or immobilized ones).

Fig. 1.31: Shapes of heterogeneous catalysts. (Reproduced with permission from [18]).

In theory, catalysts are not consumed in a chemical reaction, thus in an ideal sce-
nario for cost calculations only the cost of the initial catalyst should be considered.
In practice many industrial catalysts deactivate, requiring gradual replacement.
Nonetheless, the contribution of catalysts to the overall cost of a product is on aver-
age not very high (~3%) and subsequently a part of catalyst sales in relation to the
gross domestic product (GDP) is marginal (~0.1%). However, the share in GDP of
products made by catalysts could be as high as 25%. The global catalyst market was
estimated to be worth 27.7 billion dollars in 2015 and is projected to increase to
 1.3 Catalysis in industry and for environmental protection 31

33.5 billion USD in 2019. The main contribution to the global demand for catalysts
is associated with environmental applications (Fig. 1.32), while the second most im-
portant contribution is related to сrude oil refining (Fig. 1.33).

41%

17%
Refining Fine chemicals and intermediates
13% Petrochemicals Environmental
5%
Polymers

Fig. 1.32: Global catalyst value. (From [19]).

Gas

Desulfurization Isomerization
Light
naphtha
Gasoline
Heavy
naphtha
Reforming

Petrol Kerosene

Catalytic cracking

Gas oil Diesel

Fuel oil

Asphalt
Residues bitumen

Fig. 1.33: Future perspectives in catalysis. (From [20]).

Most growth in catalyst demand is outside the Organization for Economic Co-
operation and Development and is driven to a substantial extent by environmental air
quality and fuel specifications. The annual growth rate of emission control catalysts
(~12%) is higher than the average for the catalyst manufacturing industry (~8%).
32 1 The basics

In addition to tightening the regulations for mobile emissions and for fuel
quality, other trends driving future demand for catalysts are: improving process
efficiency by more efficient use of feedstock and higher yields; and increasing en-
ergy efficiency, expanding simultaneously the feedstock base to include coal, nat-
ural gas and biomass.
The contribution of catalysts to improving the environment is not only related
to end-of-pipe technologies (for example, the cleaning of wastes), but also to the
prevention of pollution by avoiding the formation of waste and unwanted by-
products through improving processes and selectivity in the existing ones.
Tab. 1.2 gives some examples of catalytic processes in oil refining for the pro-
duction of gasoline, diesel, kerosene, heating oils, while a photo of catalysts used
in refining as well as in the synthesis of petrochemicals is presented in Fig. 1.34.

Tab. 1.2: Major catalytic processes in oil refining.

Process Catalyst Conditions

Hydrodesulfurization NiS/WS/AlO – °C,  bar H

CoS/MoS/AlO
Hydrocracking MoO/CoO/AlO – °C, – bar H
Ni/aluminosilicate
Catalytic cracking of heavy fractions zeolites – °C, – bar
Catalytic reforming of naphtha Pt(+other metal)/alumina – °C, – bar H
Isomerization of light gasoline Pt/alumina – °C, – bar

Fig. 1.34: Refining and petrochemical catalysts (photo of the author).

 1.3 Catalysis in industry and for environmental protection 33

For various reasons there is a current trend to consider other feedstocks –

rather than oil – as a source of fuel, olefins and aromatics. Using feedstocks such as
natural gas, coal, and especially biomass, would require fundamental changes in
the chemical industry and refining.
It should be noted that the transport sector uses nearly 28% of all energy gener-
ated, and oil and gas are mainly used for the production of fuels, while the chemical
industry accounts for only 5–10% of the oil consumption. Because the first generation
of fuels and chemicals obtained from biomass (for example, bioethanol) has many
shortcomings, lignocellulosic biomass, not competing with the food chain, has at-
tracted a lot of attention. Lignocellulose contains cellulose (≈40 wt%), hemicellulose
(≈25 wt%), extractives (~5%) and lignin (the balance). With 700 billion metric tons
available on Earth, cellulose is the most abundant organic material.
Cellulose is a linear polysaccharide, built from glycoside units (the number of
monomers can be as high as 15,000), which are linked by β-1,4-glycoside bonds,
resulting in the formation of a crystal structure of cellulose with intra- and intermo-
lecular hydrogen bonds. As there is not much space around the glycoside (bridging)
oxygen in cellulose, there is limited access to this bond for homogeneous or hetero-
geneous catalysts. Hemicellulose (a polysaccharide) is composed of several types of
monomers, which imposes steric hindrance because of the presence of side chains
and prevents the formation of a crystalline structure. Lignin is a three-dimensional
polymer composed of propylphenol units. Since structure of biomass is different
from petroleum, technologies of biomass processing in general and in particular
involving catalysis differ from those widely used in petroleum refining. The
oxygen-to-carbon ratio in lignocellulosic biomass is close to unity, which is un-
suitable for the production of transportation motor fuels, i.e. hydrocarbons. At the
same time, a minor amount of oxygen is needed in gasoline, and thus the strategy
for using biomass for fuels should include substantial, but not complete, removal
of oxygen. In addition, the reactions in petroleum refining are often conducted in
the gas or liquid-phase in the presence of organic media, while catalytic reactions
of sugars and polyalcohols occur in an aqueous environment. Acidic catalysts
based on aluminosilicates and zeolites might be unstable and prone to deactiva-
tion because of the impurities and quality of the lignocellulose, which depends
substantially on the type of biomass and geographical location.
The same feedstock shift towards renewable resources is currently being inves-
tigated for the production of chemicals. Many intermediates used in the chemical
industry contain oxygen, which means that the carbon-to-oxygen ratio is close to
that for biomass in at least some chemicals such as methanol, acetic acid, and eth-
ylene glycol, and in the range 0.3–0.6 for other important chemicals. These sub-
stances do not require a very high degree of oxygen removal. Some of the typical
reactions used in biomass processing (oxidation, hydrogenation, cracking, etc.) are
similar to processes used in the chemical industry. Conversely, because of a large
number of hydroxyl and other functional groups, dehydration, aldol condensation,
34 1 The basics

ketonization, and decarboxylation, which are less frequent in the chemical indus-
try, would be required in biomass processing, when and if the concepts of biorefi-
nery are realized in industry.
In large-scale production of chemicals with oil and gas as the main feedstock,
there are many important reactions such as various hydrogenations (Ni, Co, Cu, Pd,
or Pt catalysts), dehydrogenations (on oxides, Ni, Fe, Cu), oxidations (over metals,
reducible oxides), and acid-catalyzed (zeolites, aluminosilicates) reactions, such as
alkylation, hydration, dehydration, condensation, etc. Some of these reactions are
presented in Tab. 1.3, but this is very far from being conclusive.

Tab. 1.3: Some catalytic processes for the synthesis of basic chemicals (From [18]).

Process or product Catalyst Conditions

Hydrogenation
Methanol synthesis ZnO–CrO – °C, – bar
CO + H → CHOH CuO–ZnO–CrO – °C,  bar
Fat hardening Ni/Cu – °C, – bar
Benzene to cyclohexane Raney Ni liquid-phase – °C, 
bar
noble metals gas-phase  °C, – bar
Aldehydes and ketones to alcohols Ni, Cu, Pt – °C,  bar
Esters to alcohols CuCrO – °C, – bar
Nitriles to amines Co or Ni on AlO – °C, – bar
Dehydrogenation
Ethylbenzene to styrene FeO (Cr, K oxide) – °C, . bar
Butane to butadiene CrO/AlO – °C,  bar
Oxidation
Ethylene to ethylene oxide Ag/support – °C, – bar
Methanol to formaldehyde Ag cryst. ~ °C
Benzene or butane to maleic anhydride VO/support – °C, – bar
o-Xylene or naphthalene to phthalic VO/TiO – °C, . bar
anhydride VO–KSO/SiO
Propene to acrolein Bi/Mo oxides – °C, . bar
Ammoxidation
Propene to acrylonitrile Bi molybdate (U, Sb – °C, – bar
oxides)
Methane to HCN Pt/Rh nets –, °C,  bar
Oxychlorination
Vinyl chloride from ethylene +HCl/O CuCl/AlO – °C, – bar
Alkylation
Cumene from benzene and propene HPO/SiO  °C, – bar
Ethylbenzene from benzene and ethylene AlO/SiO or HPO/  °C, – bar
SiO
 1.3 Catalysis in industry and for environmental protection 35

As mentioned above, environmental catalysis constitutes a substantial part of

all catalyst demand.

Photos presented in Figs. 1.35 and 1.36 are not directly related to catalysis, reflecting, however,
different problems associated with emissions in the environment. Pictures in Fig. 1.35 were taken
by the author on two successive days (Saturday and Sunday) from the same hotel room in
Beijing. Apparently, a lower demand for coal utilization during Saturday resulted in better visibil-
ity on the next day. The other photo (Fig. 1.36) is somewhat older and not directly related to the
personal experience of the author. During a drive along the Cuyahoga River in downtown
Cleveland, Ohio, my brother, who is residing in one of the suburbs nearby, showed the place
where the river caught fire because it became so polluted with chemicals.

Fig. 1.35: Different levels of smog in Beijing (photo of the author).

Historically, supported Pt/Al2O3 catalysts were first used in the 1940s for the cata-
lytic purification of off-gases by oxidation. The catalytic control of NOx emissions
from nitric acid and power plants, as well as various automobile exhaust emissions,
became driven by legislature. As non-catalytic methods require high temperatures
and non-selective methods use an excess of reducing agents, the SCR (Selective
Catalytic Reduction) of NOx had been developed to operate at moderate tempera-
tures and with low amounts of reductants. For stationary sources, ammonia SCR is
applied, with the ammonia dosed in stoichiometric quantities:

4NH3 + 4NO + O2 ) 4N2 + 6H2 O

In the presence of sulfur, undesired side reactions

SO2 + 1=2 O2 − > SO3
NH3 + SO3 + H2 O − > NH4 HSO4
36 1 The basics

Fig. 1.36: A fire tug fights flames on the Cuyahoga River near downtown Cleveland, Ohio, where oil
and other industrial wastes caught fire on June 25, 1952 [21].

can also occur. Apart from the parallel side reactions, selectivity is almost 100% at
the same conversion level. The potential drawback in NH3 SCR is ammonia slippage
that requires careful control of ammonia dosage.
Although different catalysts with different operation ranges can be used for am-
monia SCR (Fig. 1.37A), vanadium pentoxide, supported on titania and wash-coated
on a ceramic monoliths, is often applied in the form of blocks, which are mounted
in a larger structure (Fig. 1.37B).
The catalytic converters for automotive exhaust emissions, whose level is driven
by local legislature and is thus different in various parts of the world, are also mono-
lithic supports (ceramic or metallic) coated with platinum and other noble metals.
The monolith structures are demonstrated in Fig. 1.38.
The converters are placed behind the engine and the conventional silence mod-
ule (Fig. 1.38C). For conventional three-way catalysts (Fig. 1.39) the operation of a
catalytic exhaust control system is critical, particularly the monitoring of the air-fuel
ratio (Fig. 1.40). In an excess of fuel the concentration of unburned hydrocarbons
is high, while with an excess of air the conversion of NOx is diminished, caused
by the lack of the reducing agent. The following reaction might occur (where HC
are hydrocarbons):
 1.3 Catalysis in industry and for environmental protection 37

100 Distribution
Metal-zeolite layer
80
NOx conversion (%)

Pt V2O5
60 Monolith
layers
40
4 NH3 + 4 NO + O2 4 N2 + 6 H2O
Module
4 NH3 + 2 NO + O2 3 N2 + 6 H2O
20

0 Honeycomb
0 100 200 300 400 500 600 monolith
(A) Temperature (°C) (B)

Fig. 1.37: Catalysts for ammonia SCR: (A) temperature dependence of NOX conversion,
(B) monolithic blocks.

Fig. 1.38: Typical converters: (A) ceramic, (B) metallic, (C) placement.

NOx,HC, CO

Fuel Three-way
Engine Exhaust gas
Air catalyst (Pt/Rh)

Fig. 1.39: Three-way catalyst.

2CO + 2NO = N2 + 2CO2

1
HC + NO ! N2 + 2CO2′ .CO + O2 = CO2
2
HC + O2 ! CO2 + H2 O
38 1 The basics

100 HC
CO
Conversion (%) 80

60
O2
40
l=1
20 NOx
Diesel
mode
0
12 13 14 15 16 17 18
Air to fuel (wt/wt)

Fig. 1.40: Effect of the air-fuel ratio on emissions.

The three-way catalysts contain Pt, Pd and Rh. The latter is the most expensive
component in three-way catalysts, being, however, needed because of its activity
and high selectivity in reduction of NO to dinitrogen with low formation of ammo-
nia. Effective utilization and stabilization of Rh is thus important. Palladium oxide
displays moderate activity and selectivity in NO reduction, while Pt is the least se-
lective resulting in ammonia.
Multilayered washcoating strategy is applied in the design of three-way cata-
lysts to separate catalyst components, which can otherwise interact leading to deac-
tivation. Such multilayered washcoats comprise one or more porous, high surface
area supports, which are thermally stable, for example, La-stabilized alumina and
stabilized zirconia in different layers. Ceria–zirconia mixed oxide acts as an oxygen
storage. Hydrated alumina is used as a binder.
Table 1.4 illustrates a typical composition of three-way catalysts [22].

Tab. 1.4: Typical composition of three-way catalysts. (Reproduced with permission from [22]).

Component Composition Function

Supports – 10–12% La2O3/AI2O3 with Al2O3 High surface area, porous carrier; enables
being usually a mixture of γ, δ and θ preparation of well-dispersed precious
– ZrO2 metals, especially Pt and PdO, prevent
their sintering
Lanthana stabilizes alumina against loss
of surface area above  °C
ZrO is a noninteracting support for Rh,
which is used in a separate layer

Catalytic PdO, Pt, Rh (usually .– g/L on monolith) Pt, PdO oxidize CO and hydrocarbons; Rh,
phase(s) PdO reduce NO, using CO and
hydrocarbons as reducing agents.
 1.3 Catalysis in industry and for environmental protection 39

Tab. 1.4 (continued )

Component Composition Function

Oxygen Solid solution of  at% of ceria and % ZrO/CeO stores oxygen during oxidizing
storage of zirconia; about –% of washcoat cycle releasing it during reduction cycle;
material zirconia and rare-earth oxides (PrO and
NdO stabilize CeO against sintering)

Additives Less than  wt% of total catalyst Different types are used (e.g., nickel
and oxide diminishes the formation of HS)
promoters

More stringent regulations regarding fuel usage resulted in the introduction of

lean burn and diesel engines with approximately 25–30% less fuel consumption.
Engine manufactures in Europe mainly tuned engines to minimize particulates,
which increases NOx emissions. In the USA engine out emissions are mainly tuned
to minimize NOx leading to a high concentration of particulates (Fig. 1.41).

10 Typical
engine out –
European OE
Engine tuned
NOx (g bhp-h–1)

to minimize particulate Typical

engine out –
US OE
Engine tuned
to minimize NOx

0
0 0.22
PM (g bhp-h–1)
High temperature Low temperature

Fig. 1.41: Engine emissions map.

Conventional three-way catalysts do not work under an excess of air, so other tech-
nologies have been introduced for diesel and lean burn small, medium and large
(off-road) engines.
Figure 1.42 illustrates the NOx storage concept developed by the Toyota Motor
Corporation, which involves periodic operations and low sulfur fuels. In the lean
period (excess of air) NOx is oxidized into NO2 and stored as nitrates, while during
the rich (in fuel) period metal oxide is regenerated. The sulfur level in fuel has to be
minimized, as the storage material (Ba) has a higher affinity to sulfates than nitrates.
40 1 The basics

O2 CO2,H2O
CO NO2 Lean (in fuel) period
NO approx. 60 s
Pt Metal oxide NO is oxidized over Pt
HC
and stored as nitrates
Al2O3

HC N2, H2O, CO2

Rich (in fuel) period
CO approx. 8 s
Pt MNO3 Metal oxide is regenerated
from nitrates
Al2O3

Fig. 1.42: NOx storage concept. (From [23]).

Desulfurization, if done to regain activity, requires rich conditions and elevated tem-
peratures (> 700 °C) provoking thermal aging. NOx sensors are needed to control the
periodic operation and desulfurization, calling for an advanced engine control.
SCR of NOx emissions from diesel engines might use either ammonia (or in fact
urea for easier dosage, Fig. 1.43) or hydrocarbons.
Hydrocarbons SCR cannot rely on passive control as the amounts of unburned
hydrocarbons from the engine are not sufficient for the reduction of NOx over a cat-
alyst. Instead it requires an active control (the addition of fuel after diesel particu-
late filter upstream SCR catalyst, Fig. 1.44).
Catalysts in exhaust gas cleaning should operate constantly under highly transient
and very demanding conditions (extremely high flow rate: GHSV < 200,000 h−1, almost
zero pressure drop, simultaneous oxidation and reduction, presence up to ~12 vol%
H2O in the feed). Silver on alumina was recently reported to be an efficient catalyst for
hydrocarbons SCR with a potential to be employed industrially for off-road engines.
High fuel penalty (Fig. 1.45) and negligible low temperature activity could be among
the drawbacks of the catalyst, restricting its application.
Nevertheless, hydrocarbon lean NOx catalyst system using a silver-based cata-
lyst was developed by General Electric for large locomotive applications, and is pro-
vided by Tenneco for on- and off-road emission control markets.
A special case worth mentioning is development of SCR systems for different
types of ships, including large cruise ships (Fig. 1.46), as they use marine fuels with
a high sulfur content.
For example, already from 2009 the sulfur content in on-road diesel was limited
to 10 ppm in some regions of the world, while for ships operating outside desig-
nated emission control areas the current limit for sulfur content of marine fuel oil is
3.50% mass/mass. Obviously, SO2 emissions for fuels containing large amounts of
 1.3 Catalysis in industry and for environmental protection 41

Urea

AMOX
Engine DOC DPF SCR

M-zeolite
Diesel fuel + Air CO2 + H2O + NH3 + NO2 + O2 N2 + H2O

Diesel fuel Urea

Porous wall

Disel
oxidation
Gas
catalyst
NH3
600 °C
Scoot (C) + Air CO2 + H2O destruction
catalyst

Wall flow filter

with alternate
plugged channels

Cu-SSZ-13
N2 + H2O
NH3
NOx

Fig. 1.43: SCR with ammonia in diesel engines. (From [24].) DOC, diesel oxidation catalyst; DPF,
diesel particulate filter; AMOX, ammonia oxidation catalyst.

Fuel Fuel

Air Engine DPF SCR

Fig. 1.44: Hydrocarbons SCR.

sulfur are quite high being orders of magnitude higher than from on-road diesel en-
gines. A substantial cut to 0.50% m/m will apply after 1 January 2020. An even
stricter limit of 0.10% m/m is already applied in emission control areas such as
Baltic Sea area, the North Sea area, the North American area, covering designated
42 1 The basics

Large fuel penalty 9–10%

Fig. 1.45: Laboratory-scale installation at Turku University of Applied Sciences for HC SCR using a
monolithic block with a silver catalyst.

SCR system Silencer

Engine Exhaust pipe

Fig. 1.46: Urea SCR for marine applications (courtesy of Prof. L. Pettersson, KTH).

coastal areas of the United States and Canada, the United States Caribbean Sea area
around Puerto Rico and the United States Virgin Islands.
Heterogeneous catalysts, such as platinum group metals, also have numerous
applications in the removal of volatile organic compounds and hazardous air pollu-
tants from the chemical and pharmaceutical manufacturing processes. Monoliths
are applied in addition to pellets, which require more volume than monoliths and
 1.3 Catalysis in industry and for environmental protection 43

have higher pressure drop. Pt and Pd can be used for CO oxidation, and combina-
tions of Pt, Pd, Rh and Ir are applied while in the presence of CO and hydrocarbons
various. In special cases of certain pollutants, such as halogenated organics, chro-
mium containing catalysts can be used.
An interesting example of hazardous air pollutant removal is ozone abatement
in high-flying commercial aircrafts [25]. Intake air fed to the passenger cabin above
12,000 m altitudes can contain up to ca. 1–4 ppm of ozone, which can cause head-
ache, chest pain as well as irritations of the eye, nose and throat. This altitude is
typically used because air resistance is decreased upon elevation of altitude dimin-
ishing the fuel consumption. Removal of ozone to the level of 0.1 vppm is done
using 1% Pd on alumina supported in a monolith at 100–160 °C (Fig. 1.47).

Pd+/AI2O3//monolith (160 °c)

2O3 3O2
1–4 ppm 0.1 ppm
O3 O3
In upper Passenger
atmosphere cabin

Fig. 1.47: Pd catalyst for ozone removal.

High space velocities of 200,000–500,000 h and low concentration of ozone require

efficient contact of ozone with the catalyst. Because of variable flows, pressure drop,
lightweight requirements to decrease fuel penalty, vibration and mechanic shock, ce-
ramic and more lately metallic monolith were applied. A monolith washcoat similar
to the automobile converter maximizes geometric surface area and mass transfer at
the same time allowing minimum pressure drop.
Catalyst deactivation is possible by contaminations of sulfur, phosphorus and
hydrocarbon compounds present in compressor lubricating oils, hydraulic fluids
and ground-level air. To eliminate odor from hydrocarbons entering the cabin be-
cause of long waits behind other aircrafts on the runaway, a small amount of plati-
num is added to the catalyst. Another source of contamination is deicing fluids,
which can enter the air intake system. Catalyst regeneration can be done by apply-
ing washing solutions after 10,000–20,000 flight hours.
In addition to the application of catalysts in various environmental and bio-
mass-related reactions and the use of CO2 as a feedstock, it can be predicted that
more selective catalysts will be developed for the production of basic and specialty
chemicals (for example, in oxidation processes). Moreover, catalysts will contribute
44 1 The basics

to the production of new functional polymers with unique properties as well as pro-
duction of fine chemicals, pharmaceuticals and consumer electronics.
In particular fine chemicals are usually produced in several reaction steps
using stoichiometric methods and result in high waste-to-desired product ratio, the
E-factor, proposed by Sheldon (Tab. 1.5) [26].

Tab. 1.5: The annual production of different chemicals and the amount of
formed by-products per amount of desired product.

Industry Production (t year–) By-products (kg kg–)

Oil refining – .–.

Bulk chemicals – > –
Fine chemicals – –
Pharmaceuticals – –

One of the reasons for high E factors and the use of classical synthetic methods in
the synthesis of fine chemicals and pharmaceuticals – besides the fact that syn-
thetic chemists are unfamiliar with heterogeneous catalysis – is that commercially
available general purpose heterogeneous catalysts do not possess the required se-
lectivity. Although selectivity of these catalysts, bearing metals of nanoparticle size,
could be enhanced by carefully adjusting the size and environment of metal nano-
particles, it obviously requires dedicated time-consuming work. When the priority
is in getting the product to the market as fast as possible, such process development
might not be seen as an important factor.
Atom efficiency is another metric used to evaluate how sustainable or “green”
the process is. It is defined as the molecular weight of the desired product divided
by the total molecular weight of all products. For stoichiometric oxidation of a sec-
ondary alcohol:

3C6 H5 − CHðOHÞ − CH3 + 2CrO3 + 3H2 SO4 ! 3C6 H5 − CO − CH3 + Cr2 ðSO4 Þ3 + 6H2 O

an atom efficiency is 42%, while the theoretical E-factor is 1.5.

If the same reaction is conducted catalytically:

C6 H5 − CHOH − CH3 + 1=2 O2 ! C6 H5 − CO − CH3 + H2 O

an atom efficiency is 87%, with the E-factor equal to zero if water is not considered
as a waste.
Various hydrogenations, carbonylation and selective oxidation reactions are of
note among 100% atom efficient processes.
 1.4 Fuel cells and electrocatalysis 45

1.4 Fuel cells and electrocatalysis

A fuel cell is an electrochemical cell, where typically the chemical energy of the fuel
(hydrogen) and an oxidant (oxygen) is converted to electricity in a continuous mode,
provided there is a supply of reactants. In hydrogen–oxygen fuel cell (Fig. 1.48), only
water as the by-product in the overall chemical reaction is generated.

Oxygen Fuel cell H2O (L/g)

Fuel (H2) H2+1/2 O2 → H2O Electricity

Fig. 1.48: Principle of a fuel cell as an electrochemical energy conversion device. (Reproduced with
permission from [27]).

Fuel cells are used for primary and backup power for different buildings in remote
areas, as well as for vehicles operating with fuel cells (e.g., automobiles, buses,
boats, motorcycles or submarines).
The Gibbs free energy, ΔGo, of the overall fuel cell reaction

H2 + 1=2 O2 ! H2 O, ΔGo = − 236kJ=mol

is related with the equivalent electric cell potential ΔEo by

ΔGo = − nFΔEo (1:4)

where F is the Faraday constant and n denotes the number of electrons exchanged
in the overall chemical process.
Fuel cells consist of an anode, a cathode and an electrolyte allowing hydrogen
ions to move from anode to cathode. Two coupled, however, spatially separated half-
cell redox catalytic reactions at the anode and cathode sides are presented in Fig. 1.49.
On the anode (often platinum) surface, dihydrogen is split into protons and
electrons in the hydrogen–oxygen reaction (HOR), while water is generated from
oxygen protons and electrons on the cathode side (often nickel) in the oxygen-
reduction reaction (ORR). Different types of ion conductors can be used, including
KOH, solid polymers, solid oxides (yttria, zirconia), phosphoric acid, lithium and
potassium carbonates. A typical fuel cell produces a voltage from 0.6 to 0.7 V at full
load. A schematic view on the processes in a fuel cell is given in Fig. 1.50.
Besides HOR and ORR relevant for fuel cells, Fig. 1.50 also shows the oxygen
evolution reaction and hydrogen evolution reaction.
Requirements for a certain cell current (jcell) imply that the anode potential
shifts more positive by ηact,HOR. A negative shift of the cathode potential by ηact,ORR
even further diminishes the observed cell potential V, which is thus smaller than
46 1 The basics

Electrical output power

Electron flow

e–
H2 O2
Eo

Chemical
input
power

Proton
flow
H+

Electrolyte
Anode Cathode

H2 –> 2H+ + 2e– 1/

2 O2 + 2H+ + 2e– –> H2O

Fig. 1.49: Principle of a hydrogen and oxygen fuel cell galvanic element. (Reproduced with
permission from [27]).

4 H+ + 4 e– 2 H2 O2 + 4 H+ + 4 e– 2 H2O

HOR OER
j/mA cm–2

jcell

V ƞact, ORR
0V
ƞact, HOR 1.23 V
–j
cell

HER ORR

Electrode potential/V

Fig. 1.50: Schematic origin-of-activation overpotentials in a hydrogen-oxygen fuel cell.

(Reproduced with permission from [27]).
 References 47

the difference between the standard potentials of the hydrogen electrode (0 V) and
the oxygen electrode (1.23 V).
While Pt-based metals or alloys are often considered as efficient catalysts for
ORR, there are several challenges in applying them, including agglomeration of
nanoparticles, Ostwald ripening, oxidation of noble metals and leaching of the
nonnoble component in alloy catalysts. Designing of alternatives to Pt catalysts, a
preferably cost-efficient, more stable and not containing precious metals, is an im-
portant research task.

References
[1] Berzelius, J.J. (1836) Quelques idees sur une nouvelle force agissant dans les combinaisons
des corps organiques, Ann. Chim. Phys. 61: 146
[2] Liebig, J. (1840) Die Organische Chemie in ihre Anwendung auf Agricultur und Physiologie.
Braunschweig.
[3] http://www.nobelprize.org/nobel_prizes/chemistry/laureates/1909/ostwaldlecture.html
[4] Arnaut, L.G., Formosinho, S.J., Burrows, H. (2006) Chemical Kinetics: From Molecular
Structure to Chemical Reactivity. Amsterdam, Elsevier.
[5] Laidler, K.J. (1987) Chemical Kinetics. New York: HarperCollins.
[6] Robson Wright M. (2004) Introduction to Chemical Kinetics. Wiley.
[7] Rothenberg G. (2008) Catalysis: Concepts and Green Applications. Chichester, Wiley-VCH,
Weinheim.
[8] Balandin, A.A. (1969) Modern state of the multiplet theory of heterogeneous catalysis. Adv.
Catal. Rel. Subj. 19: 1.
[9] Rostrup-Nielsen, J.R. (2012) Perspective of Industry on modeling catalysis. In: Deutschmann,
O., Ed. Modeling and Simulation of Heterogeneous Catalytic Reactions: From the Molecular
Process to the Technical System. Weinheim, Germany: Wiley-VCH, pp. 283–301.
[10] Lilja, J., Aumo, J., Salmi, T., Murzin, D. Yu., Mäki-Arvela, et al. (2002) Kinetics of esterification
of propanoic acids with methanol over a fibrous polymer-supported sulphonic acid catalyst.
Appl. Catal. 228: 253.
[11] Geus, J.W., van Dillen, A.J. (2008) Preparation of supported catalysts by deposition-
precipitation. In: Knözinger, H., Schueth, F., Weitkamp, J. Handbook of Heterogeneous
Catalysis. Weinheim, Germany: Wiley-VCH.
[12] Binder, A., Seipenbusch, M., Muhler, M., Kasper, G.J. (2009) Kinetics and particle size effects
in ethene hydrogenation over supported palladium catalysts at atmospheric pressure.
J. Catal. 268: 150.
[13] Moulijn, J.A., Makkee, Mi., van Diepen, A.E. (2011) Chemical Process Technology. Chichester,
Wiley.
[14] Thomas, J.M., Thomas, W.J. (1997) Principles and Practice of Heterogeneous Catalysis.
Weinheim, Wiley.
[15] Chorkendorff, I., Niemanstverdriet, J.W. (2003) Concepts of Modern Catalysis and Kinetics.
Weinheim, Germany: Wiley-VCH.
[16] Bond, G.C. (1962) Catalysis by Metals. London: Academic Press.
[17] Murzin, D.Yu., Kul’kova, N.V. (1995) Kinetics and mechanism of the liquid-phase
hydrogenation. Catal. Today 24: 35.
[18] Hagen, J. (2006) Industrial Catalysis: A Practical Approach, Weinheim, Germany: Wiley-VCH.
48 1 The basics

[19] TCGR’s Intelligence Report: Business shifts in the global catalytic process industries.
[20] Dutch National Research School, Future Perspectives in Catalysis, (2009) Available at: http://
www.vermeer.net/pub/communication/downloads/future-perspectives-in-cata.pdf
[21] http://image.cleveland.com/home/cleve-media/width600/img/opinion_impact/photo/cuya
hoga-river-a55214b544fc6cac.jpg
[22] Bartholomew, C.H., Farrauto, R.J. (2006) Fundamentals of Industrial Catalytic
Processes, Second Edition, Wiley.
[23] Klingstedt, F., Arve, K., Eränen, K., Murzin, D.Yu. (2006) Towards improved catalytic low-
temperature NOx removal in diesel powered vehicles. Accounts Chem. Res. 39:273–282.
[24] www.catalysts.basf.com/mobilesources.
[25] Farrauto, R.J., Armor, J.N. (2016) Moving from discovery to real applications for your catalyst.
Appl. Catal. B. Gen. 527: 182.
[26] Sheldon, R.A. (2007) The E factor: fifteen years on. Green Chem. 9:1273.
[27] Strasser, P. (2012) Fuel cells, in Chemical Energy Storage. Ed. R. Schlögl. Berlin: De Gruyter.
2 Engineering catalysts

2.1 Catalyst design

Heterogeneous catalysts, as discussed in Chapter 1, are nanomaterials and thus

their active sites should be designed on the nanometer level, which requires under-
standing of catalytic phenomena on the molecular level.
While at the reactor level bulk density (mass per unit volume of bulk catalyst,
packed density) is important, at the mesoscale level particle density (mass per unit
volume of pellet, apparent density) and true solid density (mass per unit volume of
solid) play a crucial role.
These latter parameters are related to catalyst porosity as catalytic materials
usually possess a well-developed pore structure (Fig. 2.1).
Fig. 2.1 illustrates different types of pores. Open pores, which could be intercon-
nected, transport (passing) or dead-end, are accessible while reactants cannot ac-
cess closed pores. Porosity here could be defined as the volume of pores per total
volume of the solid, including the pores. The following parameters are typically
used to characterize porosity of catalytic materials: specific surface area (total sur-
face area per mass of solid, m2 g–1), specific pore volume (total pore volume per
mass of solid, cm3 g–1), pore size (nm) and its distribution (which can be broad or
narrow, mono- or multi-modal).
For pore size the following classification, proposed by International Union of
Pure and Applied Chemistry (IUPAC) is commonly accepted (Fig. 2.2), even if it is a
bit awkward to refer to materials with a pore size of less than one nanometer (for
example, zeolites) as microporous materials.
Shaped catalysts particles (mm range) are applied industrially on a macro-
scopic level. Thus, catalysis engineering requires knowledge of catalytic materials,
their activity, stability, mechanical strength, porosity, preparation and characteriza-
tion, understanding of reaction mechanisms and kinetics, as well as catalyst deacti-
vation. Preparation and characterization of catalytic materials is described in
Chapter 2, while the principles of kinetic modeling, including a description of deac-
tivation, will be presented in Chapter 3.
In addition to intrinsic kinetics, transport phenomena and reactor engineering
constitute an essential part of catalytic engineering. This is related to the fact, that
under industrial conditions, the rates of chemical transformations are very often af-
fected by the rates of other processes, and particularly by mass transfer processes.
The main mass transfer effects in heterogeneously catalyzed processes are: internal
diffusion inside the porous catalyst particles/layers, external diffusion in the laminar
film surrounding the catalyst particles, and – in multiphase reactors – the gas-liquid
or liquid-liquid mass transfer resistance at the phase boundary. The description of
mass transfer effects and diffusion phenomena will be addressed in Chapter 3.

https://doi.org/10.1515/9783110614435-002
50 2 Engineering catalysts

Interconnected
(open)
Ink bottle Funnel Uniform
shaped shaped size

Dead end
Passing (open)
(open) Closed
(A) (B)

Fig. 2.1: Pore structure with different types of pores. A ([1]) and B ([2]).

Micropores Mesopores Macropores

Activated carbon, Mesoporous materials Ceramics

zeolites

2 nm 50 nm

Fig. 2.2: Classification of pores.

Finally, for the understanding of the events occurring at macroscopic or reactor

level (m scale) not only microscopic understanding (nm scale – active site) is re-
quired, which is mainly the goal of chemists working in the field of heterogeneous
catalysis, but also description at the mm (shaped catalyst pellet) and micron scale.

2.1.1 Being in shape

The shaping of catalysts and supports is a key step in the catalyst preparation pro-
cedure. The shape and size of the catalyst particles should promote catalytic activ-
ity, strengthen the particle resistance to crushing and abrasion, minimize the bed
pressure drop, lessen fabrication costs and distribute dust build-up uniformly.
While small particle size increases activity by minimizing the influence of internal
and external mass transfer, bed pressure drop (Fig. 2.3) increases. Thus, there is an
apparent contradiction between the desire to have small catalyst particles (less dif-
fusional length, higher activity) and to utilize large particles displaying lower pres-
sure drop.
 2.1 Catalyst design 51

Pressure
p2 Δp = p2 – p1

A Unreacted A
p1 p2
B + C + ...
Catalyst bed

Fig. 2.3: Pressure drop in fixed-bed reactors. Courtesy of Dr. E. Toukoniitty.

There are no precise guidelines about what should be the exact value of pressure
drop. It is decided on a case-by-case basis depending on a particular process technol-
ogy. As a rule of thumb, the size of catalyst particles in fixed beds is exceeding 1–2
mm to avoid high pressure drop, even if larger particles of 10–12 mm are also applied.
In addition to the size, the shape is important affecting the bed porosity ε. Bed poros-
ity for spheres, Raschig rings and cylindrical particles vary between 0.35 and 0.4, 0.5
and 0.8 depending on the wall thickness and 0.3 and 0.35, respectively.
Thus, the best operational catalysts have a shape and size that represents an
optimum economic trade-off. The requirements of the shape (Fig. 2.4) and size are

Sphere Cloverleaf extrudate Irregular granule

Ring extrudate Pellet Cylinder extrudate

Trilobe Metal gauze Monolith

Fig. 2.4: Various catalyst shapes.

52 2 Engineering catalysts

mainly driven by the type of reactor. For reactors with fixed beds (see Chapter 3),
relatively large particles are applied (several mm) to avoid pressure drop. For mov-
ing-bed reactors, spherical particles are preferred because they allow a smooth
flow. Catalyst powders of various sizes are utilized in slurry three-phase reactors
and in fluidized bed reactors, where mechanical stress is found because of colli-
sions between catalyst particles and with the reactor walls and formation of shear
force due to cavitations at high velocities. In the case of slurry reactors, resistance
to attrition is important, the size of the particles should allow easy filtration, while
the bulk density is defined by settling requirements when easy settling is required.
For fluidized bed reactors, attrition resistance is important, as well as particle size
distribution.
Compared to other shapes, spherical particles have low manufacturing costs,
but possess relatively high-pressure drop and large diffusional length, and are not
that common. Similarly, irregular granules, which are low surface area materials,
have applications limited to just a few reactions, for example, ammonia synthesis.
Extrudates of various profiling are very common (details of extrusion will be pro-
vided in Section 2.3.7.3).
Pressure drop can be regulated by making special types of extrudates, ranging
from cylindrical to rings to cloverleaf extrudates. In the particular case of natural
gas steam reforming (Chapter 4), many types of catalysts with different shapes are
available from catalyst manufacturers. Catalysts in the forms of rings, monoliths
with several holes, wagon wheels as well as some other complex geometrical
shapes (Fig. 4.61) afford a low diffusional path in addition to low pressure drop.
The mechanical stability typically deteriorates with an increase of complexity and
decrease of the wall thickness. Extrudates with a typical aspect ratio of length to
diameter of approximately 3–6 have poorer strength compared to pellets (tablets),
as the pellets possess good mechanical strength and a regular shape. This is a
very common type of catalysts used in many hydrogenation, dehydrogenation
and oxidation reactions.
Monoliths are mainly applied when high fluid flow rates are required (off-gas
cleaning for example), as they have low pressure drop.
Low residence time and low pressure drop is a feature of low surface area metal
gauzes, which are utilized in very few specific cases, such as very exothermal oxida-
tion of ammonia to NO, when longer residence times lead to excessive temperatures
and volatilization of the active catalytic phase.
Different shapes can be ranked in the following order according to their rela-
tive pressure drop: monolith<rings<pellets<extrudates<powder. It is also notable
that for fixed-bed applications pressure drop should be high enough to allow
even flow distribution across the bed, but not too high which would then lead to
compressing and recycling costs. In some instances egg-shell type of catalysts are
industrially applied to suppress internal mass transfer limitations (Fig. 2.5).
 2.1 Catalyst design 53

Uniform Egg-shell Egg-white Egg-yolk

Fig. 2.5: Various concentration profiles in catalyst particles: active phase.

Such egg-shell catalysts with larger grain sizes have the same or similar intrin-
sic activity as powders of a smaller size with a uniform distribution of the active
phase.
The choice of catalyst shapes is thus not straightforward and involves careful
considerations of hydrodynamics, heat and mass transfer limitations, potential
pressure drops, mechanical strength, thermal resistance to sintering and phase
transition, more efficient heat conductivity for strongly exo- and endothermic reac-
tions, as well as manufacturing methods and associated costs.
Moreover, negative effects of a noncatalytic phase (carriers, binders, rheology
improvers, lubricants, etc.) and solvents on catalytic behavior should be avoided.
As a rule of thumb for the same surface-to-volume ratio the manufacturing
costs are higher for complicated shapes (monoliths, wagon wheels, etc.) compared
to more simple ones (i.e., extrudated cylinders).
Catalysts development performed in academia in the laboratory scale is mainly
limited to powder catalysts with a size that ensures an absence of mass transfer lim-
itations. Catalyst testing is typically conducted in batch and flow reactors, although
the use of structured catalysts, such as monoliths, solid foams and fibers as well as
microstructured devices, is growing.
Cordierite is widely used as a support material for monoliths (2MgO × 2SiO2 ×
5Al2O3), especially where the operating temperature is below 1,200 °C. The surface
area of cordierite is low and the active metal cannot be deposited directly on the
monolith. Therefore it is first wash-coated, for example, with alumina, to increase the
surface area (Fig. 2.6A) resulting in a wash-coat thickness of 15–100 μm. Fig. 2.6B
shows a monolith block for a heavy duty vehicle that does not have any open pas-
sages between the channels.
As the thickness of the coating for squared channels is much higher than in
other parts (Fig. 2.6A), especially at the corners, other geometries are possible, al-
lowing more uniform wash-coating (Fig. 2.7).
The manufacturing of monolithic catalysts is inherently more costly than of
other shapes such as pellets or powders. The economic benefits of using monoliths
should be thus clearly demonstrated by exceeding higher catalyst costs and invest-
ments in research and development. In particular when the annual volume of each
catalyst is small, it is difficult to justify the dedicated research. In addition, cordier-
ite has some limitations in terms of durability when in contact with alkali and alka-
line-earth above 700 °C.
54 2 Engineering catalysts

Alumina
Silica
Carbon

(A) (B)

Fig. 2.6: Monoliths: (A) wash-coating principle and (B) a monolithic block for exhaust gas
abatement from heavy duty vehicles.

Fig. 2.7: Geometry of channels in monoliths.

Fig. 2.8 shows one of the operation modes of the liquid-phase applications, the
use of cordierite is limited to a pH range of roughly pH 4–9.
Other types of structured materials have been recently suggested for use as cat-
alysts for mainly academic purposes. Some of them are also produced commercially
and are used in few specific applications related to catalysts.
Similar to monoliths, catalytic foams (Fig. 2.9) require wash-coating, as the sur-
face area of the foams made of α-alumina with several additives is just few m2 g–1.
Foams, although less mechanically stable than metal or cordierite monoliths, have
at the same time several advantages. The pressure drop can be compared to packed
beds with a similar flow pattern. Because of the template preparation method (which
will be covered in Section 2.3) a desired shape can be produced to fit the reactor size.
Wash-coating is also required for spinning disk reactors, where the two disks are
co- or counter-rotating, with each at a different rotational speed (Fig. 2.10). Although
efficient mixing and heat transfer can be achieved in such reactors, advantages of
using them for heterogeneous catalytic reactions are yet to be demonstrated.
As a final example of different catalyst shapes, it is worth mentioning fiber cat-
alysts with polyethylene as a base material further modified by grafting different
functional groups (pyridine, amines, carboxylic and sulfonic acid, or combination
 2.1 Catalyst design 55

Impeller

Monoliths

(A) (B) Sampling valve

Fig. 2.8: A screw impeller reactor housing 12 pieces of structures catalysts. (A) Monolithic blocks
(left) and a corrugated metallic monolith (right). (B) General principle of operation.

Fig. 2.9: Foam blocks, made of α-alumina, require wash-coating and a subsequent introduction of
the active catalytic phase.

of them). Such materials are produced commercially and could be used as acid
catalysts similar to ion exchange resins or can be utilized in metal scavenging. For
catalytic reactions requiring a metal function, such metals can be immobilized on
the ion exchange fiber (Fig. 2.11) with a subsequent reduction with hydrogen or a
chemical reducing agent.
The fiber based catalysts can be produced in several shapes, with the diameter
of the fiber ranging from 10 to 50 μm, ensuring a short diffusional path. The swell-
ing of such fibers in many solvents and low mechanical strength are a disadvan-
tage. This is significantly improved in some activated carbon cloth materials, which
have very high specific surface area (above 1,500 m2 g–1), are mechanically rather
strong and can be used (after deposition of the catalytic phase) in slurry and fixed-
bed reactors (Fig. 2.12).
56 2 Engineering catalysts

Spinning disks reactor

Heat transfer fluid

Products

Light reactant Heavy reactant

(gas/liquid) (gas/liquid) High gravity
High shear

Products

Fig. 2.10: Spinning disk reactor. (From [3]).

Functional
group

Metal

Functional
group
SEM micrograph of the catalyst fibers
fiber diameter = 10 μm, fiber length = 4,000 μm

Fig. 2.11: Polyethylene fiber catalyst.

(A) (B)

Fig. 2.12: Active carbon cloth. (A) General appearance. (B) Application in a fixed-bed reactor with
several layers of carbon cloths.
 2.1 Catalyst design 57

2.1.2 Catalysis informatics and high-throughput experimentation

In the recent years, catalysis informatics and high-throughput experimentation

became an essential part of catalyst development [4], enabling optimization of
operating conditions and discovery of new catalyst materials. Data mining in ca-
talysis informatics should help in predicting the required catalyst composition
and structure, allowing to achieve the required catalytic performance. To this
end analysis of available data on the catalytic reactions (activity, selectivity, sta-
bility, reaction parameters) as well as surface and bulk characterization should
be done with appropriate statistical tools and physical models to integrate this
information.
Catalytic informatics is indispensable in high-throughput combinatorial studies
that generate enormous amount of data. As an example, it can be mentioned that a
catalyst with five transition-metal components can give over 150 million possible
compositions.
On top of that not only the elemental composition should be considered but
other parameters as well, such as the precursor type, preparation (impregnation,
deposition, ion exchange, etc.) and post-treatment methods (calcination, reduc-
tion), catalyst shape and reaction parameters to name a few [5].
High-throughput experimentations in heterogeneous catalysis involve high-
throughput synthesis, characterization and screening techniques using robotics
and special software.
High-throughput qualitative or semiquantitative primary screening focused on
discovery is typically performed on small samples and can be often done in uncon-
ventional reactors. During primary screening, both false positives and especially
false negatives should be minimized. Evolutionary strategies in high-throughput
development of heterogeneous catalysts are based on creating new catalysts from
the best catalysts of the previous generation by qualitative (replacing one or more
chemical elements by others) and quantitative mutations (changing concentration
of one or more compounds), as well as crossover (exchanging components of two
materials of the previous generation). Selection of the best catalysts is based on the
position of catalysts according to ranking.
For primary screening, one of the applied options is preparation of solid cata-
lysts as thin-film spots on the surface of wafer substrates allowing deposition of
hundreds to thousands of small spots of differently composed materials on a sup-
port. Such approach involves a limited number of preparation steps without, how-
ever, a guarantee that the catalysts would have the same structure as bulk
catalysts because of small amounts and potential interactions with the support.
Obviously, hydrodynamics as well as mass and heat transfer are different from
lab-scale reactors.
An interesting option in the high-throughput catalysis is to use single pellets
that might, but not necessarily, behave like pellets in an industrial catalyst bed.
58 2 Engineering catalysts

The next level is secondary screening when the so-called primary “hits” are
confirmed and optimized using reactors, catalyst shapes and analytics resembling
real-life applications.
Table 2.1 gives a comparison between these stages of screening.

Tab. 2.1: Comparison between stages 1 and 2 in combinatorial screening (modified from [6]).

Stage  Stage 

Throughput Very high (up to , experiments Moderate (up to  experiments per day)
per day)

Synthesis Methods from inorganic chemistry, Methods similar to industrial preparation,

small amounts (micrograms) large amounts (.– g)

Testing Fast testing, conditions different from Lab reactors, considerations of

lab reactors hydrodynamics, heat and mass transfer

Analytics Dedicated, qualitative and Quantitative (conversion, selectivity),

semiquantitative generic

Application Discovery of novel materials Improvement of known systems,

investigation of the influence of
promoters, kinetics and deactivation

Conventional fixed-bed microreactors with complete analysis of reactants and prod-

ucts and evaluation of the mass balance closure can be used in the tertiary screen-
ing phase. Alternatively, piloting can proceed directly from secondary screening.
Preparation of industrial catalysts involves shaping to obtain granulates, beads
or pellets as will be discussed later. Such parallel shaping of catalyst powders would
be desirable on commercial synthesis platforms not yet being available. Industrial
preparation should be also robust, easy to operate and control, cost-efficient in terms
of precursors and ecologically friendly avoiding toxic and harmful waste.
Machine learning can assist high-throughput screening efforts by finding corre-
lations between structure and performance [7]. For analysis of macroscopic cataly-
sis data, partial least squares, principal component analysis and artificial neural
networks (ANN) are used [8, 9]. These and other statistical classification and regres-
sion models help in extracting information from the experimental data, being lim-
ited in their ability to generalize beyond the input data. For ANN, the training set
must be representative of the test set. After a proper training, in the subsequent ex-
ecution phase the ANN is fed with new cases providing an output, which should be
properly validated.
A particular challenge in heterogeneous catalysis is finding adequate descriptors
because many parameters in catalyst preparation (cluster size, phases, synthesis
 2.2 Toolbox in catalysis 59

conditions, pre- and post-treatment, etc.) can influence activity, selectivity and stabil-
ity beyond just chemical composition.
Quantum chemical modeling can be utilized to generate larger datasets than in
experiments and complement experimental data, followed by training machine
learning models.
Each catalyst comprising a dataset of different catalysts is described based on
the important physicochemical properties (electronic structures, physical and
atomic properties) defining each material. Thereafter, machine learning tools can
be used to find proper descriptors for stability, activity and selectivity.
More details on the theoretical background and quantum chemical calcula-
tions in catalysis will be given in Section 2.2.17. Here, it can be mentioned that
the energy of the d-band center with respect to the Fermi level is an often used
descriptor, as it is connected to the interactions between adsorbate valence states
and the d-states of a transition metal surface. Thus, after correlating the band
center with adsorption energies, the next step is to relate the adsorption energy
of the relevant reactants with catalyst activity through linear scaling relations.
Computation of the d-band center typically requires quantum mechanical (QM)
calculations at the DFT (density functional theory) level.
QM calculations are computationally expensive, thus being limited to small cat-
alytic systems. Development of interatomic potentials (i.e., mathematical functions
for computing the potential energy of a system of atoms) with machine learning
after training with data generated by QM can substantially accelerate simulations,
keeping accuracy comparable with QM methods. Figure 2.13 describes a workflow
in catalyst design involving machine learning.
As a conclusion, machine learning combined with computational modeling and
experiments is an emerging path in knowledge-based development of heteroge-
neous catalysts.

2.2 Toolbox in catalysis

2.2.1 General overview of the characterization methods

Catalyst characterization is required for understanding relationships among physi-

cal, chemical and catalytic properties, i.e., relating catalyst structure and function,
as well as elucidating catalysts deactivation, designing procedures for regeneration
and choosing catalyst properties to minimize such deactivation. Characterization
includes the determination of physical properties such as pore size, surface area,
particle strength and morphology of the carrier and location of active species within
the carrier, elemental and molecular surface composition, elucidation of the nature
of adsorbed species, active phase – support and adsorbate – adsorbate interactions,
to name a few. Monitoring changes in the physical and chemical properties of the
60 2 Engineering catalysts

Ab initio Data
𝜕
iħ—Ψ = ĤΨ
𝜕t

Machine-learned Catalyst synthesis

potential
Potential
functions

Catalyst design

Catalyst structure Catalyst prediction

Less
Active
Active

Mechanism Microkinetic
simulation
N
j M i
ri=∑⟮kj V i ∏Ck VK⟯
j=1 k=1

Fig. 2.13: Approaches to catalyst design involving machine learning (Reproduced with permission
from [10]).

catalyst during preparation, activation and reaction stages is important for proper-
ties control. Each tool has advantages and limitations and several tools are typically
needed to provide complimentary data.
In order to get information about the structure surface species various vibrational
spectroscopies, such as infrared (IR) spectroscopy, can be applied, which although
being of low sensitivity, can be arranged in situ. High-resolution electron energy loss
spectroscopy (HREELS), Raman and sum frequency generation (SFG) are among the
other methods used. Electron spectroscopy of chemical analysis or X-ray photoelec-
tron spectroscopy (XPS) and Auger spectroscopy is mainly applied to the study of the
surfaces, but can also give information about the adsorbed species.
IR spectroscopy is used also to study the nature of surfaces, either measuring
the spectra of solid compounds directly (for example, investigating acidic OH
groups) or by utilizing probe molecules, for which important information about the
 2.2 Toolbox in catalysis 61

number of acid sites and their type (Lewis, Brønsted) can be obtained. Electron mi-
croscopy in various forms, XPS, and EXAFS (extended X-ray fine structure), are
among methods widely employed to characterize surface structures (Fig. 2.14),
while X-ray diffraction is a part of the characterization toolbox aimed at studying
the bulk of the material.

IR photons X-rays
IR spectroscopy X-ray diffraction

IR photons X-rays
Catalyst

Electrons Electrons

SEM/TEM XPS
Electrons X-rays

Fig. 2.14: Input–output structures of some analytical methods.

In electron microscopy the overall contrast is caused by the different adsorption of

photons or diffraction phenomena. Transmission electron microscopy (TEM, Fig. 2.14)
is suitable for the examination of supported particles in the range down to few nano-
meters. Scanning electron microscopy (SEM) provides topographical images formed
from backscattered primary or low energy secondary electrons. An energy dispersive
X-ray spectroscopy is often attached to an SEM instrument. Electron spectroscopy of
chemical analysis (ESCA) or photoelectron spectroscopy (XPS) is a surface sensitive
technique, with the depth of the sample being in the range of – two to 10 atom layers.
Surface science methods like XPS require an ultra-high vacuum, creating what
is often called a pressure gap. Such conditions of high vacuum are necessary, as
they provide a clear path for the emitted electrons (or other particles) and prevent
collisions between the electrons and gas molecules. The average distance molecules
travel between collisions is inversely proportional to pressure.
More detailed information about EXAFS, nuclear magnetic resonance (NMR),
electron spin resonance (ESR) and Mössbauer spectroscopy will be provided in the
corresponding sections.
As this textbook is written for chemical engineers, the description of the
methods is uneven, in some instances being very basic and focusing more on the
application and interpretation side of the methods, rather than on the detailed
physics behind them. The limitations of the methods are specifically addressed,
as this information is of real importance for chemical engineers, who are not typi-
cally experts in physical methods of catalyst characterization.
Finally, a few words about classical elemental analysis. A catalyst is dissolved
in strong acids under (for example) microwave heating, followed by analysis with
62 2 Engineering catalysts

inductive coupled plasma-MS (ICP-MS, when a sample solution is introduced into

the core of inductively coupled argon plasma, which generates temperature of ap-
proximately 8,000 °C) or atomic absorption spectroscopy (AAS). In this way, essen-
tial information about the catalyst bulk composition can be obtained as it is not
related directly to surface properties of the catalyst.

One example from my industrial experience is worth mentioning in the context of elemental anal-
ysis. Once, when making such analysis of a dehydrogenation catalyst for an industrial client, a
high sulfur content was unexpectedly observed. It turned out that the purchase department of
that company (without informing the plant operator) acquired a large quantity of cheap coal-
derived benzene, instead of the usual petroleum-based one. As a consequence, sulfur was able
to penetrate through two upstream reactors (one of them with another catalyst), finally resting on
the dehydrogenation catalyst. Small savings in the feedstock price resulted in large spending for
the catalysts, which had to be replaced earlier than scheduled because of deactivation.

2.2.2 Adsorption methods

Adsorption methods provide information about the total surface area of the catalyst
and porosity (by physisorption measurements) and the number, type and nature of
adsorbed species (by chemisorption). Temperature-programmed methods (such as
desorption) are also extremely useful, for example, in investigating heterogeneity of
surfaces. Chemisorption and thermodesorption methods can be combined with
microcalorimetric measurements, providing information about heat of adsorption,
acidity distribution, surface heterogeneity and lateral interactions between adsorbed
species.
Either static or dynamic methods could be applied in studying adsorption. In
static methods adsorption is performed gravimetrically or volumetrically. The volu-
metric method uses two chambers, one a dosing section, while a sample is located
in the other one. The precision of the analysis depends on the accurate determina-
tion of the volumes, including the dead space in the sample chamber, careful tem-
perature and pressure measurements. In gravimetric measurements there is no
need for such precise calibration of volumes, as the quantity of adsorbed gas is
measured directly using a microbalance.
Dynamic methods can be arranged in single flow or pulse flow modes. The first
option relies on a flow of a gas containing the molecules to be adsorbed through a
catalytic bed. Although no vacuum system is required, making the method easy to
use, the single flow method is used much less than the pulse technique because of
the possibility of slow adsorption and a need to employ pure gases. In the pulse tech-
nique there is a sequence of pulses going through the sample until no further adsorp-
tion is seen. The amount of adsorbed molecules, which should be rather strongly
bound to the surface, is determined by summation of the gases adsorbed in all pulses.
 2.2 Toolbox in catalysis 63

2.2.3 Physisorption methods

Low-temperature adsorption at a gas condensation temperature (i.e., for nitrogen

this temperature is 77 K) is generally accepted as a standard procedure for the deter-
mination of surface areas and pore size distributions of a wide range of porous ma-
terials. Physisorption measurements performed at a constant temperature allow the
determination of an adsorption isotherm, whose features provide important infor-
mation about the porous solid. The theory of nitrogen adsorption that was devel-
oped in the late 1930s by Brunauer, Emmett and Teller (thus the well-known name
of the BET method and BET theory) has substantial shortcoming because of the as-
sumptions involved. Moreover, it does not include capillary condensation in the
pores in the mathematical treatment, leading to large deviations between experi-
mental data and model predictions at higher values of relative pressures. With clear
understanding of all the flaws of the BET theory, the BET method is still very widely
used throughout the world to calculate the monolayer capacity, although some
other methods are also applied. Pore size distribution can be calculated from the
desorption branch of the isotherm. Evaporation from the mesopores is not as easy
as condensation, resulting in the differences between adsorption and desorption
branches, or in other words, hysteresis. By examining the isotherm step-by-step in
the domain of hysteresis, the mesopore volume and the mesopore size distribution
can be obtained.
The monograph by J.B. Condon [11] is an excellent source of information on var-
ious methods for measuring and calculating surface area and porosity.

2.2.3.1 Surface area determination

Surface area measurements can be performed with commercially available instru-
ments (see Fig. 2.15).
The following procedure could be applied for taking measurements (Fig. 2.15B).
The lines (V1) are filled with a known pressure of nitrogen (starting from as
low as 10–20 Torr up to 700–800 Torr). Thereafter, the valve to the flask with
the known quantity of solid material (V2) is open and the pressure (pʹ) is sta-
bilized. As plineV1 = pʹline (Vl + V2) + RTnadsorbed moles the amount of adsorbed
nitrogen can be calculated.
Different types of isotherms are obtained, depending on the type of the support
material (Fig. 2.16).
In a type I isotherm there is a monolayer adsorption on an ideal surface with sat-
uration and there are no interactions between adsorbed species. In type I, as well as
in types II and III, there is no hysteresis, showing that these materials are free from
mesopores. For a type II isotherm, multilayers starting at point B are formed. For a
type III isotherm, the incoming molecules are bound stronger to the surface when it
is gradually occupied, which is a sign of lateral interactions. Isotherm VI represents a
64 2 Engineering catalysts

High vacuum

P V1
Pressure
gauge
V2

(A) (B) Adsorbate Adsorbent

Fig. 2.15: (A) Instrument for sorption measurements (Carlo Erba Sorptometer 1900). (B) Principle
scheme of measurements.

I II III
nad

nad

nad

B
p/p0 p/p0 p/p0

IV V VI
nad

nad

nad

B
p/p0 p/p0 p/p0

Fig. 2.16: IUPAC classification of isotherms.

layer-by-layer adsorption. Types IV (the most typical for catalytic materials) and V
isotherms exhibit hysteresis in mesoporous. The formation of poly-layers and then
capillary condensation also occur during adsorption, but there are different equa-
tions governing evaporation and condensation during adsorption and desorption
(as will be explained later), thus for the same amount adsorbed there is a difference
in relative pressures observed in the gas-phase.
The hysteresis behavior could be even more complicated, as shown in Fig. 2.17A,
where the H1 type of isotherm corresponds to a narrow distribution of uniform pores,
while the H2 type reflects more complex (non-uniform) pore structures. The shape of
pores in H1 and H2 isotherms are nearly cylindrical, while slit-shaped pores with uni-
form and non-uniform sizes and shapes result in H4 and H3 isotherms, respectively.
The specific surface area can be calculated when knowing the monolayer
capacity, using the following expression:
 2.2 Toolbox in catalysis 65

Type H1 Type H2 Type H3 Type H4

Adsorbed volume

Adsorbed volume

Adsorbed volume

Adsorbed volume
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
p/p0 p/p0 p/p0 p/p0

Fig. 2.17: Hysteresis behavior for different types of pores: adsorbed amount as a function of
relative pressure.

s = nm Am NA (2:1)

where s is the specific surface area (m2 kg−1 or typically m2 g−1), mm is the mono-
layer capacity (mol g−1), Am is the area occupied by one molecule (m2 molecule−1)
and NA is the Avogadro number equal to 6.023 × 1023 molecules mol−1. Tab. 2.2 con-
tains values of the areas occupied by common adsorbates. Typically, nitrogen is
used, however, krypton and argon adsorptions are also used, especially for low sur-
face area materials (< 1 m2 g−1).

Tab. 2.2: Values of the areas occupied by typical adsorbates.

Adsorbate Boiling point (K) Am (nm/molecule)

N . .
Ar . .
CO . .
Kr . .

In practice, the adsorbed volume is measured and the volumetric monolayer capac-
ity (recalculated for standard pressure and temperature) is reported. Thus, instead
of eq. (2.1) we get:

Vm Am NA
s= (2:2)
V

where Vm is the volumetric monolayer capacity, V is the volume of one mole at stan-
dard pressure and temperature (22.4 × 10−3 m3 mol−1).
There is, therefore, a need to calculate the value of the monolayer capacity from
experimental data for physisorption when multilayers are formed. The monolayer ca-
pacity can be calculated using the method developed by Brunauer, Emmett and
66 2 Engineering catalysts

Teller for multilayer adsorption, which was based on the approach of Langmuir, who
considered adsorption in a monolayer.
Layers of adsorbed molecules (Fig. 2.18) are divided in the first layer by the heat
of adsorption ΔHad,1 and in the second and subsequent layers by ΔHad,2 = ΔHcond’
equal to the heat of condensation.

Fig. 2.18: Schematics of desorbed layers.

Adsorption in the first layer can be described as a0PS0 = b1S1 where a0 and b1 are
adsorption coefficient on the zero layer (bare surface) and desorption coefficient
from the first layer, respectively, S0 denotes the number of empty sites available
for adsorption and S1 is the number of adsorbed molecules in the first layers.
Consequently S1 = (a0P/b1)S0 = yS0 with y = (a0P/b1). Analogously for the second
layer a1PS1 = b2S2, S2 = (a1P/b2)S1 = xyS0 with x = a1P/b2. As only adsorption in
the first layer is different: a1 = a2 = a3 = …. = an; b2 = b3 = …. = bn and subse-
quently a2PS2 = b3S3. S3 = (a2P/b3)S2 = x2yS0. For adsorption in the nth layer:
Sn = xn – 1yS0 = xn(y/x)S0 = xncS0 with c = y/x. The total amount of adsorbed mole-
P
∞ P∞ P
∞
cules is defined as nSn = ncxn S0 = cS0 nxn .
n=1 n=1 n=1
P∞
The monolayer capacity should also include the empty sites Sn = S0 +
P∞ P∞ n = 0P∞
cx S0 = S0 + cS0
n
x . Making use of infinite geometrical progression
n
nxn )
n=1 n=1 n=1
P∞
x
2 xn ) x ′ the equation for the monolayer capacity can be easily obtained:
ð1 − xÞ 1−x
n=1
x
cS0
Nads Vads ð1 − xÞ2 cx
= = = (2:3)
Nmono Vmono S + cS x ð1 − xÞð1 − x + cxÞ
0 0
1−x
As adsorption-desorption in the nth layers can be considered as liquid-vapor equi-
librium, condensation-evaporation equilibrium gives aP0 = b where P0 is the satura-
tion pressure. For adsorption of nitrogen P0 = 1 bar at 77 K. Further transformations
give x = (a/b)P = P/P0.
Linearization of eq. (2.3) results in:
1 x 1 c−1
= + x (2:4)
V 1 − x cVm cVm
 2.2 Toolbox in catalysis 67

which can be presented in the graphical form (see Fig. 2.19), allowing for the calcu-
lation of the monolayer volume:
1
Vm = (2:5)
i+s

1 x
V 1–x
c –1
=s
cVm

1
= Vm
1 i+s
=i
cVm

Fig. 2.19: Linearization of BET equation.

In eq. (2.5) the slope is s = (c – 1)/cVm and the intercept is given by 1/cVm, allowing
for the calculation of the monolayer volume as the reciprocal value of the sum of
the slope and the intercept.
The linearization should be performed for the relative pressures between 0.05
and 0.3, as at higher values the deviations are becoming substantial because of var-
ious shortcomings in the model; for example, not accounting for capillary conden-
sation in the pores.
By taking the volume of the monolayer from eq. (2.2) it is possible to calculate
the specific surface area. Typically the plots for catalytic materials are presented
with the adsorbed volume calculated in cm3 g–1 at standard temperature and pres-
sure (STP). For N2 adsorption at 77 K the specific surface area (m2 g−1) is then the
monolayer volume (cm3 g−1 at STP) multiplied by coefficient 4.3532. Some typical
values for catalytic materials are given in Tab. 2.3. While reporting the surface areas
of catalytic materials one should be cautious, keeping in mind the precisions of vol-
umetric flow measurements. Reporting the values that are directly available from a
physisorption instrument, for example, 124.564 m2 g−1 does not make much sense
even if similar values appear on a regular basis in the literature. In addition, errors
in the estimation of surface areas for materials with low surface areas can be very
high, thus it is recommended that there is a sufficient amount of the sample in a
measurement unit (~40 m2).
Tab. 2.3 also contains values of the mean diameter of pores. For physisorption
of nitrogen the monolayer thickness is 0.354 nm, which means that in the case of
zeolites or activated carbon applications of BET theory, relying on an infinite
68 2 Engineering catalysts

Tab. 2.3: Typical values of surface areas for catalyst supports.

Catalyst supports Mean dp (nm) SBET (m g−)

Silica gel  

 
 
γ-AlO  
 
Zeolite .– –
Activated carbon  –,
TiO – –
Aerosil SiO – –

number of adsorbed layers is more than questionable. Even for supports like alu-
mina, a pore diameter of ~5 nm is translated in around seven layers. As the BET
method is only valid in a small pressure interval, and the physical interpretation of
it is not very easy, other less popular methods are also used for the calculation of
the monolayer capacity.
In the t-plot method, introduced by de Boer, it is assumed that in a certain iso-
therm region the micropores are already filled-up, whereas adsorption in larger pores
occurs according to some simple equation, characteristic for a large class of solids.
This equation should approximate adsorption in mesopores, macropores and on a
flat surface in a narrow pressure range just above complete filling in micropores, but
below vapor condensation in mesopores. Such an assumption may be valid, if micro-
pores are small and super-micropores are not present. Thus the thickness of the ad-
sorbed layer depends only on the relative pressure but not on the material type and
could be expressed, for example, by the equations in Fig. 2.20.
Adsorption within this pressure region, when the thickness is independent on
the material type, can be described by a simple linear dependence (Fig. 2.21).
Such linear dependence can be easily understood by expressing the thickness
of the adsorbed layer through the number of adsorbed monolayers (the ratio of the
number of adsorbed moles nad to the monolayer capacity nm) and the monolayer
thickness (0.354 nm or 3.5410−10 m for nitrogen):
nad
t= 0.354 nm (2:6)
nm

And subsequently:
nad
St = nm Am NA = 0.354 × 10 − 9 ðmÞAm NA (2:7)
t

From eq. (2.7) the number of adsorbed moles depends linearly on the thickness of the
adsorbed layer with the slope inversely proportional to the surface area of mesopores.
 2.2 Toolbox in catalysis 69

1.00
0.333
Thickness of adsorbed

–5.00
t = 0.354
ln(p/p0)
layer t (nm)

Halsey

0.5
13,99
t = 0.1
0.034–log(p/p0)
Harkins–Jura–de Boer
0.00
0.00 0.50 1.00
p/p0

Fig. 2.20: Adsorbed layer thickness as a function of relative pressure.

nad

Smesopores

Fig. 2.21: Adsorbed amount as a function of

adsorbed layer thickness for a material with micro-
t and mesopores.

A dependence like the one presented in Fig. 2.21 is made through plotting an
experimentally obtained value of nad or Vad for a particular value of the relative
pressure and relating it to the thickness of the adsorbed layer for the same pressure
using correlations of t vs relative pressure.

Exercise 2.1: When nitrogen was adsorbed on a catalyst at −196 °C the following data were ob-
tained. Numerical values are given below. What is the specific surface area of the catalyst?
Calculate the pore size distribution.

Adsorption Desorption

Rel. Volume, Rel. Press, Volume, Rel. Press, Volume,

Press, adsorbed (p/p) adsorbed, (p/p) desorbed,
(p/p) (cm g−) (cm g−) (cm g−)
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
70 2 Engineering catalysts

Adsorption Desorption

Rel. Volume, Rel. Press, Volume, Rel. Press, Volume,

Press, adsorbed (p/p) adsorbed, (p/p) desorbed,
(p/p) (cm g−) (cm g−) (cm g−)

. . . . . .

. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . .
. . . .
. .  .

500
450
400
Adsorbed volume

350
300
250
200
150
100
50
0
0.0 0.2 0.4 0.6 0.8 1.0
P/P0

Fig. E2.1. Adsorbed volume as a function of relative pressure.

For microporous materials like zeolites and active carbon with the pore size ~0.5–2
nm there are no theoretical grounds to use the BET method with the assumption of
multilayer adsorption. An ideal Langmuir equation usually overestimates the surface
area. A t-plot method could be used instead for the determination of the micropore
volume and the surface area.
For microporous materials a plot of the adsorbed volume vs the thickness of the
adsorbed layer, defined by Harkins-Jura-de Boer equation (Fig. 2.20), gives a
straight line (Fig. 2.22A). Therefore even a one point analysis at a certain value of
 2.2 Toolbox in catalysis 71

W VA

t-plot

Micropore volume

1.00 1.00
(A) p/p0 t (Å)

W Micro+meso VA Slope = SA of meso

+ external surface

t-plot
Full up of micropore

Slope = total SA

1.00 1.00
(B) p/p0 t (Å)

Fig. 2.22: Application of the t-plot method for determination of micropore volume.

relative pressure is sufficient, as at this point the surface is determined by the ratio
between the volume adsorbed and the layer thickness at this particular pressure.
Fig. 2.22B provides an illustration of the t-plot method for materials containing both
micro- and mesopores.
Instead of the t-plot method for the determination of the micropore volume, a
method developed by Dubinin and his co-workers [12] could be applied. The
Dubinin–Radushkevich equation relates the volume of adsorbate and the total
volume of the micropores Vm:
"  2  #
RT p0
V = Vm exp − B ln2 (2:8)
β p

where B and β are constants and T is temperature. A linear relationship should be

found between ln V and ln2(p0/p). Although the curve is bending at low values of
log2(p0/p) extrapolation to zero gives the value of Vm.

2.2.3.2 Determination of pore size and pore size distribution

In the capillary condensation region (p/p0 > 0.4) there is capillary condensation in
the pores during both adsorption and desorption (Fig. 2.23). Evaporation from the
72 2 Engineering catalysts

Adsorption followed by Capillary evaporation

capillary condensation parallel with desorption

Fig. 2.23: Representation of pore filling during adsorption and desorption. Adsorption proceeds
through multilayer formation, while desorption is meniscus controlled.

pores is not as easy as condensation, resulting in differences between adsorption

and desorption branches (hysteresis).
In a cylindrical pore with two radii of curvature, the saturation vapor pressure
is given by the Kelvin equation:
 
p0 1 1 σV cos θ
ln = + (2:9)
p r1 r2 RT

where p0 is saturation pressure, σ is the surface tension of the liquid (8.72 mN m−1
for N2), V is the molar volume of liquid (N2 = 34.6 cm3 mol−1), θ is the contact angle
of the liquid with the gas (zero for liquid nitrogen). Equation (2.9) shows the result
of the interactions with the pore walls making condensation occur at a lower rela-
tive pressure than the saturation one. As the relative pressure is increased, conden-
sation occurs first in the pores of smaller radii and then progresses. The Kelvin
equation gives the possibility of estimating pore radii from a gas adsorption iso-
therm. At any equilibrium pressure p, pores of radii less than r will be filled with
condensed vapor. Application of the Kelvin equation to all points of the isotherm at
relative pressures greater than that corresponding to the monolayer volume (where
capillary condensation begins to occur) yields information concerning the volume
of gas adsorbed in pores of different radii.
Desorption occurs from hemispherical surfaces with both radii of curvature
equal to each other, thus eq. (2.9) is rearranged:

2σV cos θ
r= p +δ (2:10)
RT ln p0

The modified Kelvin equation (eq. 2.10) is the basis of Barrett, Joyner, Halenda and
Dollimore-Heal methods and contains δ which is the thickness of the adsorbed
layer and can be calculated from the Wheeler equation:
  1
p0 − 3
δ = 0.734 ln ðnmÞ (2:11)
p
 2.2 Toolbox in catalysis 73

Finally an expression is obtained for the estimation of the pore radii from the nitro-
gen desorption branch of isotherm:
  1
0.93 p0 − 3
r = p + 0.734 ln ðnmÞ (2:12)
ln
0 p
p

The following procedure is used to calculate the pore size and pore volume distribu-
tion. The desorption branch of the experimentally measured nitrogen physisorption
is applied. Each point gives the amount adsorbed and the relative pressure. By mul-
tiplying this amount by the molar volume of liquid nitrogen (34.68 cm3 mol−1) the
pore volume is obtained. The modified Kelvin equation gives the radius at a particu-
lar value of relative pressure. The next step is to plot the volume of pores vs radius
and then calculate dV/dr at different values of r. This can be done for example, by
dividing the incremental difference in volumes (V2 – V1) by the incremental differ-
ence in corresponding radii (r2 – r1) and relating (V2 – V1)/(r2 – r1) to the average
between these two radii [(r1 + r1)/2]. A typical result is presented in Fig. 2.24.

1.2
Pore volume (cm3/g/nm)

0.9

0.6 5% Ru-10% Sn/SiO2

0.3

0
0 1 2 3 4
Pore radius (nm)

Fig. 2.24: Pore volume distribution as a function of pore radii.

Besides the suggestion that desorption proceeds from a hemispherical meniscus

while adsorption occurs on a cylindrical meniscus and thus for adsorption in
eq. (2.9) r1 = r and r2 = ∞ there are other explanations in the specialized literature,
such as changes of the contact angle caused by surface cleaning, different radii of
meniscus in adsorption and the desorption presence of interconnected pores behav-
ing differently during these processes.
For pore size distribution analysis it is not only hysteresis in nitrogen physisorp-
tion isotherm that can be used, but also mercury porosimetry. Because of the non-
wetting nature of mercury (the angle of contact of mercury with solids, θ, is above
74 2 Engineering catalysts

90° and ~140°) its intrusion into the pores is met with resistance and the pore radius
can be related to the applied pressure following the Washburn equation:

2σ cos θ
r= (2:13)
Δp

where σ is surface tension normally taken as 480 mN m−1. A comparison between

surface area determination by mercury penetration and nitrogen physisorption typi-
cally shows a discrepancy for microporous materials, which is associated with Hg
inability to penetrate small (micro)pores, uncertainty of the contact angle and sur-
face tension values and the probability of the cracking or deforming of samples.
However, mercury porosimetry remains as the most reliable method for investiga-
tion of macroporous materials.

2.2.4 Chemisorption

In addition to physisorption, which is used for the determination of specific area of

supports, the exposed surface area of the active (metal) phase and metal dispersion
can be evaluated using chemisorption. The metal dispersion D is defined as:

Ns
D= 100% (2:14)
NT

where ns is the number of surface atoms and nT is the total number of atoms.
In essence, chemisorption is the titration of surface sites by a certain adsorbate
(hydrogen, oxygen, CO), which reacts only with the active phase to form a mono-
layer. The understanding of stoichiometry (as shown in Fig. 2.25 for CO adsorption)
is important for getting the correct values of metal dispersion. In some instances,
such as the determination of Cu surface area by chemisorption with N2O, there is a
possibility of subsurface oxidation, which in pulse chemisorption (explained
below) depends on temperature (prominent above 100 °C), amount of the sample,
metal loading and the catalyst bed geometry.

O O O O O

C O C C C C C

Fig. 2.25: Different stoichiometry for CO chemisorption.

From the experimentally measured adsorbed volume the number of adsorbed moles
is calculated. Through knowing stoichiometry, the number of surface atoms is
 2.2 Toolbox in catalysis 75

determined. The total number of metal atoms and subsequent metal dispersion can
be calculated by knowing metal loading.
There is a relationship between the metal dispersion and the diameter of a
metal cluster on a support that can be calculated using the following approach.
Imagine a spherical cluster of the size d. The surface of this cluster (πd2) is com-
posed of metal atoms, each with a surface corresponding to the cross section of a
particular atom SMe. Then the number of surface atoms is defined as NS = πd 2/SMe.
The total number of atoms in the cluster is proportional to the mass of the cluster
and the Avogadro number and inversely proportional to metal molecular mass:

mcluster NA Vcluster ρNA πd3 ρ NA

NT = = = (2:15)
M M 6 M
where ρ is the metal density. The metal dispersion is computed from (2.14) and
(2.15):

mcluster NA Vcluster ρNA 6M

D= = = 100% (2:16)
M M dρSMe NA

which is finally related to the diameter of metal clusters:

6M
d= 100% (2:17)
DρSMe NA

Some equations, which can be used for calculating the metal particle size (in nm)
are given here for several metals: Pt: 108/D(%); Pd: 107/D(%); Ru: 101/D(%); Ni: 97/
D(%). An assumption in these calculations is the metal surface area. If only dense
metal surfaces are considered (like 111) with smaller SMe, somewhat larger values of
metal clusters will be obtained. Because of uncertainties in the stoichiometry, the
presence of spillover, the possible formation of subsurface species and not neces-
sarily spherical cluster shape, verifying the estimation of metal particle size with
chemisorption by some other physical methods is recommended. For example, di-
ameter d can be measured or calculated from several techniques, such as high-
resolution electron microscopy, X-ray diffraction or XPS. The overall number of ad-
sorbed molecules in a monolayer could be determined by temperature-programmed
desorption, while infrared spectroscopy using a probe molecule is widely applied
for determination of the number of acid and basic sites in zeolites and mesoporous
materials. These methods will be described later in detail.
In static chemisorption methods, after a high temperature evacuation the iso-
therm is measured, followed by a subsequent low-temperature evacuation, where
physisorbed molecules are removed. In the second phase of adsorption measure-
ments only these weakly adsorbed molecules are chemisorbed back on the surface.
Subtraction of the weakly adsorbed amounts from totally adsorbed ones gives
the amount of strongly and irreversible adsorbed gas, which is then used for
76 2 Engineering catalysts

Total chemisorption
after high-T evacuation
nad (mmol g–1)

Strong chemisorption
Irreversible

Weak chemisorption
after subsequent low-T evacuation
Reversible

p (kPa)

Fig. 2.26: Principles of the back sorption method.

calculating the surface area and metal dispersion. The principles of the back sorp-
tion method are illustrated in Fig. 2.26.
Although historically the static methods for chemisorption were applied, dy-
namic methods are also often used for measuring chemisorption (Fig. 2.27A).

1.4
Cumulative volume (mL g–1 STP)

1.2
1.0
0.8
Catalyst Detector 0.6
H2, CO
0.4
0.2
0.0
–0.2
0 1 2 3 4
Pulse Response Peak number
(A) (B)

Fig. 2.27: Dynamic method for measuring chemisorption. (A) pulse-response sequence.
(B) adsorbed amounts as a function of pulse number.

Determination of the metal dispersion of a supported metal catalyst (for example,

palladium on a mesoporous support, Fig. 2.27B) could be performed by measuring
CO chemisorption. The sample should be placed into a U-tube and reduced by hy-
drogen, while increasing the temperature linearly, 5 °C min−1 to 200 °C, then main-
taining it for 2 h. After the reduction, the catalyst sample should be cooled down to
40 °C under helium flow. Chemisorption of CO is then achieved by pulsing
CO through the system and completely saturating the catalyst (Fig. 2.27B). The
 2.2 Toolbox in catalysis 77

stoichiometric ratio between metal (Pd) and CO is assumed to be unity (in fact it
could depend on the cluster size, as the stoichiometry might change). The mean
particle size ds is calculated knowing the metal dispersion.

2.2.5 Temperature-programmed methods

There are several temperature-programmed techniques available for catalyst char-

acterization. A certain process (catalyst reduction, oxidation of coke, desorption of
pre-adsorbed species or their chemical reactions) is followed while the temperature
is increased linearly. This non-spectroscopic technique is inexpensive and applica-
ble to real and model catalysts.

2.2.5.1 Temperature-programmed reduction

This method is based on the reduction of a catalyst by hydrogen during temperature
ramping, while monitoring hydrogen consumption with a thermal conductivity de-
tector. Hydrogen consumption profiles are plotted as a function of temperature (or
time) as shown in Fig. 2.28.

0.09 Treated at 200 °C with NH3

prior to Pd addition

0.06
Treated at 600 °C with NH3
H2 uptake (a.u.)

prior to Pd addition

0.03
Treated at 400 °C with NH3
prior to Pd addition
0

–0.03
0 50 100 150 200 250 300 350 400
Temerature (°C)

Fig. 2.28: TPR of palladium supported on N-functionalized carbon nanotubes, treated with
ammonia at different temperatures prior to addition of palladium.

The number of peaks in the thermogram can be interpreted either by the presence of
different chemical species, which are reduced at different temperatures, or by the suc-
cessive reductions of one type of species. The degree of reduction or the initial oxida-
tion state (if reduction is complete) can be elucidated from the consumed amount of
78 2 Engineering catalysts

hydrogen. Finally, the positions of the peaks in temperature point the active phase
interactions with the support. Strong interactions with the support will result in a shift
in the peak maximum to higher temperatures, giving a broader peak (Fig. 2.29A).

Unsupported Supported Small nanoparticles: Large particles:

nanoparticles: nanoparticles with almost intrinsic rate TPR broadened
strong support of reduction by mass transport
interaction limitation

dM dM
dT dT

(A) T (B) T

Fig. 2.29: TPR curves. (A) Influence of support. (B) Influence of the cluster size.

In the reduction processes, where mass transport limitations are prominent, large
particles are reduced at higher temperatures and the peak is broadened (Fig. 2.29). In
the reduction processes controlled by nucleation (reduction of nickel oxides) the tem-
perature of the maximum is, conversely, increasing with a decrease in the particle
size, thus large particles are characterized by narrow and intense peaks, while small
clusters have a broader peak at higher temperatures.
As the position of the peaks may vary depending on the metal support interac-
tions, particle size and heating rate, one should be very cautious in interpreting the
thermograms and in selecting the reduction temperature from, for example, litera-
ture data, as this temperature is very sample- and condition-specific.
TPR is useful is analyzing bimetallic systems, as the number of peaks and the
shift of the peaks compared to monometallic systems provide information about the
interactions between the metals.

2.2.5.2 Temperature-programmed desorption

Temperature-programed desorption (TPD) involves heating a sample with a rising
temperature and simultaneously detecting the residual gas in the vacuum using a
mass analyzer. As the temperature rises, certain absorbed species will have enough
energy to escape and will be detected as a rise in pressure for a certain mass. A vac-
uum is needed to prevent re-adsorption, which is thus neglected in the mathemati-
cal treatment. The desorption rate goes through a maximum. At low T the coverage
 2.2 Toolbox in catalysis 79

is high, but the rate constant is low and overall the desorption rate is low. As the
temperature increases the rate constant and thus desorption rate increases. At in-
termediate temperatures the desorption rate is high as both the coverage and the
rate constant are high. At higher T the coverage is virtually zero and although the
rate constant is enormous, the desorption rate is virtually zero. The temperature
of the peak maximum provides information on the binding energy of the bound
species (Fig. 2.30A) as well as decomposition path for complex organic molecules
desorption (Fig. 2.30B).

50,000
500
Mass intensities (rel. units)

40,000 400

2.
30,000 300

T (°C)
20,000 200

1.
10,000 100
4.
3.
0 0
1,500 2,000 2,500 3,000
(A) Time (s)

o-Xylene

Benzene
CH4
H2
MS signal (au)

Toluene

C2H4
C2H4
CH4

323 423 523 623 723 823 923

(B) Temperature (K)

Fig. 2.30: (A) TPD of CO from Pd/C catalysts after stearic acid decarboxylation. 1, CO; 2, hydrogen;
3, CO2; 4, methane. (B) TPD of o-xylene. (From [13].)
80 2 Engineering catalysts

Activation energy of desorption can be determined from TPD experiments at differ-

ent heating rates β = dT/dt. For the first-order desorption kinetics –dθ/dt = θke–ΔEdes /RT
with a linear heating rate T = T0 + βt the temperature dependence of the coverage is:

  ΔEdes
dθ θk −
− = e RT (2:18)
dT β

which after differentiating gives:

0 1
ΔEdes ΔEdes
d2 θ k @ Edes − dθ −
− = θΔ 2 e RT + e RT A (2:19)
dT 2 β RT dT

At the peak maximum (T = Tp) it holds that – d 2θ/dT2 = 0, therefore:

dθ ΔEdes
− =θ (2:20)
dT RTp2

Elimination of the term dθ/dT and further simple manipulations give a possible de-
termination of Edes from TPD experiments with different heating rates β (Fig. 2.31).
The slope in Fig. 2.31 is equal to R/ΔEdes:

1 R T2 R kR
= ln + ln (2:21)
T ΔEdes β ΔEdes ΔEdes

2.6

2.5
103/Tp

2.4

2.3

2.2

7 8 9 10 11 12
ln(Tp2/beta)

Fig. 2.31: Determination of activation energy of first-order desorption.

 2.2 Toolbox in catalysis 81

Exercise 2.2: The following TPD data were obtained with different heating rates (Fig. E2.2).
Calculate the activation energy of desorption

2.1

1.8 20°C min–1

105.5°C
1.5
15°C min–1
NH3 concentration (%)

99.5°C
1.2
10°C min–1
0.9 90.7°C

0.6 78.4°C
5°C min–1
69.9°C
0.3
3°C min–1
0.0

0 30 60 90 120 150 180 210 240

Temperature (°C)

Fig. E2.2: TPD at different heating rates.

2.2.5.3 Temperature programmed oxidation (TPO)

In temperature-programmed oxidation (TPO) the nature of solids deposited on de-
activated catalysts is analyzed, giving valuable information about the regeneration
conditions. Oxygen consumption is measured as a function of time. The method is
often used to investigate coke deposited on supported metal catalysts and zeolites
(Fig. 2.32).
In many cases, such as the one shown in Fig. 2.32 for metal (nickel and vana-
dium) containing zeolitic catalysts used in fluid catalytic cracking (discussed in de-
tail in Chapter 4), TPO profiles usually contain several peaks that reflect the form of
carbon structure on the spent catalysts. When these peaks are not well-resolved as
in Fig. 2.32, deconvolution is needed to make proper assignments.
A higher temperature peak (peak N in Fig. 2.32) is associated with more gra-
phitic- like or highly aromatic coke with high H/C ratio. Peak M is assigned to cata-
lytic coke produced by acid-catalyzed cracking, which is more reactive to oxygen
than the graphitic coke, but less reactive than the coke associated with metals
(peaks K and L).
82 2 Engineering catalysts

1500

1000

500
TCD response (a.u.)

0
530 Peak M
1000
Peak L
Peak K 439 Peak N
500 340 613

–500
0 200 400 600 800
Temperature (°C)

Fig. 2.32: TPO of coke for an equilibrated fluid catalytic cracking (FCC) catalyst after cracking of
sour heavy gas oil. [Redrawn from [14]].

2.2.6 Calorimetry

Microcalorimetry is a very useful technique – conceptually easy but difficult to imple-

ment. Typically, during the measurements adsorption isotherm is obtained simulta-
neously with calorimetric data. In a differential scanning calorimeter there are two
sample holders. Thermal behavior caused by the temperature difference between the
sample under investigation and the reference sample is recorded (Fig. 2.33).

Furnace
Reference Sample

Heat flow
Balance controller

Fig. 2.33: Sketch of a differential scanning calorimeter.

In adsorption calorimetry there is a stepwise introduction of the adsorbate at a con-

stant temperature. For each adsorption step with a slow increase of pressure the
adsorbed amount and the evolved heat must be determined. The differential heat
can then be calculated from the evolved heat in relation to the number of molecules
adsorbed in a particular step qdiff = [dQ/dn]T,A as illustrated in Fig. 2.34A, which
 2.2 Toolbox in catalysis 83

demonstrates that the heat of adsorption remains approximately constant up to

nearly one-third of the monolayer, but decreasing thereafter. There is a wealth body
of experimental data for different systems (supported metals, zeolites, single crys-
tals) that confirms the decrease of the heat of adsorption with coverage. An exam-
ple is shown in Fig. 2.34B. A linear decrease in the heat of adsorption with coverage
corresponds to logarithmic (Temkin) adsorption isotherm θ = lnða0 pÞ=f
Differential heat of adsorbed CO (kJ mol–1)

60
250
55 (a) C2H2–Pd/AI2O3
50

Differential heat (kJ/mol)

45 200 C2H2–PdZn
40
35 150 C2H4–Pd/AI2O3
30
C2H4–PdZn
25
100
20
15
10 50
5
0 0
0.000 0.001 0.002 0.003 0.004 0 10 20 30 40
Adsorbed amount of CO (mmol g–1) Adsorbate amount (μmol/g)

Fig. 2.34: Differential heats of adsorption as a function of coverage for (A) CO at 40 °C on Pd/N-CNT
[15] and (B) ethylene on palladium catalysts [16]. Reproduced with permission.

2.2.7 X-ray diffraction

X-ray diffraction (XRD) is a versatile, non-destructive technique that reveals de-

tailed information about the chemical composition and crystallographic structure
of natural and manufactured materials.
Diffraction patterns are obtained from diffracted X-rays of a certain wave length
λ (for example, CuK α radiation, λ = 1.5418 Å) by rotating the ionization chamber 2θ
degrees in the scan mode within a range of 2θ values while the crystal is rotating θ
degrees. The crystal planes are then reflected. They are separated by spacing d
(Fig. 2.35A) when the Bragg law nλ = 2dsinθ is fulfilled, with reflection order n being
an integer. Diffraction patterns are compared with available databases to identify
crystalline phases present in the sample.
Because of imperfection of the crystal, noise and the interference of the back-
ground, the obtained diffractograms (Fig. 2.35B) are not perfect. Width of the peak
provides information about the crystallite size, which is inversely proportional to
the full width at half maximum (FWHM) of the diffraction peak according to a
widely used empirical Scherrer equation. Instead of the line width at half height, a
more accurate parameter is the ratio between the area of the peak and its intensity.
84 2 Engineering catalysts

Wave 8,000
front
6,000

Counts
4,000
Θ
2,000
d

0
0 2 4 6 8 10
(A) (B) 2ϴ

Fig. 2.35: XRD: (A) principle and (B) an XRD pattern for Na-MCM-41.

Commonly used catalytic supports such as silica and alumina typically have high
surface area and low crystallinity. For example, silica has broad diffraction peaks
(2θ = 20 and 25°), and also few broad diffraction peaks are obtained even for the most
crystalline aluminas. Conversely, mesoporous silicas such as MCM-41 (Fig. 2.35B)
have a high degree of organization and reflections found at small angles (below 8°).
Zeolites are crystalline materials and have very distinct diffraction patterns.
When reflections are not seen in the pattern, there could be several reasons
why. In addition to a complete absence of a certain phase, low concentrations of
the active phase could be an explanation because catalytic material should be pres-
ent in an amount larger than 1%. There could be not enough crystallinity in the ac-
tive phase, as an amorphous phase will not be distinguished from the base line.
Interference between the diffraction lines of the active phase and the support might
mask the presence of an otherwise crystalline phase. Small metal clusters (below
3 nm) cannot be detected (Fig. 2.36A), thus EXAFS is most frequently used for com-
pounds exhibiting only a short range order. Because of broadening of the peaks,
the size of metal clusters is usually overestimated from XRD analysis.
XRD can be used to study if metal sintering is happening during catalysis.
Fig. 2.36B illustrates a particular case of fatty acid hydrogenation for which there was
no sintering of palladium. It should be mentioned that not all reflections are seen in
an XRD pattern. Thus, for face-centered structures only those reflections are present,
where the values of indices (Fig. 2.36B) are unmixed (all even or all odd).
Figure 2.37 illustrates information that can be obtained from an idealized dif-
fraction pattern.

2.2.8 X-ray photoelectron spectroscopy and X-ray fluorescence

Analytical techniques, such as XPS based on the detection of ejected electrons, are
the most suitable methods for the analysis of surfaces because they probe a limited
depth of the sample (~5–10 first monolayers).
 2.2 Toolbox in catalysis 85

002(C)
8000

0.8 wt%-Ir-MCM-41 (9.7 nm)

Intensity

6000 111(Pd)
200(Pd)
220(Pd)
0,98 wt%-Ir-MCM-41 (6.4 nm) Initial
4000 10(C)
004(C)
1.1 wt%-Ir-MCM-41 (propably <3 nm)
Spent
40 45 50 10 20 30 40 50 60 70
2ϴ (°) 2ϴ (°)
(A) (B)

Fig. 2.36: XRD of supported metals. (A) Ir/MCM-41 catalysts prepared by different methods. (B) Fresh
and spent (after hydrogenation of fatty acids) Pd/C catalysts. Reproduced with permission from [17].

7,000

6,000
Peak area (= integral Intensity)
→ real measure for peak intensity
5,000 → ∙ crystal structure ( contents of the unit cell
∙ phase amount ( in a phase mixture)
Intensity [counts]

4,000
Peak width
→ crystallite size, defects (strain, disorder0
3,000 ∙ full width at half maximum (FWHM),
also known as “half width”
Peak height (= maximum intensity) → peak profile dependent!
2,000
→ approximation for peak intensity ∙ integral breadth
(= Integral Intensity / maximum intensity)
1,000 → less dependent from peak profile

0
20 25 30 35 40 45 50 55 60 65 70

Peak position 2theta [°] Peak shape (peak profile)

→ d-spacing → crystallite size, defects (strain, disorder)
→ lattice parameters (metrics of the unit cell)

Fig. 2.37: Information content of an idealized diffraction pattern. Reproduced from [18]. Copyright
© 2015 Recent Research in Science and Technology.

A limit of these techniques for applications in heterogeneous catalysis is repre-

sented by the high vacuum environment required rather than real catalytic working
conditions such as atmospheric or even high-pressure environments. Nevertheless,
XPS can give valuable information regarding: catalyst composition, i.e., the elements
present; chemical nature of the elements; chemical nature of neighboring (coordinat-
ing) atoms; dispersion of active phase and support; and location of active phase in
the particle.
In XPS measurements, ionization of the inner electron shells of atoms occurs
because of exposure to monochromatic X-rays (Fig. 2.38).
86 2 Engineering catalysts

Photoelectron
Kinetic
energy
EV
ϕ

EF
Valence
band Binding
energy

Photon

Core hole

Core
levels

Fig. 2.38: Principles of XPS.

The difference between the kinetic energy of the emitted electrons and the
X-rays energy is measured, which is then related to the binding energy:

hv = EB + EKin + Ecorr (2:22)

where hv is the energy of photons, Ekin is the kinetic energy of emitted electrons, EB
is the binding energy of emitted electrons, Ecorr is the correction factor, also includ-
ing the work function ϕ.
XPS is a surface sensitive technique as only electrons close to the surface can
escape without energy loss, while those deeper in the bulk lose part of their kinetic
energy while travelling towards the surface and cannot escape.
Several peaks are expected because of the spin-orbital coupling. The binding
energy or position of a peak is specific for an element and depends on the oxidation
state (electrons are stronger bound to atoms in higher oxidation states), chemical
environment (for example, neighboring atoms, types of ligands, their number, etc.).
Position of the main peak and satellite peaks, caused by energy loss through plas-
mons, help to identify the elements, and thus XPS is often called ESCA (electron
spectroscopy of chemical analysis).
Quantitative analysis of XPS spectra takes into account the charging of the sam-
ple during measurements (and thus shift in the binding energy), which can be per-
formed by relating the peaks to a reference peak, for example, C 1s peak at
284.6 eV. In addition, when the measurements take a significant time there is a pos-
sibility of surface atoms reduction altering the ratio between the peaks correspond-
ing to different oxidation states. As an example, XPS measurements of supported Pt
catalysts that contain chlorine are presented in Fig. 2.39, demonstrating that the
ratio between Pt+2 and Pt+4 is changing during measurements.
 2.2 Toolbox in catalysis 87

PtCl2 Al PtCl4

269 – 341 min (c)

75 – 245 min (b)

30 – 75 min (a)

68 69 70 71 72 73 74 75 76 77 78 79 80
Binding energy (eV)

Fig. 2.39: Pt 4f photoelectron spectra of 5 wt% Pt/silica fibers catalyst during prolonged X-ray
bombardment.

As already apparent from Fig. 2.39, deconvolution of XPS spectra with the con-
tribution of several oxidation states can be challenging.
XPS data could be also used for determination of the metal particles size esti-
mations based on the dispersed phase/support phase XPS intensity ratio, as elabo-
rated by Davis [19].
X-ray fluorescence (XRF), widely used for elemental analysis, is the emission of
characteristic “secondary” (or fluorescent) X-rays from a material that has been ex-
cited by bombarding with high-energy X-rays or gamma rays (Fig. 2.40). The term
“fluorescence” is applied to phenomena in which the absorption of higher-energy
radiation results in the re-emission of lower-energy radiation.

2.2.9 Infrared and Raman spectroscopies

Vibrational spectroscopies are applied in order to study chemical bonds in chemi-

sorbed species as well as the solid substances. In the IR domain (200–4,000 cm−1)
the absorption of energy excites bonds to a higher vibrational state with only excita-
tions modifying the dipolar moment taking place. Different chemical bonds absorb
IR-light at certain wave lengths and the absorbed intensity can be analyzed, allowing
88 2 Engineering catalysts

hν
hν

Fig. 2.40: Principles of X-ray fluorescence.

characterization of the bonds. Typically emissions of a fast pulse of light are analyzed
by Fourier transform (FTIR). Transmission mode is applied for transparent samples
(Fig. 2.41), which are either self-supported or KBr diluted wafers prepared under me-
chanical pressure.

IR

IR

Transmission Diffuse reflectance

Fig. 2.41: IR spectroscopy arrangements.

Transmission is widely applied for solids and the spectrum is a result of the trans-
mission of catalyst vs transmission of a reference. In diffuse reflectance mode
(DRIFTS) catalyst grains are dispersed in a diffusing matrix, whose spectrum should
be subtracted.
Infrared spectroscopy can be used for catalyst characterization by, for exam-
ple, direct measurement of catalyst IR spectrum and analyzing positions of OH
groups, as these positions and shapes provide information about the coordination
(Fig. 2.42).
Measurements of interactions with probe molecules, such as ammonia, acetoni-
trile, tert-butylnitrile or pyridine, allow the determination of catalysts acidity, while
basic sites could be probed with, for example, CO2. There is a long tradition dating
back to the 1950s and 1960s of using IR studies of adsorbed ammonia and pyridine
molecules for the characterization of solids. The advantages of ammonia as a probe
molecule for acidity are related to the molecule size, which is able to penetrate even
very narrow pores and small cages of zeolites. Moreover, ammonia does not con-
taminate vacuum installation. However, ammonia forms dimers and the spectra are
not as well-resolved as for pyridine adsorption. Fig. 2.43 illustrates adsorption of
ammonia and pyridine on proton (Brønsted, H+) and Lewis (L) sites. The band
 2.2 Toolbox in catalysis 89

Vicinal Isolated Geminal

H
H H
H
O O O
O O

Si Si Si Si
O

3,720, 3,520–3,720 cm–1 3,750 cm–1 3,742 cm–1

Fig. 2.42: Silanol groups on amorphous silica.

NH4+
1,450
N
PyL
1,450 Lewis acid
NH3L
1,620

PyH+
1,545

1,700 1,600 1,500 1,400 cm–1

Fig. 2.43: IR spectra of ammonia and pyridine adsorption on Brønsted and Lewis acid sites.

intensity should be related to the concentration of acid sites by applying extinction

coefficients.
Although pyridine is most frequently used for qualitative and quantitative acid-
ity studies its disadvantage is related to the size, disallowing entrance to pores
smaller than 0.5 nm. Acetonitrile, another often-used probe molecule, is not basic
enough although it is small.
IR can be also used to elucidate metal dispersion. Fig. 2.44 shows adsorption of
Pt/Si-SBA–15 catalyst, the data were obtained at University of Pierre and Marie Curie.
A band at 2,079 cm−1 is characteristic for CO linearly adsorbed on Pt nanoparticles,
while bridging CO bands adsorbed on two Pt atoms (1,860 cm−1) are favored on larger
faceted Pt particles. Thus, judging by low intensity of the latter signal, it can be stated
that large particles of Pt are absent and the catalyst is highly dispersed.
Note that the investigation of catalysts with high metal loadings is problematic
because of low transmission.
FTIR is widely applied in mechanistic studies, allowing the analysis of ad-
sorbed substrates, reaction intermediates and deactivation, as well as in situ reac-
tion monitoring. Fig. 2.45 illustrates FTIR spectra obtained by flowing a mixture of
90 2 Engineering catalysts

0.08

0.07 Pt-CO
2,079
0.06
Si-OH-CO
0.05 2,157
Absorbance

0.04

0.03

0.02

Pt2-CO
0.01
1,860
0.00
2,300 2,200 2,100 2,000 1,900 1,800 cm–1

Fig. 2.44: Adsorption of CO on Pt/Si-SBA–15.

1.4
1,605

1,250
1,295
1,550

1.2

1.0
Absorbance

0.8
NO + O2 (80 min)

0.6 NO + O2 (60 min)

NO + O2 (40 min)
0.4

NO + O2 (15 min)
0.2
NO + O2 (6 min)

0.0 NO + O2 (1 min)
2,500 2,000 1,500 1,000 cm–1

Fig. 2.45: Time-on-stream in situ FTIR on Ag/alumina at 250 °C in a flow of 1,000 ppm NO and
6 vol% O2 in He. (Reproduced with permission from [20]).

NO and oxygen through a silver on alumina catalyst. Slow formation of the peaks at
1,250, 1,295 and 1,550 cm–1 and a shoulder at 1,605 cm−1 is attributed to generation
of monodentate nitrate (1,250 and 1,550 cm−1) and bidentate nitrate (1,295 and
1,605 cm−1).
 2.2 Toolbox in catalysis 91

As a general comment on in situ and operando spectroscopies, not all observed

species participate in the reaction, and some spectator species could be registered
just because they do not react.
In addition to IR, ultraviolet-visible spectroscopy in the ultraviolet-visible spectral
region (50,000–2,500 cm−1) is also frequently applied for the analysis of solid catalysts
and adsorbed species. It uses light in the visible and adjacent ranges (near-infrared,
12,500–3,300 cm−1). UV-vis spectra obtained in situ at the Fritz Haber Institute during
isomerization of n-butane on mordenite zeolite [12] show bands at 250–370 nm
(Fig. 2.46) that are ascribed to neutral conjugated double-bond oligomers, while bands
at 370–500 nm correspond to carbocationic oligomers.

1.0

293
~4 h
0.8
Kubelka Munk function

2h
0.6
1h

0.4
330

0.2 395 450

Activated

0.0
250 300 350 400 450 500 550 600
Wavelength (nm)

Fig. 2.46: UV-vis spectra in butane oligomerization over H-mordenite at 534 K with n-butane/
nitrogen ratio 5:95. Reproduced from [21].

While reflectance or transmission is transformed into adsorbance in FTIR (Fig. 2.45),

in UV-vis the Kubelka Munk transform is applied in diffuse reflectance (Fig. 2.46),
relating the ratio between the absorption coefficient and the scatter coefficient of
the sample at a given wavelength to reflectance.
A variation of IR is attenuated total reflectance (ATR) that enables samples to be
examined directly in the solid or liquid state without further preparation (Fig. 2.47).
Contrary to IR, Raman spectroscopy relies on different wavelengths of the scat-
tered light, which is a feature only of a small fraction (10−6) of emitted light. The
selection rules are also different in Raman and IR spectroscopies, and some types of
molecular motions appear in Raman and IR, and some only in Raman or IR. Namely,
a criterion for Raman scattering is polarizability, while for IR it is a transition in the
92 2 Engineering catalysts

Sample in contact
with evanescent wave

To detector
Infrared
beam ATR crystal

Fig. 2.47: Schematic of ATR-IR. (Reproduced from [21].)

vibrational states that induce a change in the dipole moment. A particular vibration
mode in Raman is silent if polarizability is not changed by vibration.
Raman spectroscopy is more useful for studying the catalyst itself and since
water does not interfere with Raman spectroscopy it could be used in aqueous sys-
tems, where IR is difficult. Conventional Raman signals have lower sensitivity that
requires an enhancement, which is achieved by applying lasers. Raman spectros-
copy can be combined with “operando studies”, where the spectra of the catalyst
are measured during a catalytic reaction (Fig. 2.48).

Operando reactor

Catalyst Thermocouple well

Inert packing

Online gas Raman system NH3

chromatograph Renishaw system 1000 Vent
Gas feed
system O2

Heat C3H8
screen

He

Heating element Temperature

controller
Operando system

Fig. 2.48: Operando system. (Reproduced with permission from [22].)

 2.2 Toolbox in catalysis 93

IR-visible SFG spectroscopy has recently started to be applied for catalytic materi-
als. It uses two input laser beams overlapping at the surface of a material, generating
an output beam with a frequency that is the sum of the two input beams. One of the
two input beams is a visible wavelength laser at a constant frequency, while the
other is from infrared laser that is tuned to scan the system and obtain the vibrational
spectrum (Fig. 2.49). SFG has the disadvantage of a relatively small frequency range
(1,600–4,000 cm−1) and lower resolution and sensitivity, compared with IR.
Vi
sib

Tun
le

ab
le I
la

R la
se

ser
r

ω sum
ω Vi

ω
IR
s

Sample surface

Fig. 2.49: Principle of IR-visible sum frequency generation spectroscopy.

Conversely, this method affords the possibility of obtaining vibrational spectra

in situ at both buried and exposed interfaces, as well as under equilibrium and reac-
tion conditions. This is in contrast to traditional vibrational spectroscopies when
signals are generated from the surface.

2.2.10 Catalyst particle size measurements

The mean particle size of a catalyst can be determined by a light scattering tech-
nique using a laser diffraction and commercially available particle size analyzers.
The determination is based on He-Ne laser light scattering as light is passed
through the particle suspension. The scattering profile is collected by a Fourier
Transforming lens and the various outputs are handled and converted to particle
size distributions by an integral computer.
The particles size of powdered catalysts can be measured using a counter type
apparatus that has one or more microchannels that the separate two chambers
containing the electrolyte solutions. When a particle flows through one of the mi-
crochannels, it results in the electrical resistance change of the liquid-filled micro-
channel. This resistance change can be recorded as electric current or voltage
pulses, which can be correlated to size, mobility, surface charge and concentra-
tion of the particles.
94 2 Engineering catalysts

2.2.11 Electron paramagnetic/spin resonance

The technique relies on resonant absorption of microwave radiation by paramag-

netic ions or molecules with at least one unpaired spin in the presence of a static
external magnetic field. From an experimental perspective, the measurements are
organized by keeping the wavelength constant and recoding variations in the mag-
netic field. As a result, complex spectra in a narrow range of energy are obtained
with the superposition of different paramagnetic species and several spectra com-
ponents caused by anisotropy of the environment.
EPR is applied for structural characterization of solid and liquid systems and
only radicals (species with unpaired electrons) can be detected. Transition metal
cations in specific oxidation states (for example, Cr3+ d3, F3+ d5 high spin), O– and
O2 are most relevant to catalysis.
Signal positions depend on the electron configuration of the molecule or cat-
ion and on the energy difference between the d orbitals for paramagnetic transition
metal cations. Delocalization of electrons on metal ions and ligands is reflected
from the number of lines and their intensities, while coupling constants provide
information on the bonding between the cation and the ligands. In addition, probe
molecules can be used to study the species present in the catalyst. For example,
Fig. 2.50 shows the ESR spectra recorded at 77 K for the fresh iron-modified beta

3,000
ESR spectra 4.Fe-Beta-25 IE
recorded at 77 K
2,000 g = 4.25

1,000
g = 1.99 Fresh sample

–1,000 g = 4.25
Intensity (a.u.)

After evac at 673 K 4 h

–2,000 g = 2.0023
g = 4.26
–3,000
g = 3.93
–4,000 After 1 doz of NO
and evac at RT 5 min
–5,000

–6,000
0 100 200 300 400 500
Field (mT)

Fig. 2.50: ESR spectra of fresh, evacuated and NO adsorbed Fe-H-Beta-25 (Reproduced with
permission from [23]).
 2.2 Toolbox in catalysis 95

zeolite, for the same material after evacuation at 300 °C for 4 h and after adsorp-
tion of NO.
The ESR spectrum for the fresh sample displays signals in two different re-
gions of the magnetic field, a broad line that can be assigned to randomly oriented
Fe3+ ions in oxides and/or hydroxides species. The signal at g = 4.25 in a low field
region originates from Fe3+ ions in tetrahedral coordination. After evacuation at
673 K for 4 h a signal at g = 4.25 is more visible and a signal from iron oxide and
hydroxide species is less intensive. The new narrow ESR signal (g = 2.0023) can
originate from coke formed from residual template, but it also can be caused by
Fe3+ in octahedral coordination, as isolated ions in cationic position attributed a
sharp line at g = 2 to superoxide ions (O2 − ) that are associated with iron ions.
Adsorption of NO results in ESR signals in the magnetic field range 300–350 mT
and in another one in the low field (g = 3.9). After short evacuation a rhombic sig-
nal appears, with three distinct values of the diagonal terms (g = 2.09, g = 2.06,
g = 2.02). The low field value (g = 3.93) originates from (NO)2 biradicals. Based on
this, it can be assumed that most of the iron is tetrahedral Fe3+ but that some Fe2+
is also present.
In addition to the identification of paramagnetic species on the surface or
within cavities of zeolites/mesoporous materials, EPR can be used for identification
of radicals, which could escape into the gas-phase. Such events could occur, for ex-
ample, in partial oxidation of hydrocarbons or in hydrocarbon-assisted SCR of NOx
in which the generation of radicals was detected by matrix isolation EPR [24].
Propagation of the surface reactions into the gas-phase is beneficial for the overall
NOx conversion and could be used in engineering of catalytic converters. For exam-
ple, a catalytic unit for reduction of NOx can contain several catalytic beds
(Fig. 2.51; 3a–3e), that are arranged separately from each other in a longitudinal di-
rection in the catalyst unit, and a number of second reaction zones (Fig. 2.51;
4a–4e) downstream of each respective catalytic bed.

2 5 4a 4b 4c 4d 4e 7 6
L1

L2

1 3a 3b 3c 3d 3e

Fig. 2.51: Catalytic converter. 1, catalytic unit; 2, exhaust pipe; 3, catalytic zones; 4, gas-phase
reaction zones; 5, inlet; 6, outlet; 7, oxidation catalyst. (Reproduced from [25]).
96 2 Engineering catalysts

2.2.12 Mössbauer spectroscopy

Mössbauer spectroscopy provides information about the oxidation state, coordina-

tion number, ligand type, electron distribution, the strength of binding to the sur-
rounding of the resonant atom and the electric charges of the neighbors. The
technique is based on absorption by a nucleus of the resonant atom of photons
emitted by a radioactive source (for iron this is 57Co). The wavelength emitted has a
constant value, which is slightly varied by the Doppler effect, i.e., by moving the
gamma-ray source towards and away from the sample at varying speeds, typically a
few mm s−1 (Fig. 2.52). For convenience, Mössbauer spectra allowing observations
of the hyperfine interactions correspond to the intensity of the transmitted X-rays
as a function of the Doppler velocity.

Radioactive Sample Detector

source

Fig. 2.52: Mössbauer apparatus.

During emission or absorption of photon a free nucleus recoils because of the con-
servation of momentum with a certain recoil energy. Resonance is achieved when
the loss of the recoil energy is overcome by having the emitting and absorbing nu-
clei in a solid matrix. Identical environments of emitting and absorbing nuclei pro-
duce a single absorption line spectrum, as shown in Fig. 2.53. As the resonance
occurs only in the case of exact matching the transition energies of the emitting
and absorbing nucleus the effect is isotope specific.

Emitter nucleus Absorber nucleus

g g

Detector

Fig. 2.53: Simple Mössbauer spectrum from identical source and absorber.
 2.2 Toolbox in catalysis 97

The strength of the signal reflects the relative number of recoil-free events and is
dependent upon the γ-ray energy, thus the Mössbauer effect is only detected in iso-
topes with very low-lying excited states. The resolution is dependent on the lifetime of
the excited state. Because of these constraints the technique is limited to a few iso-
topes, including (relevant for catalysis) 57Fe, 119Sn. For isotopes such as 193Ir or 197Au, a
very short lifetime of a radioactive source prevents their characterization by Mössbauer
spectroscopy.
When the source and absorber are identical the absorption peak is seen at zero
velocity (0 mm s−1), while various types of changes in the environment of the reso-
nant atoms lead to changes in the spectra. The isomer shift, which is useful for de-
termining valence states, ligand bonding states and electron shielding, is caused by
changes in the s-electron environment, allowing differentiation between Fe+2 and
Fe+3 for example.
The nuclear energy levels are split (quadruple splitting) in the presence of an
asymmetrical electric field caused by asymmetric electronic charge distribution or
ligand arrangement. Magnetic splitting of nuclear levels with a spin of I in the pres-
ence of a magnetic field into (2I + 1) substrates is caused by dipolar interactions of
the nuclear spin moment with the magnetic field.
Thus 57Fe Mössbauer spectroscopy is the method for the study of valency, coordi-
nation and magnetic properties of the iron species in for example iron-modified zeo-
lites, contrary to the limited abilities of ESR or IR. In these materials, in addition to
extra-framework iron species (for example, not isomorphously substituted), isolated
Fe ions with a different oxidation state that are chemically anchored to the frame-
work via –O–Fe bridges, are present in addition to dinuclear Fe-O-Fe species con-
fined in the channels, oligomeric Fe oxi species in the large cavities and oxide-like
clusters. Typically, 57Fe Mössbauer spectra (see Fig. 2.54) are fitted to get the best cor-
respondence with the data by using the line width peak intensity, isomer shifts and

100
100
98
Transmission (%)

Transmission (%)

99.8 96

94
99.6
92

90
–2 –1 0 1 2 –10 –8 –6 –4 –2 0 2 4 6 8 10
(A) Velocity (mm s–1) (B) Velocity (mm s–1)

Fig. 2.54: Room temperature 57Fe Mössbauer spectra of (A) Fe-ZSM-5 prepared by ion exchange and
(B) spectrum with sextet: α-Fe2O3. (Reproduced with permission from [26]).
98 2 Engineering catalysts

the quadruple coupling constant as fit parameters. The quadruple coupling constant
measures the splitting in mm s−1 between the two lines in the doublet.
Fig. 2.54(A) displays 57Fe Mössbauer spectra consisting of a single paramagnetic
doublet for Fe-ZSM-5 zeolite, prepared by ion exchange. In fact, the shape of spectra
depends very much on the preparation methods. In the case of ion exchange prepara-
tion methods the presence of Fe3+ ions in octahedral coordination gives intense central
doublets (0.3 < isomeric shift < 0.4 mm s−1 and 0.5 < quadruple splitting < 1.1 mm s−1),
while the appearance of additional doublets with the isomeric shift ~1.2 mm s−1 and
1.5 < quadruple splitting < 2.8 mm s−1 (absent in Fig. 2.54A) is associated with Fe2+ ions
in an octahedral coordination. Agglomerated iron oxide species could result in a sextet
(absent in Fig. 2.54A, but present in Fig. 2.52B).
If Fe-ZSM-5 is prepared by isomorphous substitution a broad singlet with an
isomer shift of 0.52 mm s−1 will be observed in the spectra, corresponding to tetra-
hedrally coordinated high spin Fe3+ ions and reflecting the incorporation of iron
into the framework. If the same catalyst is calcined, migration of iron to an extra-
framework position and increased interactions between extra-framework Fe3+ and
the zeolite lattice will lead to an unresolved doublet in the Mössbauer spectrum.
Alternative methods of Fe-ZSM-5 catalyst preparation (impregnation, reductive
solid state ion exchange and chemical vapor deposition) result in spectra, which
could be deconvoluted into a hyperfine magnetic sextet and two quadruple doublet
lines, confirming the presence of large α-Fe2O3 particles and either small α-Fe2O3
particles or Fe3+ in a strong distorted environment.

2.2.13 X-ray absorption spectroscopy

This is an element-specific type of spectroscopy that requires synchrotron produced X-

rays with energy close to the absorption edge of the element core shells. Information
about the oxidation state of the element is retrieved from the edge (Fig. 2.55) as the
edge shifts to higher energy with an increase of oxidation state.
X-ray absorption near-edge spectroscopy (XANES, Fig. 2.55), in addition to al-
lowing elucidation of the oxidation state, also provides information on the electro-
negativity of the neighboring atoms and symmetry of the atomic environment.
Electrons from the K edge correspond to s-orbitals, while the L edge corresponds to
p-orbitals. EXAFS is used to acquire information on the coordination shells around
the atom, number of surrounding atoms, bond lengths, dynamic and static disorder
in the inter-nuclear distances. It is especially useful for materials without long-
range organization when XRD cannot be applied. The waves of atoms interfere with
the wave of the emitted after-ionization electrons, resulting in absorption oscilla-
tions seen in EXAFS spectra. Fourier transform is applied to decompose the EXAFS
signal. Each peak in Fig. 2.56, presented here for illustration, corresponds to a shell
of scattering atoms at a certain distance. A number of parameters is needed for the
 2.2 Toolbox in catalysis 99

2.0

Normalized absorption
EXAFS 1.6
Absorption (a.u.)

4
3
1.2
2
1
XANES 0.8 1 PdO
2 2% Pd/HBETA-300, 120°C, He
3 3.9% Pd/HBETA-150, 120°C, He
0.4
4 4.7% Pd/HBETA-25, 120°C, He
5 4.1% Pd/HBETA-22, 120°C, He
0.0
Energy (eV) 24,300 24,400 24,500 24,600
(A) (B) Photon energy (eV)

Fig. 2.55: X-Ray absorption spectroscopy. (A) XANES and EXAFS regions. (B) Pd K-XANES spectra of
calcined Pd/H-beta catalysts and PdO.

1
FT (Xk2)

–1

–2

–3

0 1 2 3 4
Uncorrected distance (Å)

Fig. 2.56: Model fits of Pd K EXAFS spectra taken from a calcined 2% Pd/H-beta-300 catalyst.

data fitting thus precise simulation is challenging especially when too many differ-
ent atoms and/or distances coexist. Blurring of a signal is possible when there is
too large a discrepancy in the atom environment, as the EXAFS signal is an aver-
aged linear combination of all sites of a particular element.

2.2.14 Nuclear magnetic resonance

Nuclear magnetic resonance (NMR), being isotope specific, concerns isotopes with
a nuclear spin different from zero. It is based on splitting of the nuclear spin ground
level by an external magnetic field of constant value and the subsequent absorption
100 2 Engineering catalysts

of energy from electromagnetic radiation of varying wavelengths. Absorption signal

positions (chemical shift) depend on the nuclei electron environment shielding nu-
cleus spin from the external magnetic field. For a nuclear spin I the number of lines
is 2nI + 1, which are separated by a coupling constant, the latter linked to the dis-
tance between the nuclei. The inherent sensitivity of NMR is low, although it is pos-
sible to get larger signals by using highly dispersed or porous materials. For the
solid materials, broadening of the signals is overcome by rotating the sample 54.74°
with respect to the magnetic fields (magic angle spinning, MAS).
1
H NMR is applied for the characterization of OH groups, while 13C NMR can
be used for studying adsorption of isotopically marked organic compounds. The
chemical shifts in 29Si MAS NMR are linked to the number of silicon atoms pres-
ent as second neighbors. An example of the 29Si MAS NMR spectra for a mesopo-
rous material H-MCM-41 is given in Fig. 2.57, showing several broad overlapping
lines. A computer fit to the spectrum gives the lines at −108.5 ppm (65% of the
total intensity), −100.6 ppm (13%), −94.8 ppm (19%) and −80 ppm (3%). The
strongest peak at −108.5 ppm can be attributed to Si(OSi)4 sites, while the com-
ponents at −100.6, and −94.8 ppm are caused by the sites with silanol groups Si
(OSi)3OH and Si(OSi)2(OH)2, respectively. As the material has a considerable
amount of Al in the framework, the contribution to the signal is expected from
the sites Si(OSi)3(OAl), Si(OSi)2(OAl)2, Si(OSi)(OAl)3, and Si(OAl)4 in the region
of −80 ppm to −100 ppm.
Intensity

Intensity

0 –40 –80 –120 –160 –200 200 150 100 50 0 –50 –100
ppm ppm

Fig. 2.57: 29Si (left) and 27Al (right) MAS NMR spectra of H-MCM-41 including the computer fits.
(Reproduced from [27] with permission from the PCCP Owner Societies).

The 27Al MAS spectrum of H-MCM-41 (Fig. 2.57) shows a signal with a maximum at
54 ppm caused by tetrahedrally coordinated framework aluminum and a broad
asymmetric line around 0 ppm caused by octahedrally coordinated aluminum sites.
 2.2 Toolbox in catalysis 101

A small component arising from the computer fit was attributed tentatively to the
distorted tetrahedral or five-coordinated Al sites. The quantitative analysis of the
spectrum yields the overall absolute number of Al sites, as well as the respective
values for framework tetrahedral, non-framework tetrahedral or five-coordinated
and octahedral aluminum. The Si/Al ratio can be estimated from the 27Al line
intensity.
NMR can also be used to characterize microporosity of zeolites applying 129Xe,
whose chemical shifts depend on the inner void structure of the solid and charge
gradients.

2.2.15 Imaging of catalysts

Several techniques are applied for the imaging of catalysts.

2.2.15.1 Electron microscopy [28, 29]

Electron microscopy uses a particle beam of electrons to illuminate a specimen and
create a highly-magnified image (Fig. 2.58). Electron microscopes have a much
greater resolving power and can obtain much higher magnifications than light mi-
croscopes that use electromagnetic radiation. Electron microscopy is usually com-
bined with elemental (EDX) or structural analysis (XRD). In EDX X-rays are emitted
by the atoms for which inner shell energy levels are ionized by electron excitation
and then collected.

Primary electron beam Cathode

Anode
EDX, WDX Lens

X-rays Backscattered electrons

Sample
Secondary lens
electrons SEM
Photons Auger electrons

Image
TEM, lens
dark field Diffracted Loss
electrons electrons

Transmitted TEM,
electrons bright field Final image

Fig. 2.58: Different electron microscopes. (Reproduced with permission from [29]).
102 2 Engineering catalysts

The TEM uses a high voltage electron beam (100–300 keV) to create an image.
The electrons are emitted by an electron gun and the electron beam is accelerated by
an anode focused by electrostatic and electromagnetic lenses, and transmitted
through the specimen that is in part transparent to electrons and in part scatters
them out of the beam. The magnification range in TEM is 102–107. The sample is put
on a grid and is placed in a vacuum chamber. Images are formed because of different
lateral absorptions of the beam, with heavy atoms being darker. A 10–7 Torr is usually
used for high-resolution electron microscopy to avoid noise caused by gas molecules.
The practical limit of detection is ~1 nm, although several times lower resolution is
possible. Support and the supported phase are distinguished where there is good
contrast between them (Fig. 2.59). Discrimination of the two phases is easier when
their d-spacing and shape are different. Oxide supports are more difficult to distin-
guish than the metallic phase. The same difficulty applies to metals of the first transi-
tion row compared to heavier metals from the third row. Challenges in images are
also present in the case of poorly organized supports, such as γ-alumina, which ex-
hibits agglomeration of small crystallites, and on well-crystallized supports with
many d spacings in images.

3 nm

Fig. 2.59: A metal cluster on a support: TEM image.

For small particles with poor contrast with the support, good images are obtained
when electrons are scattered at high angles, which is far from the position of the
objective aperture in classical TEM. In the case of high-angle annular dark field
(HAADF) metal particles are visible as bright sports over a dark background.
For calculation of cluster size distribution (Fig. 2.60) and the average size of
particles, several hundreds of clusters are counted that counterbalance premature
conclusions based on eye-catching zones (unexpectedly large particles or unusual
particle shape).
 2.2 Toolbox in catalysis 103

40

30
Frequency (%)

20

10

0
0 1 2 3 4 5 6 7 8 9 10 11 50 nm
(A) Particle size (nm) (B)

Fig. 2.60: HRTEM of Au/Al2O3 catalyst. (A) Image. (B) Cluster size distribution.

Another recent development in HRTEM is associated with 3D imaging, when

the sample is tilted with a small (1°) increment and the images are taken after each
change.
Although HRTEM analysis became very popular and is widely applied, it is im-
portant to remember the challenges and shortcomings associated with the method,
such as the inability of imaging of clusters well below 1 nm, representativity of the
specimen area, potential damage during preparation as well as electron beam dam-
age. High energy of the incident electron beam may cause changes in the micro-
structure of crystallites or the reduction of surface ions. Applicability of TEM for
in situ studies is limited by losing resolution at above ~5 mbar. In fact, all electron
microscopy techniques rely on high vacuum because of the large attenuation of the
electron beam in any atmosphere.
In general as technically relevant catalysts should be studied under certain gas
atmospheres it is difficult to extend vacuum-based techniques, including TEM, to
in situ experiments.
SEM has a magnification range 10–105. SEM produces images by probing the
specimen with a focused electron beam (accelerating voltage 4–30 keV) that is
scanned across a rectangular area of the specimen. Electrons emitted or scattered
by the sample are then detected. Generally, the image resolution of SEM is about
one order of magnitude poorer than that of a TEM. Fig. 2.61 shows SEM images for
ZSM-5 zeolite.
Scanning transmission electron microscopy (STEM) is a merger of TEM and
SEM, allowing the same resolution as TEM and the ability to perform microanalysis
at a high resolution similar to SEM.
104 2 Engineering catalysts

5 µm

Fig. 2.61: SEM image of zeolite ZSM-5.

2.2.15.2 Scanning probe microscopy

In scanning probe microscopy the very sharp probe is brought in contact with a sam-
ple. In scanning tunneling microscopy (STM) the probe does not touch the surface
(Fig. 2.62A). A constant tunneling electric current is maintained, and the method is
thus limited to conducting materials. Very high resolutions (height 0.01 nm, length
0.1 nm) can be achieved.

Laser
– Battery +
Current Laser
detector

Mirror

Tunneling current

Computer
Microscope tip
Sample
Sample surface scanned
(A) (B) back and forth Display

Fig. 2.62: Principles of (A) STM and (B) atomic force microscopy (AFM). (Modified from [29]).

STM studies of catalytic systems under realistic conditions (high pressure and high
temperature) are possible and have been recently achieved.
In AFM the probe can touch the surface (Fig. 2.62B). As a constant small force is
maintained the method is not limited to conducting materials and is suitable for all
 2.2 Toolbox in catalysis 105

surfaces. The method allows a high resolution (length 2–10 nm, height 0.1 nm).
Desulfurization reactions in oil refining are an example of applying STM to cataly-
sis. Triangular MoS2 nanoparticles have sulfur atoms located in the edges, which
react with hydrogen, forming hydrogen sulfide. The resulting vacancy is replaced
by sulfur from the sulfur-containing oil molecules. STM studies by F. Besenbacher
and co-workers [30] (Fig. 2.63) demonstrated that the morphology of MoS2 nanopar-
ticles strongly depends on their size. When the particles contain less than 100
atoms, the relative sulfur content increases above the stoichiometric value for the
bulk MoS2. Subsequent changes occur in the edge structure to counterbalance this
increase and to lower sulfur to a molybdenum ratio and improve stability of the
nanoparticles that is deteriorating with sulfur increase. Refer to a review article [31]
for a discussion of the progress in the application of scanning probe microscopy in
catalysis.

Fig. 2.63: STM of MoS2. (Modified from [30]).

2.2.15.3 Electron microprobing

Determination of an active phase profile in catalysts grains can be performed using an
electron probe microanalyzer, which is an analytical tool used to non-destructively
determine the chemical composition of small volumes of solid materials. It works in a
similar way to SEM by bombarding the sample with an electron beam and collecting
the signals coming from the sample.
Another non-destructive method is based on applying XPS with a variation of
the emission angle.
Destructive methods involve mechanical and chemical erosion by, for example,
progressively removing surface atoms caused by ion bombardment. Another tech-
nique is laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS),
which combines the micrometer-scale resolution of a laser probe with the speed,
sensitivity and multi-element capability of ICP-MS, and rivals other microbeam
106 2 Engineering catalysts

techniques such as the proton microprobe and secondary ion mass spectrometry. A
pulsed laser beam is used to ablate a small quantity of sample material, which is
transported into the Ar plasma of the ICP-MS instrument by a stream of Ar carrier
gas. Fig. 2.64 gives an example of LA-ICP, showing an LA-ICP-MS spectrum for
Cu-H-MCM-41. The intensity of copper is following the Al-content, suggesting that
copper species are balancing the framework charge, and thus were ion-exchanged
into H-MCM-41. Some minor heterogeneity of the aluminum distribution in the sam-
ples is visible. Intensity fluctuations are also observed in the copper contents and
few intensive peaks in the sodium intensity are observed, indicating that traces of
sodium are still present in the samples that were synthesized first as Na-MCM-41
and then exchanged with protons.
Intensity (counts s–1)

Al

Cu

Na

0 20 40 60 80 100 120 140 160 180

Time (s)

Fig. 2.64: LA-ICP of Cu-H-MCM-41. (Reproduced from [27] with permission from the PCCP Owner
Societies).

2.2.15.4 Magnetic resonance imaging

In spatially unresolved conventional NMR a nucleus of non-zero nuclear spin
quantum number is placed in an external magnetic field. By exposing the system
to an electromagnetic energy of appropriate radio frequency a resonant absorp-
tion occurs between non-degenerate nuclear spin energy levels at a frequency
proportional to the strength of the external magnetic field. The precise energy-
level splitting, specific to a given isotope, is slightly modified by the electronic
environment of the nuclei, making it possible to identify the presence of particu-
lar molecular species.
In order to obtain spatial resolution a spatially varying magnetic field is applied
in addition to the large static field, with the resonance frequency of species within
a sample becoming a function of position and strength of the applied gradient.
 2.2 Toolbox in catalysis 107

In pulse encoding methods, an initial and a subsequent pulsed field gradient

are applied after a certain observation time, allowing the monitoring of the distance
travelled during a known time and thus the measuring of diffusion, dispersion and
flow processes.
Magnetic resonance imaging can be used for micro-imaging of a single catalyst
pellet with a spatial resolution of ~30–50 μm, making it possible to reveal the effect
of the catalytic manufacturing process on the catalytic pellet micro- and meso-scale
structure, for example, tortuosity. At the same time, this spatial resolution does not
make it possible to resolve individual pores in a pellet. MRI can be applied to char-
acterizing metal ion distribution in the pellet during the catalyst preparation and in
the final material, as well as to spatial distribution of coke in a spent catalyst pellet.
It is possible to investigate internal phase distribution and liquid flow fields in-
side fixed-bed reactors and monoliths reactors at spatial resolution of 100–500 µm,
although measurement constrains do not allow for the study of large-scale indus-
trial size reactors.
In the case of trickle bed reactors, two-phase flow investigations make it possi-
ble to measure the values of hold-up and wetting of the catalytic materials, consis-
tent with gravimetric data and model predictions.
Fig. 2.65 shows that full pre-wetting of the bed was followed by immediate es-
tablishment of the trickle flow, yielding significantly larger catalyst wetting and
hold-up compared to other methods of wetting.

Fig. 2.65: MR images of gas and liquid distribution within a

fixed-bed reactor operating at air and water velocities of 66 and
4.6 mm s–1, respectively. The bed was fully pre-wetted. Red,
gray, and black pixels identify water, air, and packing
elements, respectively. (Reproduced with permission from [32]).

In addition to MRI related to catalytic engineering, recently there has been an inter-
est in the imaging of reactions in situ.

2.2.15.5 Positron emission tomography imaging

Positron emission tomography (PET) is now a well established diagnostic technique
in nuclear medicine, generating 3D images of radiolabeled molecules distributed
within living human organs.
108 2 Engineering catalysts

Positron-emitting short-lived radioisotopes such as 11C (half-life 20.4 min) must

be produced on-site, which is done by the irradiation of an appropriate target mate-
rial with energetic beams of protons or deuterons generated by a cyclotron.
Encounters between positrons and electrons lead to their mutual annihilation and
resulting electromagnetic radiation. Rotating a tomograph during a scan (~10–15)
allows for recoding photons emitted over 360° in the plane, making it possible to
have a 3D reconstruction of the imaged object.
Data for 11C methanol transformations in a catalytic bed of a zeolite are presented
in Fig. 2.66. The image was acquired with a small PET scanner, developed by
E. Sarkadi-Priboczki and co-workers from the Institute of Nuclear Research in
Debrezen. Red and yellow colors correspond to higher radioactivity, while green and
blue are related to lower radioactivity concentrations. The 3D image exhibits an inho-
mogeneous distribution of compounds on the catalyst, reflecting non-uniformity of
the catalyst bed in terms of activity.

11C-methanol

3D image

Fig. 2.66: 3D image of methanol conversion on a zeolitic material at 620 K. Courtesy of Dr. E. Sarkadi-
Priboczki.

2.2.15.6 Integrated laser and electron microscopy

In fluorescence microscopy (FM) a focused laser light scans a catalyst, detecting
fluorescence emission from the focal points. The technique (contrary to electron
microscopy) has limited spatial resolution and only reveals fluorescent structures.
The integrated laser and electron microscope (iLEMI developed by B. Weckhuysen
and co-workers [33]) is an imaging tool, combining the strengths of both methods.
In order to generate a fluorescence signal, fluorogenic thiophene oligomerization
is used, catalyzed by the acidic sites of the zeolites. The product, oligomerized
thiophene, emits green fluorescence with the intensity of the fluorescence relating
 2.2 Toolbox in catalysis 109

to acidity. In this way a spatial resolution of ~20 nm can be achieved. The iLEM
set-up shown in Fig. 2.67A consists of a custom-designed laser scanning FM
mounted on a side port of a TEM with the laser beam perpendicular to the path of
the electron beam. For FM imaging the grid with the catalyst sample is rotated to
face the laser beam, while FM is partially retracted and the grid is tilted to enable
TEM imaging.
The technique was applied to study a fluid catalytic cracking catalyst, which
will be discussed in detail in Chapter 4. The active phase, zeolite Y, is embedded in
a matrix consisting of clay, silica, and alumina. Using iLEM, the catalyst acidity,
which is related to fluorescence intensity, could be correlated to the catalyst parti-
cle structure. Two different areas were found in the catalyst, one associated with
the zeolite component giving the fluorescent products, while the other areas are
mainly composed of the matrix material (Fig. 2.67B,C).

TEM TEM

FM FM

(A)

(B) (C)

Fig. 2.67: (A) The iLEM set-up, and iLEM analysis of a sectioned FCC catalyst particle. (B) FM image.
(C) TEM image, taken from the same region. (Reproduced with permission from [33]).

When the zeolitic structure is exposed to steam, Al atoms from the framework can
be extracted, creating extra-framework Al species. This results in lower amounts of
Brønsted acid sites and thus eventual deactivation. The iLEM images of hydrother-
mally deactivated catalysts show much lower fluorescence intensity compared to
the fresh FCC catalyst particles, moreover a large fraction of the crystals is damaged
and the large regions of clay are lost.
110 2 Engineering catalysts

2.2.16 Catalytic reactions: product analysis

Various reactors for studying catalytic reactions will be discussed in Chapter 3.

Investigations of catalytic reactions, in addition to providing information on kinetics,
activity, selectivity and deactivation, are also important in the reaction mechanisms.
Detailed product analysis is thus an essential element of each and every catalytic
study. In vary many cases it can be a bottleneck, as although products might not be
unknown their quantification is necessary. Moreover, rapid analysis may be needed.
A brief overview of chromatographic methods is provided here. Chromatography is a
physical method of separation in which the components to be separated are distrib-
uted between two phases, one of which is stationary, while the other moves in a defi-
nite direction. Chromatographic methods are widely used nowadays for analytical
purposes, not only for off-line analysis but also for on-line determination of minute
amounts, as well as large- scale preparative separations. In fact, not only monomers,
but also polymers and oligomers, can be separated by chromatography, although in
the former case it is essentially a group separation. There are several forms of chroma-
tography that use different mobile and stationary phases, with the two main forms of
instrumental chromatography being liquid (LC) and gas chromatography (GC).

2.2.16.1 Gas chromatography

Gas chromatography (GC) (Fig. 2.68) provides qualitative and quantitative determi-
nation of organic components, introduced into the column for separation either di-
rectly or in the derivatized form.

Retention time:
Time for peak maximum
after injection

Injector Detector
Flow
Gas
control
Columns:
Nonpolar (volatility)
Polar (volatility/polarity)

Columns:
Temperature
i.d. = 0.1–0.75 mm
programming
film = 0.1–0.5 μm
1 = 5–50 m
Oven with GC-column

Fig. 2.68: General scheme of GC. Typically the set-up includes also a liquid injector along with an
evaporator.
 2.2 Toolbox in catalysis 111

Vaporized products passed through the column are identified in a detector,

whose response is recorded as a chromatogram. Thus, in order to be analyzed by
GC, compounds in the samples must be able to get volatilized and additionally pos-
sess thermal stability.
Capillary columns are made of fused silica with a stationary phase as a thin
film of liquid or gum polymer on the inside of the tube. The most common station-
ary phases are siloxane polymer gums with different substituents providing differ-
ent polarities. The polymers are usually cross-linked in the column by photolytic
or free-radical reactions, bringing strength to the polymer films. Wall-coated
open-tubular columns with a liquid-phase coated directly on the inner walls, and
support-coated open-tubular columns are applied. In the latter case a stationary
phase is coated on fine particles, deposited on the inner walls. Among non-polar
columns, those based on dimethyl polysiloxane could be mentioned. HP-5 with
95% dimethyl polysiloxane and 5% phenyl groups is slightly more polar. An even
greater number of polar columns employ polyoxyethylene or polyester liquid-
phases.
Capillary columns are available in a wide range of internal diameters, lengths
and liquid film thicknesses. Although longer columns provide better separation, it
comes at the expense of the analysis time, meaning that this time is increasing,
which is usually not desitrable. In addition, longer columns lead to higher pressure
and thus problems with the injection. Columns with thicker films have higher ca-
pacity, but usually require higher temperature, while thin film columns are suited
for large molecules with low volatility. In principle, the analysis of components
with up to 60 carbon atoms is possible.
Different types of injection systems are used in GC. Split mode, where the in-
jected material after evaporation is split between the column and an outlet, affords
rapid volatilization and homogeneous mixing with the carrier gas. Most of the sam-
ple will pass out through the split vent and only a small proportion will flow into
the column. Splitless systems provide more reliable quantification, allowing analy-
sis of rather high molecular-mass compounds as, for example, triglycerides.
Two types of detectors are mainly applied. Flame ionization detectors (FID) op-
erate in a destructive mode with a current measuring circuit that responds to ions
created by combustion in the flame. These detectors have high sensitivity for hydro-
carbons, but do not detect water. Thermal conductivity detectors (TCD) are non-
destructive measuring changes in conductivity caused by product in the carrier
stream. They are combined with FID.
On-line coupling of capillary columns with mass spectrometers is routine nowa-
days enabling convenient structure identification.
An important issue, which is sometimes forgotten, is that the sensitivity for dif-
ferent compounds is varying for a detector; thus different peak areas are not neces-
sarily in proportion to the weight concentrations. Knowledge of response factors is
therefore necessary and calibration for components should be done properly,
112 2 Engineering catalysts

especially for compounds with various functional groups. Commonly, internal stan-
dard compounds are applied, for example, compounds that are not present in the
sample itself are purposely added. Chemically they should be similar to the sample
compounds with close retention time, however, with no peak overlapping.
What is not that often considered is a necessity to not have any leftovers in the
samples from the previous injections, as heavy compounds, which are difficult to
vaporize, could remain in the GC column and significantly influence subsequent
analyzes. Thus regular control of retention times and response factors, as well as
column cleaning or replacement in due time, should not be overlooked. For some
samples related to the analysis of complex mixtures even pre-fractionation could be
necessary.
In addition to such advantage of GC as accurate quantification based on inter-
nal standards, a possibility to be combined with a mass spectrometer and complete
automation regarding injection and analytical runs, very high resolution should
also be mentioned.
Conversely, only molecules up to about 1,000 mass units can be analyzed, as
they should be stable at high temperatures. Therefore, samples should sometimes
be processed before the analysis and this is important for polar compounds, for ex-
ample, acids, which should be derivatized. GC and GC-MS analysis in the vapor
phase require volatile derivatives that do not adsorb onto the column wall.
Different derivatizations for different substances are recommended and discussed
in the specialized literature on GC.
Recent advanced in gas chromatography are associated with use of multi-
dimensional analysis (two-dimensional or 2D GC). In this version of GC a portion of
the compounds separated on the first column (first dimension) is sent to a second
column (second dimension) for further separation according to different physical
principles. For example, if the first column is apolar with separation through differ-
ent vapor pressure of components mixture, the second column is polar with polar-
ity-based separation. An example of 2 GC is provided in Fig. 2.69.

2.2.16.2 Liquid chromatography

These chromatographic methods use liquids such as water or organic solvents as
the mobile phase. Silica or organic polymers as well as anion-exchange resins are
used as a stationary phase. Separation is performed either at atmospheric pressure
or at a high pressure generated by pumps, which is often called high-performance
liquid chromatography (HPLC) with solvent velocity controlled by high-pressure
pumps, giving a constant flow rate of the solvents. Solvents are used not only as
single solvents but they can also be mixed in programed proportions. In fact, even
gradient elution could be applied with increasing amounts of one solvent added to
another, creating a continuous gradient and allowing sufficiently rapid elution of
all components.
 2.2 Toolbox in catalysis 113

aro. RCP

OP
aro. R

cis-decalin
RCP
sat.

trans-decalin

sat. ROP

Fig. 2.69: GCxGC chromatogram of tetralin hydroconversion products. Aro- aromatics Sat. saturated;
RCP-ring-contraction products; ROP- ring-opening products. (Reproduced with permission from [34]).

The most commonly used columns contain small silica particles (3–10 μm) coated
with a non-polar monomolecular layer. For low-polar compounds the mobile phase is
an organic solvent, while reversed phase HPLC employs mixtures of water and aceto-
nitrile or water and methanol as eluents and is applied for non-ionized compounds
soluble in polar solvents. UV-Vis and diode-array detectors enabling recording of UV-
vis spectra, or detectors based on refractive index monitoring, are applied.
An important form of HPLC is size-exclusion chromatography (SEC), which is
widely applied for the determination of molecular-mass distributions of oligomers
and polymers. In SEC (Fig. 2.70), solutes in the mobile phase (for example, tetrahy-
drofuran) are separated according to their molecular size. Smaller molecules pene-
trate far into the porous column packing material and thus elute later than larger
ones.
Among the advantages of LC are its non-destructive character and absence of
derivatization. This technique can handle both small and large amounts and it can
also be used for preparative isolation of compounds from mixtures. Contrary to GC
there are almost no, or at least much fewer, limitations in terms of the molecular
size. In addition LC can be combined with mass spectrometry, once again without
derivatization.
114 2 Engineering catalysts

Chromatogram

Time

Fig. 2.70: Size-exclusion chromatography.

Thermally unstable and polar compounds can thus be analyzed as such, and
the molecular mass in triple quadruple or ion-trap LC-MS can be up to 3,000 m/z,
while time- of-flight versions allow up to 16,000.
Although LC-MS provides better sensitivity and selectivity than GC-MS and is
excellent for quantification of selected substances in complex mixtures, this tech-
nique is not very suitable for rapid and reliable identification of unknown com-
pounds, mainly because fragmentation is sparse at mild ionization conditions of
ionization. Moreover, spectra libraries that enable identification are typically not
available. Another shortcoming of LC-MS is the rather low sensitivity of the detec-
tors for certain compounds. Finally, it may be difficult to obtain constant pressure,
which in turn influences retention; clean, degassed solvents are needed and it
might be challenging to find the optimum solvent mixture.

2.2.17 Theory as a part of a toolbox

It is important to emphasize that in catalyst development there is now a close cou-

pling between theory and experiments where the theories are based upon quantum
chemistry and describe single molecular situations. The current level of theory can
be applied simultaneously with experiments, but these experiments cannot yet be
replaced by theoretical calculations. The high sophistication of modern theoretical
concepts requires that chemical engineers if not themselves proficient in theory
closely collaborate with theoretical chemists. Quantum chemical calculations pro-
vide valuable information about adsorption as well as activation energy.
Experimentalists, chemists and chemical engineers should, however, be careful
with interpretation of the modeling results because realistic modeling of technical
catalysts is too complicated and often drastic simplifications are included into
calculations.
Electrons are particles with wavelike properties, therefore, they do not behave
like point-like objects moving along precise trajectories. Instead their trajectory is
replaced by a position-dependent wave function (Ψ ). The basis of quantum chemis-
try is the Schrödinger equation, H ^ Ψ = E Ψ where H is the Hamiltonian for the
 2.2 Toolbox in catalysis 115

system (an operator containing derivatives), E is the energy of the system. The prob-
ability of finding an electron at any point in space varies with the distance from the
nuclei and is proportional to |Ψ |2.
Solving the Schrödinger equation is challenging because the motion of any
particle is influenced by all other particles and analytic solutions can be obtained
only for very simple systems, such as atoms with one electron. Thus approxima-
tions, being a trade-off between ease of computation and accuracy of the result,
are needed to be able to treat molecules theoretically. For example, a Born-
Oppenheimer simplification for the many-electron Schrödinger equations divides
electrons into valence and inner core electrons. The inner core electrons are
strongly bound and can be excluded from chemical binding calculations, while
valence electrons completely determine binding properties. In such approach an
atom is reduced to an ionic core interacting with the valence electrons, although
the electronic wave function is still a function of the spatial coordinates of the
electrons and their spin variables. Hartree-Fock approximation expresses the
overall wave function as a product of single particle wave functions. This ap-
proach does not include electronic correlations, which account for interactions of
electrons with each other. Additional correlations have been developed to over-
come these shortcomings of the Hartree-Fock approach.
DFT instead of using many-body wave functions uses electron density as the
basic variable. This theory, which became very popular to study many-body prob-
lem, is simpler than the wave function approach and is much less expensive com-
putationally. In this theory, the properties of a many-electron system can be
determined by using functionals, i.e., functions of another function, namely the
spatially dependent electron density. Although DFT provides an exact path to the
determination of the electronic energy of a system in a ground state, and allows
computation of reaction energies and barriers of many complex catalytic systems,
such calculations are computationally very demanding.
An example of DFT calculations [35] is presented in Fig. 2.71 for steam reform-
ing of methane (CH4 + H2O ↔ CO + H2), which will be discussed in detail in
Chapter 4.
These calculations were done for the Ni (111) surface and the Ni (211) surface,
representing a stepped surface. The Ni (211) step sites have lower activation bar-
riers, are more reactive and have more strongly binding intermediates than the
close-packed Ni (111) surface. In addition to pointing out these differences in
activity, DFT calculations were very useful in understanding deactivation by car-
bon formation. It follows from Fig. 2.71 that as carbon is bound stronger at the
stepped surfaces, they could thus be the potential sites of carbon nucleation.
Visualization of the graphene layers growth on step sites by in situ HREM in com-
bination with DFT resulted in a new approach of catalyst promotion by blockage
of the step sites.
116 2 Engineering catalysts

300
H2O 3H2

200 Ni (111)
CH4
E (kJ mol–1)

CO
100

6H
CH3, H CH2, 2H
0 CH, 3H
C, 4H O, C, 6H
Graphene
Ni (211) OH, C, 5H
–100 CO, 6H
Reaction coordinate

Fig. 2.71: DFT calculations for steam reforming of methane. (Reproduced with permission from [36]).

An alternative approach to complete DFT calculations is to develop descriptors

of catalyst reactivity, which will enable design and engineering of more active and
selective catalysts.
Thus for a methanation reaction, which is a reverse of methane steam reform-
ing, i.e., CO + H2 ↔ CH4 + H2O, the group of Norskov [37] demonstrated that the
activity of a certain metal is given by only two descriptors: the C and O adsorption
energies (Fig. 2.72).
The TOF is plotted as a function of carbon and oxygen binding energies for the
stepped 211 surfaces of several transition metals at 573 K, 40 bar H2, 40 bar CO.
The reason for volcano type behavior is that the breaking of the CO bond is
the rate-determining step for relatively unreactive catalysts such as Ni. It proceeds
by bond breaking the OH intermediate, yielding adsorbed C and OH on the sur-
face. On the more reactive metals (Fe) the surface is poisoned by adsorbed carbon
and oxygen because of the faster reaction. A good methanation catalyst thus
should have a reasonably low CO dissociation barrier and reasonably high carbon
and oxygen desorption rates as methane and water. DFT calculations identified
Ni3Fe catalysts as a cheaper alternative with a higher activity than Ni, and this
was verified experimentally.
Further improvements in DFT are needed to treat, among other issues, intermo-
lecular interactions, especially van-der-Waals forces, as well as transition states,
global potential energy surfaces and some other strongly correlated systems, such
as oxides.
 2.3 Preparation of catalytic materials 117

0
Au 2

Ag –1
–1
Pd
Pt –4
Cu
–2
–7
Rh

log (TOF) s–1

Ni
DE 0 (eV)

Ru Co –10
–3
Ni3Fe
Fe –13

–4 –16
Re

–19
–5
–22

–6 –25
–2 –1 0 1 2 3 4
DE C (eV)

Fig. 2.72: Theoretical volcano for the production of methane from syngas, CO, and H2. (Reproduced
with permission from [37]).

2.3 Preparation of catalytic materials

2.3.1 General overview

For many years, the development and preparation of heterogeneous catalysts were
considered to be more alchemy than science. A heterogeneous catalyst is a compos-
ite material, characterized by: (a) varying amounts of different components (active
species, physical and/or chemical promoters, and supports); (b) shape; (c) size; (d)
pore volume and distribution; (e) surface area. The optimum catalyst is the one that
provides the necessary combination of properties (activity, selectivity, lifetime, ease
of regeneration and toxicity) at an acceptable cost.
Although in the past the main determining factors were activity and selectivity
rather than costs of catalysts, nowadays economics and ecological concerns are of
significant importance. The cost/performance ratio sometimes prevents the use of
new technologies for manufacturing catalysts, as they are usually associated with
additional capital investment. Contrary to academic research aimed at the synthesis
of well-defined structures, which could be carefully characterized with the results
published in scientific journals, the industrial approach to catalyst development
118 2 Engineering catalysts

should take into account a number of issues, which are typically of almost no con-
cern in academic research. In addition to activity, selectivity and preparation costs
we should also mention mechanical and thermal properties (resistance to attrition,
hardness and thermal conductivity), morphology and porosity, stability against de-
activation assuring long-term performance, possibility of regeneration and repro-
ducibility of preparation. As many catalysts should be active for a number of years,
industry is also concerned about the ways of predicting the lifetime. To this end,
acceleration tests are performed where catalysts are aged using, for example,
higher temperatures. This might obviously lead to other side processes and not nec-
essarily correspond to the actual lifetime at other process parameters. Catalyst man-
ufacturers are usually concerned about the patent situation and report (if report)
preparation methods as inventions. Judging which methods are used industrially is
very difficult from the patent literature and in the majority of cases the imple-
mented methods are kept as company secrets. The production technologies are not
reported in the engineering literature except for a few processes with rather vague
descriptions.

From my personal experience even catalyst manufacturing companies personnel not involved in
manufacturing can have restricted access to production facilities.

Catalyst scale-up from the laboratory to an industrial scale is not straightforward for
a number of reasons. Typically, the precursors of industrial catalysts are the salts of
catalytically active metals, sols, oxides and natural minerals. The selection of feed-
stock is determined by their constant chemical and phase composition, absence of
undesired impurities, required humidity, costs, etc. Water is another source of impu-
rities, as it is used for dissolution, dilution, washing, etc. Research in academia on
the contrary uses expensive analytical grade reagents, while industry is working with
technical grade substances. Some unit operations are difficult to scale-up, such as
precipitation, which will be discussed below, thermal activation and catalyst forma-
tion in general. Precipitation on a large-scale could even be related with the order in
which compounds are mixed. Taking all this into consideration, catalysts scaling-up
still requires a lot of empirical expertise and knowledge.
Methods for catalysts preparation are very diverse and typically require several
steps.
The most important of these could be identified as: (a) preparation of the pri-
mary solid by precipitation, impregnation, etc.; (b) processing of these primary sol-
ids (thermal treatment, etc.) and (c) activation (reduction, sulfidation, etc.).
Traditionally, catalysts are divided into bulk and supported catalysts.
Bulk catalysts are produced by a number of methods, such as precipitation, co-
precipitation, sol-gel, flame hydrolysis, spreading or melting. Some of these methods
will be discussed in the corresponding sections. For the impregnation supported cat-
alysts the following methods are used: ion exchange, equilibrium adsorption,
 2.3 Preparation of catalytic materials 119

grafting, deposition-precipitation, immobilization of metal clusters, spreading and

wetting, chemical vapor or atomic layer deposition and the anchoring of homoge-
neous catalysts.

2.3.1.1 Preparation of the primary solids

Precipitation and co-precipitation are mainly used for production of oxidic catalysts
and support materials. Two typical examples of such catalysts are iron oxide cata-
lysts for high temperature water-gas shift, and copper oxide-alumina catalyst for
methanol synthesis. Iron oxide catalysts can be prepared by precipitation of iron
carbonate from iron sulfate in the presence of sodium carbonate with subsequent
washing, filtration and calcination.
The mixing and maturation procedures, pH and pH variations could influence
precipitation, which involves nucleation and crystal growth. Nucleation needs
conditions far from equilibrium (supersaturation), while crystal growth finally
approaches equilibrium.
Low supersaturation, which could be achieved under homogeneous precipitation
conditions when the precipitating agent is continuously supplied or produced in situ,
leads to poor dispersion. Although high supersaturation might lead to well-dispersed
catalysts they could thermodynamically be unstable, undergoing Ostwald ripening
(growth of large crystals at the expense of smaller ones).
Catalyst manufacturers typically use water during precipitation, trying to avoid
organic solvents for precipitation because solvent work-up and recycling are expen-
sive. There, are, however, examples when organo-metallic precursors are used in
organic solvents for the synthesis of supports by precipitation. A process for alumi-
num alkoxide production from aluminum metal and an alcohol, was developed by
Condea Chemie. Subsequent hydrolysis also gives an alumina slurry, hydrogen and
alcohol, which is recycled. Aluminosilicates could be produced by the same
technology.
Water could be completely avoided when dry mixing of catalysts precursors is
performed by mechanical activation. During ball milling the particles of easily de-
composable salts (formates, carbonates, oxalates) are crushed and the mechanical
energy is transferred to heat. When heat conductivity is poor, high temperature hot
spots lead to local solid reactions, resulting in catalysts with poor dispersion and
small specific surface area compared those obtained by precipitation. Iron oxide
based catalysts for dehydrogenation of ethylbenzene to styrene are produced by
mechanical activation when potassium hydroxide is put in contact with chromium
and iron oxides.
Gelation methods are based on the continuous transformation of a solution into
a hydrated solid precursor (Fig. 2.73).
The key factors are hydrolysis and condensation rates, as will be discussed later
when the synthesis of silica is described. Initially, hydrophilic colloidal solutions are
120 2 Engineering catalysts

Dissolve a precursor Start with

(metal salt or alk- preformed sols
oxide) in a solvent

Add water and acid

Destabilize sols by Formation
or base for hydro-
abjusting pH
lysis and conden-
or concentration
sation

Formation of a Formation of a gel

self-supporting gel onto a substate

Aging of the gel

Aging

Evaporative Supercritical
drying Drying of the gel drying
for
solvent removal
Solvent
removal
Formation of a Formation of an
xerogel aerogel

Obtaining different
product forms such Heat
as powder, mono- treatment
liths, thin films and Calcinatiory
membranes sintering

Fig. 2.73: Diagram for the different steps in the sol–gel preparation. (Redrawn from [38]).

generated with the micelles separated because of electric charges. Subsequent gela-
tion (Fig. 2.74) depends on the pH, micelle concentration, temperature ionic strength
and temperature. The sol-gel process produces a metastable open-structure polymer
with the units bound by chemical, dipole and van der Waals forces and hydrogen
bonds. The method allows for better control of the texture, composition, structural
properties and homogeneity of the catalysts.
Removal of the solvent should be, however, carefully done, as drying can result
in either aerogels or xerogels, the xerogels being too dense to have an application
 2.3 Preparation of catalytic materials 121

Supercritical
solvent Aerogel
extraction

Xerogel
Evaporation
Heat
Solution (Hydro)gel
of micelles Dense ceramic

Fig. 2.74: Formation of aerogels and xerogels from micelles.

in catalysis. In addition, phase segregation and porosity changes can take place be-
cause of calcinations, thus the sol-gel method is limited to the production of car-
riers such as silica and alumina and to hydrothermal synthesis of zeolites, which
will be discussed in a separate section.
In selective removal one of the components is removed. An example is the prep-
aration of skeletal metals or sponge catalysts such as Raney nickel. Aluminum is
removed from a relatively course powder of an alloy (NiAlx) by leaching with so-
dium hydroxide (Fig. 2.75A). This procedure results in a highly porous nickel cata-
lyst that possesses high activity and, also important for three-phase reactions, a
high settling rate. Sponges are composed of 2–15 nm microcrystallines that have
been agglomerated into macro pores. The surface area of sponge catalysts (which
were and are still used commercially in various hydrogenation reactions such as hy-
drogenation of sugars and oils to margarine), is 50–120 m2 g−1. The catalysts pre-
pared by selective removal have some limitations. Digestion of the less noble metal

Mixing

Melting

Cooling

Crushing

Screening

Exposure to NaOH
(A) solution (pH ca. 14) (B)

Fig. 2.75: Formation of sponge catalysts. (A) General scheme, (B) Storage of catalysts.
122 2 Engineering catalysts

from a bimetallic or polymetallic alloy results in highly pyrophoric catalyst, which

starts to burn spontaneously in air when dried and should be thus kept and trans-
ported under water (Fig. 2.75B). Difficulty in process control also means challenges
in the reproducibility of the preparation. In addition, the overall costs are high be-
cause of the high-energy consumption needed for melting.
In supported catalysts catalytically active compounds are attached to a support
with a large specific surface area. The preparation of these catalysts includes depo-
sition of the precursor on the support surface and transformation of the precursor
in the required active phase (metal, oxide, sulfide). The precursor type and concen-
tration, nature of the support, presence of promoters, and conditions of the prepa-
ration influence chemical interactions of the precursor as well as the final active
phase interactions, with the support being decisive for the performance of the cata-
lyst in terms of activity and selectivity.
Impregnation is one of the methods of depositing the active phase on a support
and could be performed by contacting a certain volume of solution with a precursor
and the support (Fig. 2.76).

Support,
often porous grains

Dipping Impregnated Active

support catalyst

Drying Reduction
Metal salt Calcination
solution
(A)

Desorption
Adsorption Evaporation
Adsorption

Solution flow Adsorption/desorption Drying

(B) into pores + diffusion

Fig. 2.76: Impregnation of supports. (A) The overall process and (B) processes in a pore.

In wet impregnation an excess of solution is used, which is eliminated by evapora-

tion or draining. Deposition is slow and can take hours or even days, therefore, there
is an apparent danger of surface restructuring. Moreover, quantitative deposition of
the precursor is difficult to achieve. In pore volume impregnation the volume of the
 2.3 Preparation of catalytic materials 123

solution containing the precursor corresponds to the volume of the pores. A similar
method is incipient wetness impregnation, the difference being that the volume of
the solution is more empirically determined for the catalyst to start looking wet.
For both methods the operating variable is the temperature, which influences
both the precursor solubility and the solution viscosity and as a consequence the
wetting time. The concentration profile of the impregnated compound depends on
the mass transfer conditions within the pores during impregnation and drying.
The maximum loading in these methods is limited to the solubility of the pre-
cursor in the solution. Several active components could be introduced in successive
impregnations with drying and even calcinations between these impregnations.
Ion exchange is based on replacing an ion in electrostatic contact with the sup-
port by another ion. The latter ions gradually penetrate into the pore space of the
support, while ions initially present in the catalyst pass into the solution until equi-
librium is established, corresponding to a certain distribution of the two ions be-
tween the solid and the solution. In synthesis of zeolites for example, a sodium ion
is replaced by ammonium during ion exchange, followed by the subsequent calci-
nation and finally resulting in proton forms of zeolites.
Anchoring of a precursor from an aqueous solution can be done by adsorption
(Fig. 2.77), which occurs on charged surfaces, as frequently used supports like
metal oxides and active carbons develop a pH-dependent surface charge. The metal
precursor should thus be selected based on the surface charge, which depends on
the isoelectric point of the oxide (pH at which the surface is neutral), pH and solu-
tion ionic strength.

Support
adsorbed
Amount

Drying
and firing

Concentration

Fig. 2.77: Adsorption.

Precipitation onto the support surface (deposition-precipitation), which requires

lower supersaturation than in bulk precipitation, should be carefully performed to
avoid precipitation in the solution bulk (Fig. 2.78). A well-dispersed and homoge-
neous active phase is reached when the OH– groups of the support interact directly
with the ions present in the solution. A constant level of supersaturation is then
124 2 Engineering catalysts

Change of T, pH
or concentration

Decrease solubility of metal

Metal salt solution

Fig. 2.78: Precipitation.

achieved, for example, by decomposing a suitable substance that releases the pre-
cipitation agent (OH–) continuously, as in the case of urea hydrolysis. Precipitation
with urea starts with dissolving it in water and its decomposition slowly at ~90 °C,
giving a uniform concentration of OH– in both the bulk and pore solutions. Thus
the precipitation occurs evenly over the support surface, making the use of urea the
preferred method for amounts higher than 10–20%. Another option is to have a
controlled and progressive addition of this precipitation agent. The use of porous
supports might lead, however, to deposition-precipitation at the external surface.
A particularly interesting example in this respect is preparation of nanosized gold
particles, which are very different catalysts from “bulk gold” in terms of their reactiv-
ity. The deposition precipitation method using urea was reported by Fokin already in
1913 [39] without certainly measuring the size of gold deposited on asbestos (!) and
utilized in oxidative dehydrogenation of methanol to formaldehyde. In fact, gold on
this support was even more active than supported silver, although utilization of sup-
ported silver catalysts became an industrial reality. This paper of Fokin written in
Russian language remained in oblivion even if it was cited in a textbook of Sabatier
published few years after the original discovery. Tremendous interest in catalysis by
gold is mainly due to much later work of Haruta and Hutching. The latter scientist
while working in industry found that gold was the best catalyst for acetylene hydro-
chlorination giving vinyl chloride monomer. Currently, carbon-supported mercury
chloride is used commercially. For example, only in China approximately 13 million
tons per year are produced in more than 90 plants with up to 100 reactors per plant
having each of about 6–7 tons of catalyst per reactor. From the overall capacity of
15,000 tons, approximately 1,000 tons per year are simply lost because HgCl2 is vola-
tile. There are currently efforts to replace this catalyst with 0.1% Au/C extrudates, re-
sulting in essentially the same productivity.
Monolithic catalysts are produced by two main methods. In the first, catalytic
compounds are incorporated into the monolith before formation of the monolithic
structure. A significant amount of the active compound is not accessible because of
a long diffusion path. Deposition of the active phase directly on the monolith is
also possible. As ceramic monoliths have low surface area wash-coating is an alter-
native way of active phase incorporation. This consists of depositing a layer of a
 2.3 Preparation of catalytic materials 125

high surface oxide, carbon or polymer onto the surface of a low surface area mono-
lith. Introduction of the active phase is done either during wash-coating or after it
by methods already presented above, such as impregnation, ion exchange, precipi-
tation, deposition-precipitation. Specific methods could be also applied, such as
thermal spraying, spin coating or chemical vapor deposition (CVD). CVD is an ex-
ample of a gas deposition method, where deposition is done by adsorption or reac-
tion from the gas-phase. Another popular method is atomic layer deposition, which
will be discussed in detail later. These methods of gas deposition result in catalysts
with controlled metal dispersion, which could be very high at significant loadings.

2.3.1.2 Treatment of intermediate solids or precursors

Treatment of intermediate solids includes filtration, washing, drying, thermal de-
composition, calcinations and activation. Activation could also be done after form-
ing the catalyst (methods briefly mentioned in Section 2.3.1.3 and described in
further detail in Section 2.3.7).
In the case of cake filtration (Fig. 2.79), a relatively thin filter medium is used and
separation is achieved on the upstream side of the medium. The particles should be
either larger than the pores of the medium or should form bridges to cover the pores.
A cake is formed from successive layers of solid deposits apparently increasing the
pressure drop for the case of dead-end filtration.

Feed

Permeate Fig. 2.79: Schematics of dead-end cake filtration [40].

An increase in the cake height during deposition of solids apparently results in an

increase in the total cake resistance. In an incompressible cake with a linear in-
crease of resistance with height, the filter cake porosity remains constant while the
cake volume increases. Filtering at constant pressure as implemented in vacuum
filtration results in decline of filtration rates as the filter cake is growing in size. Too
thick filter cakes lead to prolong filter cycles because of low filtering, dewatering
and washing rates. Steps in a filtering cycle include besides filtration per se also
126 2 Engineering catalysts

dewatering, washing, filter cake discharge, cleaning, reassembly and filling of the
filter.
In continuous large-scale vacuum filters, the suspension is introduced to the
filter at atmospheric pressure. Vacuum applied on the filtrate side of the medium
creates the driving force for filtration. The rotary vacuum drum filter is illustrated in
Fig. 2.80.

Washing water

Dew
ate
cake rin
g
on

zo
cti

ne
su

suc so
tion lid
pro
du
central ct
ion duct
suct

knife
on

suct
cti
su

on i

Filtration zone

Fig. 2.80: Schematic of a rotating vacuum filter [41].

As the drum rotates, being partially submerged in the slurry, solids trapped on the
drum surface are washed and dried. The cake discharge occurs at the end of the
rotational cycle. The drum surface covered with a cloth filter medium can be pre-
coated with a filter aid to improve filtration and increase cake permeability.
Horizontal filters such as the horizontal belt filter presented in Fig. 2.81 allow
settling by gravity before the vacuum is applied. Horizontal belt filters having a sim-
ple design and low maintenance costs have difficulties to handle very fast filtering
materials on a large scale.
The filters described earlier operate in a continuous mode. Nutsche filters,
which are basically vessels divided into two compartments with vacuum applied
to the lower compartment, can operate batch-wise. This might be necessary in
case of stringent washing requirements or if there is a need to keep batches sep-
arated. The Nutsche (Fig. 2.82) filters can handle batches of 25 m3 and a cake
volume of 10 m3 and are able thus to work with an entire charge of slurry.
Sufficient holding volume is required for fast charging and emptying of the ves-
sel. The difficulties of operation with such filters arise when cakes are slow to
form, sticky and the product deteriorates during long downtime. The operational
 2.3 Preparation of catalytic materials 127

Filter cloth
Slurry feed device Cake washing device Cloth drive roller

Vacuum tray drive

Cloth tension device

Cloth tracking device
Cloth washing
device

Vacuum pump

Fig. 2.81: Schematic of a horizontal belt filter [42].

Pressure
Filtration
drying

Cake Vacuum
washing drying

Cake
Repulping discharge

Fig. 2.82: Nutsche filter: (a) Photo [43] and (b) schematics of operation [44].

sequence starts with filtration per se when the filter is charged with slurry and
the pressure is applied. In the washing stage, the wash liquid is introduced over
the cake displacing the mother liquor. In the drying stage, air or gas purges the
cake until the desired drying level. The final step is the cake discharge and in
some instances washing the cloth or woven mesh screen with water to remove
any cake residue.
Washing is essentially a replacement of the mother liqueur with pure water,
with desorption or exchange of certain undesirable ions for those that could be
128 2 Engineering catalysts

decomposable during calcinations. Thus, for example, a known catalyst poison like
Cl– is washed away by water or replaced with NO3− .
Drying, which is the elimination of the solvent (usually water) from the pores of
a solid, is done at temperatures higher than critical temperature of water or by evapo-
ration under vacuum at low temperatures (from −50 °C to −5 °C). Inappropriate dry-
ing can result in non-uniform distribution along the catalyst grain of an active
component not strongly bound to the surface. Slow drying or freeze drying can be
used to avoid such problems. Drying of powders with the channels above 10 μm is
faster than for smaller ones. The drying rate depends on the catalyst shape being
faster for spherical particles than for cylindrical ones.
The objective of calcinations is to develop a well-defined structure for the active
agents/supports, modify the support structure by, for example, transforming γ-
alumina to α-alumina, adjust the texture and pore volume and to provide mechani-
cal resistance.
Calcination means heating without the formation of a liquid-phase and is a fur-
ther heat treatment beyond drying. It can determine the properties of the final cata-
lyst, as many processes are occurring, such as decomposition of the precursor,
spreading, sintering, and formation of new phases from the active material and the
support. Calcination is specific for each system and depends on the gas-phase.
Although air is often used, inert gases or controlled water vapor atmosphere could
be also applied.
Typically, however, calcination is carried out in air at temperatures higher than
those used in the catalytic reaction or catalyst regeneration.
In a chain-like belt, calciner gases can circulate readily through the belt and
the catalyst of a fairly larger size in the form of extrudates or pellets. Because for
catalyst shaping, lubricants are added to aid extrusion, they are burned out during
calcination, which might lead to excessive temperature increase. A remedy in such
cases is to limit the oxygen content or alternatively to perform calcination in a tem-
perature-programmed mode to control oxidation of combustible materials present
in the catalyst.
The word “calcination” is used because of historical reasons, as burning of cal-
cium carbonate (limestone) to calcium oxide (quick lime) proceeds with removal of
CO2 (i.e., a process obviously beyond drying). A limestone calciner (rotary kiln) is
shown in Fig. 2.83.
Apart from loss of chemically bonded H2O or CO2 several other processes occur
during calcination or heating, such as modification of the nature and/or structure of
the phases present, generation of the active phase and stabilization of mechanical
properties and texture modification through sintering. The heating temperature and
atmosphere must be properly chosen to obtain phases that are stable in the reaction
and regeneration conditions, at the same time avoiding excessive sintering phenom-
ena (agglomeration of crystals) as much as possible, as they have a negative effect on
the catalytic performance. In order to understand at which temperature sintering can
 2.3 Preparation of catalytic materials 129

Fig. 2.83: Panoramic view of rotary kiln (photos of the author).

occur for a particular metal it is important to know the Tamman temperature TT and
Hüttig temperature TH when lattice and surface atoms, respectively, become mobile.
Semi-empirical relations between these temperatures and the melting temperature
Tmelt

TT = 0.5 Tmelt , TH = 0.3Tmelt (2:23)

where T (K), indicate that when the melting temperature is not as high as for some
metals (Cu: 1,357 K, Ag: 1,234 K) sintering can be profound at temperatures typically
used for the thermal treatment of catalysts.
Technical arrangements of drying, calcination and activation should be per-
formed in a way that all the catalyst particles are exposed to the same conditions,
which is not possible to achieve in fixed beds. One should also realize that condi-
tions of these treatments and applied equipment in the lab and industrial scales are
very different. For example, different types of moving beds are used in industry
(Fig. 2.84).
Activation refers to other thermal treatments, such as reduction of metal sup-
ported catalysts or sulfidation of hydrodesulfurization catalysts, usually performed
in special atmospheres in the reactor at the start-up of the unit. Therefore, activa-
tion is not considered strictly to be a preparation procedure. Reduction is mostly
done with hydrogen and a higher dispersion is obtained with a higher reduction
rate unless nucleation is a limiting factor. Some catalyst manufactures do not have
hydrogen at their production facilities and use chemical reductants like formic
acid. It should be noted that reducibility of supported metals might be different to
bulk materials and the preparation of bimetallic catalysts could be complicated be-
cause of a possible preferential reduction of one metal before the other. In Fig. 2.85
hexachloroplatinate is used as an example of the complexity of calcinations and
the reduction processes.
130 2 Engineering catalysts

Fig. 2.84: Calcination. (From [45].)

[Pt(C4H7)2]s Ptdisp
300 °C
110 °C

OH
HCO
300

ion
uct 0°
C α-[PtO2]s
C

red 26
°C
0°

itu
In s
29

+ 0 °C

60
0
50
CH

°C

300 °C 500 °C
[PtIVCl6]s2– [PtIV(OH)x Cly]s [PtIVOx Cly]s β-[PtO2]s
C
70 C
60 CH

0°
0°
0°

60
+

O2
H2 Ptcryst

Fig. 2.85: Effect of pre-treatment on Pt/alumina.

Because of the presence of catalyst particles in technical catalysts with different

location depth, crystalline size, interaction strength with the support, etc., activa-
tion rate is usually diminished as the process proceeds and some catalytic species
might not be properly activated. The degree of activation (for example, degree of
metal reduction) should be examined, which might not be easy when such reduc-
tion is done at a plant site in a catalytic reactor.
 2.3 Preparation of catalytic materials 131

2.3.1.3 Forming
When a powder in a water suspension is fed through a nozzle that is spraying small
droplets into hot air, such spray-drying process results in 10–100 μm particles of an
almost identical shape. They could, in principle, be used in fluidized bed reactors.
As discussed in Chapter 1 such small particles result in large pressure drops when
applied in fixed beds, therefore forming operations are required.
Granulation of humidified powder could be performed in a horizontal rotating
cylinder or a pan rotating around a 45° axis. Prior to this the powder is crushed,
ground (in the presence or absence of a liquid, typically water) and screened. In
granulation, different-sized particles are put into motion with a spray of liquid, re-
sulting in particles of 1–20 mm with a broad size distribution and high-pressure
drop. The slurry might contain binders for better adhesion and layer-by-layer
growth of particles that are predominantly spherical after granulation.
Extrusion is the most economic and commonly applied shaping technique for
catalysts and supports. During extrusion a wet paste from a hopper at the top is
forced through a die and the emerging ribbon that passes through holes in the die
plate is cut to the desired length using a suitable device (Fig. 2.86).

Slurry

Die plate
Extrudate

Fig. 2.86: Principle of extrusion.

Usually the catalyst powders obtained after the thermal treatments behave like
sand, i.e., do not have by themselves the required moldability and plasticity, even
when water is added. Various additives are used in the formulation of pastes,
such as: (a) compounds for improving the rheological behavior (clays or starch);
(b) binders (alumina or clays); (c) peptizing agents to de-agglomerate the particles
(dilute acetic or nitric acid); and (d) combustible materials to increase the porosity
(carbon black, starch, etc.).
The operating variables include mixing time, additive content, water content,
aging and extrusion temperature. The quality of the extrudates also depends on the
drying and calcination procedure. Special shapes (trilobates, rings, hollow cylinders,
132 2 Engineering catalysts

monoliths or honeycombs) can be obtained using proper dies. Press extrudates for
viscous pastes and screw extrudates for less viscous ones are mainly used.
Pellets or tablets of few a mm are obtained by dry tableting, which is more
expensive than extrusion. A dry powder is pressed between two punches in a
press with a pressure up to 30 MPa. Small cylinders and rings are formed because
of this compression. Plasticizing agents, such as graphite, stearic acid or talc,
binders (alumina, clays) or porosity additives (i.e., polymer fibers) could be added
during tableting to achieve the desired properties.
More details on the forming methods will be given in Section 2.3.7.

2.3.2 Unsupported metals

2.3.2.1 Colloidal catalysts

Synthesis of metal blacks can be done by reduction of aqueous solutions of their
salts with hydrogen, formaldehyde formic acid and hydrazine. This procedure re-
sults in large metal particles having a surface composed mainly of face atoms.
Polymer stabilized metal nanoparticles do not undergo coalescence and agglomera-
tion (Fig. 2.87).

Repulsion
Metal
Stabilizer layer nano-
Metal particle
nano-
particle

Fig. 2.87: Principles of nanodispersed metal nanoparticles.

Metal precursors, surfactants, solvents and reducing agents are needed for the syn-
thesis of nanoparticles by colloidal methods. Typically, chlorides, nitrates, sulfates,
and acetates are selected as metal salts. Moreover, two or more precursors can be
applied to generate bimetallic nanoparticles.
Nanoparticles can be prepared, for example, by microemulsion methods.
Microemulsion is a system containing water, oil and a surfactant, with the dis-
persed phase having monodispersed droplets of the size 5–100 nm. The size of
water droplets in a continuous oil phase for water-in-oil microemulsions, surfactant
concentration and the nature of the precipitation agent are the parameters regulat-
ing the properties of nanoparticles. Nanocolloids can be synthesized by mixing
 2.3 Preparation of catalytic materials 133

microemulsions containing the metal precursor and the precipitating agent or the
reducing agent. Alternatively, the precipitating agent can be added directly into the
microemulsion containing the metal precursor. Surfactants, such as various organic
molecules and polymers, ensure colloidal stability of nanoparticles. The variations
of the precursor, reductant, surfactants, additives and conditioning are used to con-
trol the shape and size of the nanoparticles. Cationic or anionic surfactants can be
applied. Cationic cetyltrimethylammonium bromide or chloride is used together
with n-hexanol as a co-surfactant. Water-to-surfactant ratio is a parameter that con-
trols the size of particles, for example, larger particles are formed when this ratio is
higher.
Poly (N-vinyl-2-pyrrolidone) (PVP) is one of the polymers that is used effi-
ciently to control the size of transition metal nanoclusters by varying its concen-
tration: larger concentration leads to smaller nanoclusters. Polyvinylalcohol
(PVA) is used for synthesis of gold nanoparticles, which could be thereafter de-
posited on a required support. This method is particularly useful in synthesis of
gold catalysts supported on carbon, when other preparation methods result in
poor metal dispersion.
Formation of nanoparticles can be explained by simple nucleation and growth.
Prior to nucleation, in a super saturation step the concentration of monomers in-
creases up to a point of spontaneous generation of clusters. This results in a decline
of monomer concentrations, preventing nucleation of new particles. The particles
that are available in the solution start to grow. As growth can also occur during the
nucleation phase, short nucleation phases and slow growth kinetics are required,
to allow for narrow cluster size distribution.
The most favored shape with a minimum surface area and an interfacial free
energy for an fcc structure of metal nanoparticles is truncated octahedron. By intro-
ducing special additives it is possible to achieve not only the size but also shape
control. This is because of the different affinity of surfactants to different crystal
facets hindering growth of a facet with preferential surfactant binding. The shape
of Pt nanoparticles could be changed by varying the concentration of the surfac-
tants and through a control of reduction kinetics (Fig. 2.88).
One of the challenges in the colloidal preparation method is the removal of sur-
face capping molecules that could deteriorate catalytic activity. Determination of
metal cluster size (metal dispersion) by TEM and probe gas chemisorption can thus
give very different values. This is then reflected in potentially erroneous calcula-
tions of TOF. Stability of nanoparticles at high temperatures against sintering
should be also ensured. The size of particles could also change during a catalytic
reaction because of Ostwald ripening.
The shape purity of resulting nanoparticles still needs to be improved. In addi-
tion the size of particles is often rather large, exceeding 10–20 nm with conse-
quently low efficiency. Moreover, the stability of shape-controlled nanocatalysts
during the reduction and prevention of leaching are unresolved issues. New
134 2 Engineering catalysts

1. PVP
1. TTABr 2. NaBH4
2. NaBH4 3. H2

1. TTABr
2. NaBH4
3. H2

Fig. 2.88: Schematics illustrate a generic synthetic procedure for preparing Pt nanoparticles
with cube, octahedron and cuboctahedron shapes (TTABr, tetradecylammonium bromide; PVP,
poly-vinylpyrrolidone. (Reproduced with permission from [46]).

developments in the application of colloidal methods in catalyst preparation are

related to the formation of core-shell, hollow and multi-branched (dendrimatic)
mono, bi- and multi-metal nanoparticles. More details on the synthesis of precious
metal nanoparticles are provided in the review by Jin [47].

2.3.2.2 Bulk metal catalysts

Only a few catalysts belong to this group of unsupported bulk metal catalysts.
Synthesis of sponge catalysts has been briefly addressed before. Another important
example is an iron catalyst for ammonia synthesis (Fig. 2.89). Melting of iron ore in
an electric furnace occurs at 1,600 °C and during this process promoters such as
alumina and silica are introduced into the furnace, while other promoters (CaO and
KNO3 or K2CO3) are added in the subsequent chill cast step in a shallow tray when
oxygen is fed at 1,600–2,000 °C.
The melt is first cooled down and then the catalyst is crushed, separated from
the undersized particles and screened. Oversized particles are further crushed,
while smaller than required particles are recycled to the casting stage or melted in
another furnace at 1,600–1,800 °C. This melt is cooled, then crushed and sieved
into several size fractions (for example, 6–10 mm; 1.5–3.0 mm, etc.) depending on
the operation requirements. The irregular granules of iron catalysts are formed
using this method when the melt is passed through a hole and is put in contact
with a cooling liquid. The catalyst mix crystallizes rapidly when cooled below a
temperature of 570 °C to form a solid solution of wustite and alumina in the magne-
tite crystal lattice with a rather small surface area (1–2 m2 g−1).
 2.3 Preparation of catalytic materials 135

3 10

2 Fe
4
O2
11
1 8 12
7

8
9
5
6

13

Fig. 2.89: Scheme for iron oxide synthesis, 1, screen; 2, drum; 3, lift; 4, furnace; 5, feeder; 6,
crushing; 7, oxidation; 8, cooling; 9, bunker; 10, cyclone; 11, bunker of undersized fines; 12,
furnace; 13, storage of catalyst. (Redrawn from [48]).

In general, parameters such as holding time, gas phase nature, efficiency of

mechanical mixing and homogeneity of temperature in the melt are important to
achieve homogeneity of the melt composition. For alloy catalysts, the final compo-
sition is also controlled during cooling the melt by regulating the rate of cooling,
crystallization kinetics and uniformity of temperature across the bulk of the melt.
Slow cooling results in the thermodynamic equilibrium composition, while rapid
cooling (> 100 K/s) gives metastable amorphous, glassy phases. The latter can be
transformed into nanocrystalline materials, which may be also metastable in com-
position and catalytically active. Such metastability is desired for fused iron oxide
catalysts where unique crystallographic states cannot be prepared by the precipita-
tion method followed by calcination. Rapid cooling also influences the formation of
smaller crystallites.

2.3.3 Preparation of bulk oxides by precipitation

Precipitation of metal salts or hydroxides is a widespread method to produce bulk

oxides allowing for more uniform mixing and uniform distribution of active species.
Methanol and Fischer-Tropsch syntheses catalysts are examples of catalysts pre-
pared by precipitation. This method gives more control over pore size and pore size
distribution as it is not limited to the formation of a support. Thus, catalysts of dif-
ferent size and shape could be prepared. Nitrates are usually the preferred precur-
sors because of their high solubility in water, availability and comparatively low
136 2 Engineering catalysts

price. However, careful control of NOx emissions is required. Other salts, such as
sulfates and chlorides could be also applied, although their disposal could be chal-
lenging. As mentioned above, in special cases in industry more expensive organic
compounds are used for the manufacturing of supports. In addition to mixing com-
ponents well on an atomic scale, precipitation can afford high loadings at relatively
high dispersion. At the same time, low loadings give low dispersion.
Despite the apparent superficial simplicity of the method, a number of parame-
ters (Fig. 2.90) influence the quality of the final product, therefore even if the chem-
ical composition of a particular catalyst could be the same for several industrial
manufacturers, properties could be still very different.

Solvent Mixing
sequence
Supersaturation
Textural Temperature
properties; Precipitate
Particle size; composition;
rate of crystallinity Phase;
homogeneity
precipitation textural
properties

Precipitating Phase; Precipitate Textural

agent homogeneity Additives
properties

Purity;
Morphology; Phase; crystallinity;
textural Phase purity; textural
properties precipitate properties
composition

Anion pH Aging
Solution
composition

Fig. 2.90: Parameters influencing precipitation. (Reproduced with permission from [38]).

Synthesis complexity also explains why the scaling up of precipitation is difficult.

For the methanol synthesis Cu/ZnO catalyst, changing precipitation conditions
many result in different phases with varying catalytic activity. The precipitation
method requires large consumption of chemicals and gives high amounts of process
water.
For easily soluble compounds a classical theory of crystallization from oversa-
turated solutions can be applied in order to understand the mechanism of precipita-
tion. After forming a supersaturated solution precipitation is initiated by physical
or chemical means, resulting in nucleation followed by the crystal growth. The
 2.3 Preparation of catalytic materials 137

precipitate formed after aging and modification undergoes spray-drying or is first

filtered and then dried. The dried material is further calcined and then shaped.
These operations could be also performed in the reverse order.
The processes occurring during precipitation are shown in Fig. 2.91.

Precipitate
Metal ion
Concentrations

Nucleation
regime

Nucleation threshold

Equilibrium
Precursor
concentration

Time

Fig. 2.91: Dependence of concentration on time during precipitation.

During dissolution of a salt the following processes occur:

 m +
Mem + + nH2 O ! MeðH2 OÞn
 m + h im − p
MeðH2 OÞn + pH2 O ! MeðH2 OÞn − p ðOHÞ + pH3 O +

Hydrolysis products are polymerized into complexes of [Me(OH)m]np+ type, where n

depends on hydrolysis conditions and the nature of the metal. Intensive mixing im-
proves the rate of dissolution. Another important parameter is temperature. For
most substrates, such as chlorites, nitrates, ammonium salts, solubility in water im-
proves with temperature increases. However, several salts, such as sodium sulfates
and carbonates display maximum solubility as a function of temperature.
Precipitation, happening after dissolution of the precursors, consists of nucle-
ation and crystal growth. The catalytically active phase is typically metastable, thus
crystallization, which is enhanced at low temperatures, should be performed at
conditions far from equilibrium.
In general, kinetics of solid–solid phase transformations tends to be much
more slower than liquid–solid transformations because of higher activation en-
ergy related, among other reasons, to larger solid–solid interfacial energies,
smaller changes of free energy per unit volume and slower solid-state diffusion.
As a consequence, solid–solid phase transformations rarely reach equilibrium,
138 2 Engineering catalysts

resulting in the formation of metastable phases as mentioned earlier. Nucleation

and growth depend not only on temperature and supersaturation, but also on
impurities.
Supersaturation must first be achieved independent of from which phase crys-
tals can be grown. The degree of supersaturation is expressed by the concentration
difference Δc:

Δc = c − c* (2:24)

where c is the actual solution concentration and c* the equilibrium saturation

value. The supersaturation ratio S defined as

S = c=c* (2:25)

is commonly used in explaining crystallization.

The fundamental dimensionless driving force for crystallization is expressed
through the chemical potential difference of a given substance in the transferring
(solution) and the transferred (solid) states

Δμ=RT = lnða=a*Þ = ln S (2:26)

where a* is the activity of a saturated solution, implying that

S = expðΔμ=RTÞ (2:27)

Solubility–supersolubility diagram (Fig. 2.92) is often used to illustrate how concen-

tration depends on temperature.
Concentration

Supersolubility
curve Solubility
curve

Spontaneous
nucleation Supersaturatd
solution
MSZW

Unsaturated
solution

Temperature

Fig. 2.92: Solubility–supersolubility diagram with MSZW denoting the metastable zone width.

While position of the lower equilibrium solubility curve can be accurately deter-
mined, location of the upper supersolubility curve is less clear as it is influenced by
 2.3 Preparation of catalytic materials 139

many factors such as the rate of supersaturation, agitation intensity, presence of

crystals or impurities. The metastable zone width visible in Fig. 2.92 is thus charac-
teristic of a particular crystallization system.
Nucleation may occur by means of homogeneous or heterogeneous mecha-
nisms, either through primary (absence) or secondary route (presence of crystals).
In primary homogeneous nucleation, nuclei are generated from the solution, while
for heterogeneous nucleation, the solid surfaces are involved.
For homogeneous nucleation, the most likely shapes are spheres as this shape
minimizes the amount of interfacial area per unit volume. The total Gibbs energy
change, ΔGtot, involved in forming a spherical nucleus of radius r is defined
through
4 3
ΔGtot = ΔGvolume + ΔGint = πr ΔGV + 4πr2 γ (2:28)
3

where ΔGV is the free energy per unit volume released upon creating the
new second-phase particle. This term is negative to allow a phase transition. The
interfacial energy per unit area γ is positive as energy is expanded when an inter-
face is created. This term is associated with the creation of new interfacial area.
The total free energy change caused by agglomeration can be expressed as a
function of the nuclei number (Fig. 2.93).

ΔG ΔG
a
ΔG*

ΔG* b

0 0
n* n r* r
c
(A) (B)

Fig. 2.93: Dependence of Gibbs energy as a function of nuclei precursors: (A) amount and
(B) radius.

Fig. 2.93(A) demonstrates the influence of nuclei number on Gibbs energy, which is
composed of the difference in free energy between the solution and bulk species,
free energy change related to formation of the interface as in eq. (2.28) and other
changes, for example, because of strain or the presence of impurities. Obviously,
for agglomeration to occur the Gibbs energy should be negative, which happens
when the critical number of nuclei is reached. This number n* depends on the su-
persaturation ratio S. Without saturation (case a in Fig. 2.93A) nuclei are not
140 2 Engineering catalysts

formed. For higher supersaturation (compare cases b and c in Fig. 2.93A) there are
more nuclei with a smaller size, resulting at the end in a precipitate with a smaller
size. Therefore, in order to increase the number of nuclei, concentrated solutions of
precursors should be used.
An important concept in crystallization is the critical crystal size r* (0.5–5 nm)
illustrated in Fig. 2.93B, starting from which there is a fast growth of nuclei result-
ing in a large number of crystals of different sizes. For lower r* the degree of su-
persaturation should be higher. Beyond the critical nucleus size r*, the total free
energy of the system is decreasing and it is energetically unfavorable for nuclei
to continue growing, while above the critical size (r > r*), there is a continuation
in growth.
The critical nucleus size r* can be determined by taking the derivative of eq. (2.28)
and setting it equal to zero

∂ΔGtot ∂4πr3 ∂4πr2

=0= 3 ΔGV + γ = 4πðr*Þ2 ΔGV + 8πr*γ (2:29)
∂r ∂r ∂r
Giving

r* = − 2γ=ΔGV (2:30)

When nucleation occurs heterogeneously at specific sites (surfaces, interfaces or

grain boundaries), the total free energy associated with the nucleating phase is
4 3
ΔGtot = ΔGvolume + ΔGint = πr ΔGV f ðθÞ + 4πr2 γf ðθÞ (2:31)
3

which is different from homogeneous nucleation by a factor f(θ), accounting for ge-
ometry of the cap compared to a sphere.
*
The critical nucleus size rhet for heterogeneous nucleation calculated in a simi-
lar fashion as for homogeneous nucleation, that is, setting ∂ΔGtot =∂r = 0, gives the
*
same expression rhet = − 2γ=ΔGV as for homogeneous nucleation (i.e., eq. (2.30)).
For heterogeneous nucleation, activation barrier, ΔG*het, is different from ho-
mogeneous nucleation and is defined as
4
πðr*Þ3 ΔGV f ðθÞ + 4πðr*Þ2 γf ðθÞ =
ΔGtot = ΔGvolume + ΔGint =
3
    (2:32)
4 2γ 3 2γ 2 16πγ3
= π − ΔGV f ðθÞ + 4π − γf ðθÞ = f ðθÞ
3 ΔGV ΔGV 3ðΔGV Þ2

Concentration of nuclei (n*) that can be formed at any given temperature T is calcu-
lated through the change in ΔG*:

n*het = n0 exp − NA ΔG*het =RT (2:33)
 2.3 Preparation of catalytic materials 141

BecauseΔG*het strongly decreases with decreasing θ, n*het increases exponentially

with decreasing θ explaining why concentration of viable heterogeneous nuclei is
orders of magnitude higher than the corresponding value for homogeneous nuclei.
The size of the critical nuclei is often expressed through the Gibbs-Kelvin equa-
tion [49]:
 * 31
* 3n v 2γv
r = = (2:34)
4π kT ln S

where n is the number of precursor species, v is the molecular volume of precursor,

γ is the interfacial energy, and S is the supersaturation ratio.
From eq. (2.35) it can be concluded that the size of the critical nuclei depends
on ln S. It is, however, difficult to control nucleation by alteration of the supersatu-
ration ratio. A more efficient way is through the interfacial energy (surface tension).
The size of the nuclei is smaller for higher supersaturation and smaller surface
tension (interfacial energy), and the latter could be affected by solution pH, ionic
strength through changing nature and the concentration of electrolytes or adsorp-
tion of ions.
The size distribution of the particles depends on the duration of the nucleation
period; shorter nucleation results in more monodispersed particles, while if the nu-
cleation period is long a wide particle size distribution is obtained.
Although nucleation is decisive, growth, limited by surface reactions, is also
important because the crystal size depends on the ratio of rates rnucleatjon and
rgrowth. In most crystal growth processes, not only diffusion but also surface re-
actions are important. As a result, the crystal growth kinetics often depends not
only on the crystal size but also on the surface structure. The growth rate can
be represented with the first-order kinetics in the solute concentration, while
the nucleation rate is defined as proportional to the supersaturation concentra-
tion and the number of critical nuclei, corresponding to supersaturation.
Rapid precipitation can be initiated by adding seeds to the solution increasing
the concentration of the solution at the interface. High stirring favors formation of
small crystals, because the concentration at the interface is increasing, while higher
temperature increases the crystal size.
Crystallization is complicated by the so-called Ostwald ripening. Solubility of
small crystals is higher than of the larger particles. Concentration in the solution in
equilibrium with a solid particle depends on the particle size. The larger the particle
the lower the equilibrium concentration, thus particles larger than equilibrium size
will grow, while for particles smaller than the equilibrium size at a certain concen-
tration the solution is not saturated and the particles dissolve. Thus, during precipi-
tation in addition to nucleation and crystal growth larger crystals are also formed at
the expense of the smaller ones.
142 2 Engineering catalysts

Nucleation and growth are not separated in time as once few viable nuclei are
formed they start to grow. The Johnson–Mehl equation describes the overall trans-
formation rate:
. :
FðtÞ = 1 − exp − ðπ=3Þ N G3 t4 (2:35)

Relating the overall fraction of the transformed material as a function of time [F(t)]
with the nucleation (Ṅ) and the growth (Ġ) rate considered to be time independent.
The third power in the linear growth rate follows because of three-dimensional (3D,
spherical) growth resulting in the transformed volume increased in proportion to Ġ 3.
Equation (2.35) corresponds to sigmoidal transformations, when F(t)] first in-
creases exponentially in time slowing down thereafter and finally asymptotically
approaching complete transformation. A simplified version of the Johnson–Mehl
equation, known as the Avrami or Avrami–Erofeev equation, is very often used in
practice:
n
FðtÞ = 1 − e − kt (2:36)

where n is known as the Avrami exponent and k is a parameter comprising both the
nucleation and growth rates associated with the phase transformation.
From the discussion presented above it is possible to identify some key engi-
neering issues related to precipitation. As concentration is changing during the pro-
cess quality of the product can vary even for precipitation of a single component.
This can be overcome by precipitation in a buffer solution of an electrolyte with pH
corresponding to the pH of precipitation. An alternative method is continuous pre-
cipitation when pH and concentration is constant when the precursors are fed into
the reactor and the precipitated suspension is removed continuously. Application
of the proper solvent and pH also influences the crystal shape. Habitat (or crystal
shape) can be important for either technological reasons (improving rheological
properties, downstream operations, handling, etc.) or from the viewpoint of cata-
lytic behavior. The rate of cooling or evaporation, the degree of supersaturation,
crystallization temperature and presence of impurities can influence the crystal
shape. In particular, adsorption of impurities blocking the crystal surface may
change the crystal shape or decrease the growth rate. Therefore, impurities can be
removed prior to crystallization or alternatively can be added influencing the crys-
tal habit. Moreover, the precipitate can contain some impurities or ions that might
influence catalytic performance. A remedy is to apply decomposable ions, such as
nitrates, oxalates or ammonium.
Supersaturation degree can be different in different parts of the vessel. In order
to avoid local supersaturation at the points when the solutions are mixed and
achieve homogeneous precipitation, it is possible to add another component (for
example, urea), which keeps the solution homogeneous at least for some time.
 2.3 Preparation of catalytic materials 143

For multicomponent precipitates it is even more difficult to achieve homogeneity

of the precipitate. Because of different solubilities, solid-phase composition could be
different at the beginning and at the end of precipitation. For example, precipitation
of hydroxides from a mixture does not occur simultaneously, but depends on the pH.
Tab. 2.4 contains values of pH at which hydroxides start to precipitate for 0.01 M sol-
utions and corresponding values of equilibrium constants KL = ½Men + ½OH − n are
 
also given for the equilibrium MeðOHÞn $ Men + + ½OH − n . For precipitation to
occur the term [Men+][OH−]n should exceed KL. Thus, during neutralization of acidic
solutions, when mixing is sufficient, formation will be observed in the solid phase of
a hydroxide with a lower value of precipitation pH. Co-precipitation of hydroxides
with similar precipitation pH results in the generation of complex crystal structures,
inclusion of one hydroxide into another one or adsorption of one hydroxide onto an-
other one, which is formed first, etc.

Tab. 2.4: Dependence of precipitation on pH for

hydroxides for 0.01 M solutions.

Hydroxide pH KL
−
Mg(OH) . (÷.) × 
Mn(OH) . . × −
Ni(OH) . −÷−
Zn(OH) . −
Cu(OH) . . × −
Cr(OH) . . × −
Al(OH) . . × −
In(OH) . . × −
Fe(OH) . . × −

(From [50].)

For low soluble compounds the mechanism of crystal formation is different from
the classical method and follows orientation attachment when larger crystals are
formed by crystallographically oriented assembly of smaller nanocrystals.
Aging follows nucleation and growth stages and allows the system to approach
thermodynamic stability. During aging there could be changes in the particle size
caused by dissolution-crystallization and aggregation, changes in the crystal phase
(for example, titania from anatase to rutile), crystallization of amorphous phases
and changes in morphology.
Washing, drying and calcination are typical preparation steps that are per-
formed after aging. A flow scheme of industrial-scale catalyst preparation by precip-
itation is given in Fig. 2.94.
144 2 Engineering catalysts

Mixed metal Sodium

nitrates tank carbonate tank

Filter
Mixed oxides
Drier and
and carbonates
calciner

Mill

Precipitation
and aging Powder mixing
bin
Weighing
machines Catalyst pellets

Pelleting machine

Fig. 2.94: Preparation of catalysts by precipitation. (From [51].)

This scheme is rather general including the most important unit operations
needed to prepare bulk oxides type of catalysts by precipitation. It is worth consid-
ering a specific case of copper- zinc-aluminum oxides catalyst used for methanol
synthesis.
Synthesis of this catalyst includes several steps, namely the preparation of sol-
utions, mixing, filtration, washing, drying, grinding, pelletizing, calcinations and
final packaging. Precursors of high purity are needed to make a catalyst with mini-
mal amount of impurities, giving highly dispersed or amorphous precipitates and
subsequently catalysts with developed surface area. The scheme of the catalyst
preparation is presented in Fig. 2.95.
Water solutions of the corresponding salts of copper and aluminum together
with sodium carbonate are fed into the reactor in (pos. 5), where 20–40 nm crys-
tals of copper carbonate are formed mixed with aluminum hydroxide particles of
a size < 15 nm. After the addition of zinc nitrate, small crystals of zinc carbonate
are generated that are included in the copper carbonate crystals, preventing the
growth of the latter. After filtration (pos. 6) from the mother liqueur, drying and
calcination (pos. 7), the crystals are ground (pos. 8) resulting in small particles.
Graphite (pos. 9) is then added to these particles, which after additional grinding
(pos. 11) are tabletized (pos. 12). The amount of graphite in the final catalyst is
below 2 wt%. Reduction (pos. 15) results in the formation of copper crystallites in-
corporated in the matrix of Al2O3-ZnAl2O4 spinel. The presence of zinc diminishes
dehydrogenation ability of the catalyst.
 2.3 Preparation of catalytic materials 145

Cu(NO3)2 Na2(CO3)2 Al(NO3)3 Zn(NO3)2

H2O H2O H2O H2O

1 2 3 4 Gas
Air
5

7
6

14 8
10
15
9

11

16

13 12

Fig. 2.95: Flow scheme of Cu-Zn-Al oxide catalyst preparation: 1, 4 vessels for preparation of
water solutions; 5, reactor; 6, filter; 7, drying and calcination unit; 8, grinder; 9, vessel with
graphite; 10, mixing and conveyer; 11, additional grinding unit; 12, tabletizing machine; 13 and
14, conveyers; 15, reactor for reduction (or oxidation); 16, vessels for the final catalyst. (From
[52]).

2.3.4 Heteropoly acids

Heteropoly acids (HPA) or polyoxometalates are an interesting class of bulk acid

catalysts with the anion [XxMmOy](–q) balanced by protons. In heteropoly acids X
stands for a hetero atom, such as B, Al, Si, Ge and P, As or Fe, Mn, Co, Cu and
Zn. M indicates addenda atoms (Mo, W, V, Ta, Nb, Os), which are attached to hetero
atoms through oxygen atoms and q is the charge, varying from −3 to −28. Addenda
atoms typically have high charge (+5 or +6), small size with ionic radius in the
range 0.53Å–0.70Å, an expandable coordination number from 4 to 6 and the ability
to form double bonds with unshared oxygen atoms.
Well-known Keggin structures HnXM12O40 (H4Xn+M12O40, with X = Si, Ge; M = Mo,
W and H3X n+M12O40, with X = P, As; M = Mo, W, for example, H3PMo12O40 and H3
PW12O40) can possess such qualities as good thermal stability, high acidity and high
oxidizing ability and could be used as heterogeneous acid catalysis.
146 2 Engineering catalysts

Synthesis of H3PW12O40 can be done by dissolving Na2WO4 and Na2HPO4 in boil-

ing water and adding hydrochloric acid, leading to the precipitation of H3PW12O40·
nH2O, which is extracted with ether.
The thermal stability of HPA is, however, not sufficient for high temperature
applications, moreover the surface area of bulk heteropoly acids is rather low
(2–10 m2 g−1) and therefore activity is insufficient. Supported heteropoly acids can
be used to overcome these limitations. The key issues in the use of such supported
HPA are the retention of polyanion structure, proper dispersion and fixation to the
support. Typically, the choice of the support (γ-Al2O3, MgO, ZrO2, TiO2, activated
carbon, zeolites, mesoporous materials, clays or mixed oxides) is determined by a
need to have a high surface area, inertness and thermal stability.

2.3.5 Catalyst supports

Supported catalysts are often applied because they combine a relatively high dis-
persion (amount of active surface) with a high degree of thermostability of the cata-
lytic component. The important properties of the support are stability at reaction
and regeneration conditions, proper texture, thermal conductivity, mechanical
strength and low costs. Typically, different phases of alumina, silica, titania, zirco-
nia, magnesia, zinc oxide and active carbon are used.
The support should also enable production of shaped particles (mm range)
where small easily sintered crystals of the active phase (nm range) do not coalesce.
Pre-shaping of supports is an attractive option. However, care must be taken that
dispersion of the catalytic components is not modified in the following steps. With
powdered supports, the intimate mixing during deposition of the catalytic compo-
nents is easily realized during the first step, however, the following operations, in
which the grains are transformed into their required shape with desired porosity,
are more difficult and the dispersion may not be uniform.

2.3.5.1 Carbon
Active or activated carbon, together with alumina and silica, is mainly used as a cata-
lyst support. Activated carbon is preferred in applications where properties like inert-
ness, lower coke propensity compared to silica and alumina, stability at a wide range
of pH and temperatures, high adsorption capacity, wide variety of textural properties,
are of importance. Carbon supported noble metal catalysts are typically applied in
various liquid-phase hydrogenation reactions at rather low temperatures.
In oil refining the application of activated carbons is usually limited by chemical
reactivity in the presence of oxygen. This means that conventional methods of catalyst
regeneration by burning off the coke away are not applicable. Moreover, active car-
bons are mechanically weak, which prevents their applications in fixed-bed reactors
 2.3 Preparation of catalytic materials 147

and leads to generation of fines in slurry systems. Conversely, in order to recover the
active phase (expensive precious metals for example) from spent catalysts, carbon
can be easily burned away. It also should be noted that although conventional car-
bons are used mainly as powders, different forms, such as extrudates, cloths and
fibers could be prepared. Finally, the costs of carbon supports are usually lower than
other conventional support materials.
Activated carbon is produced from carbonaceous (for example, carbon rich)
source materials such as nutshells, wood and coal by physical or chemical activa-
tion. In the former case carbonization is used.

Once during a scientific presentation a PhD candidate, who happened to be the author of this
book, got a question related to selection of the support, which was active carbon. The person
who posed the questions was wondering why this particular support was used as it was active in
the reaction in addition to the metal on it. That person was assuming that the term “active” was
related to catalytic activity of the support and not the activation procedure during preparation.

Materials with carbon content are pyrolyzed at temperatures in the range 600–900 °C,
in the absence of air (usually in inert atmospheres with gases like argon or nitrogen)
(Fig. 2.96). This process is followed by activation/oxidation as carbonized material is
exposed to oxidizing atmospheres (carbon dioxide and/or steam at 400–600 °C).
Some carbon is burned away to give the porous structure. Chemical activation means
impregnation with, for instance, phosphoric acid or zinc chloride, and carbonization
at temperatures in the range of 400–600 °C. Carbonization/activation steps proceed
simultaneously. Activating chemicals remain in the structure after carbonization.
After washing, the final activated charcoal is produced. Activated carbons have a
much higher specific surface area than other types of supports (> 1,000 m2 g−1).

Peat, coal, biomass,

Carbonaceous material
sugar, polymers
High T
pyrolysis
Porous material
H2O
Partial gasification
Additional porosity Activated carbon

Fig. 2.96: Preparation of activated carbon.

All activated carbons have a porous structure, usually with a relatively small
amount of chemically bonded heteroatoms (mainly oxygen and hydrogen). In addi-
tion, activated carbon may contain up to 15% of mineral matter (the nature and
amount is a function of the precursor), which is usually given as ash content.
148 2 Engineering catalysts

The adsorptive properties of activated carbon are determined not only by its
porous structure but also by its chemical composition. In graphite, with a highly
oriented structure, the adsorption takes place mainly by the dispersion compo-
nent of the van der Waals forces, but the random ordering of the imperfect aro-
matic sheets in activated carbon result in incompletely saturated valences and
unpaired electrons, and this will influence the adsorption behavior, especially
for polar or polarizable molecules. In addition activated carbon is associated
with such heteroatoms as oxygen and nitrogen (derived either from the starting
material, activation process or post-treatment) and with the inorganic ash com-
ponents. The presence of oxygen and hydrogen in surface groups can be up to
30 mol% H and 15 mol% O, which has a large effect on the adsorptive properties
of the activated carbon. Some functional groups in carbon supports are given in
Fig. 2.97.

Quinone
O Free
Ethers
Phenolic O
radicals
HO
O
C
H OH
C O OH
O Phthalic
O
Aldehyde C
H
O
C
p-sites: O
Lactone
C6 rings, edge OH
>C=C< bonds HKCN
O
C7 and C5
Carbon C OH rings
Pyran O
chains O C O
H
Pyrone R
Carboxylic
Chromene

Fig. 2.97: Functional groups in activated carbon.

Oxidation with hydrogen peroxide and nitric acid results in the formation of C–O
bonds, lactones, quinine, and carboxyl-carbonate structures with higher acidity
generated by treatment with HNO3. Therefore, even if carbon is hydrophobic the
chemical nature could be modified to increase hydrophilicity.
In some cases, the high surface area of the carbon support may be detrimental
if the active catalytic phase is confined in narrow micropores that are not accessible
to the reactant molecules. This is important in processes where large molecules are
involved, for example, in liquid-phase reactions of fatty acids when diffusion of re-
actants and products may be hindered by the narrow porosity.
 2.3 Preparation of catalytic materials 149

Because activated carbons are lacking purity and could be hardly tunable mate-
rials with difficult to reproduce properties (even having the same surface area) sev-
eral attempts were made in academia and industry to synthesize a carbon support
that will be reproducible and mechanically stable. Thus, there is a growing interest
in making catalyst supports from carbon nanotubes and carbon nanofibers, which
could be produced in reproducible ways and make it possible to tailor the proper-
ties of these materials.
Carbon nanotubes (CNTs) that are categorized as single-walled (SWNTs) and
multi-walled nanotubes (MWNTs) possess a cylindrical nanostructure with a
length-to-diameter ratio of millions to one. They can be viewed as rolled, long, hol-
low structures with the walls formed by one-atom-thick sheets of carbon. High ther-
mal conductivity and special mechanical and electrical properties made them a
subject of intense research. The chemical bonding of nanotubes is composed of sp2
bonds, as they are stronger than sp3 bonds and thus result in materials with sub-
stantial mechanical strength. Single-walled nanotubes (SWNT) have a diameter of
close to 1 nm. Multi-walled nanotubes (MWNT) consist of multiple rolled layers of
graphene (Fig. 2.98).

(A) (B)

Fig. 2.98: Structure of (A) single wall- and (B) multiwall nanotubes.

Carbon nanotubes are prepared by catalytic vapor phase deposition of carbon that
uses an initial layer of metal catalyst particles, most commonly nickel, iron or co-
balt. The diameter of the nanotubes, which is controlled by annealing or plasma
etching of a metal layer, is related to the size of the metal particles. Nanotubes grow
at the metal catalyst sites where the carbon-containing gas (acetylene, ethylene or
methane) is dissociated on the surface and formed carbon is transported to the
edges of the catalyst particle, where the nanotubes are formed with the metal par-
ticles staying either as the nanotube base (Fig. 2.99, top) or rest at the tip of the
growing nanotube (Fig. 2.99, bottom) depending on the adhesion.
150 2 Engineering catalysts

Extrusion or root growth

CnHm

Metal
C C

Support

Tip growth

CnHm C +H2

Metal

Support

Fig. 2.99: Mechanisms of nanotubes growth. (Reproduced from [53]).

When CNT are used as catalyst supports they provide a spatial restriction on
the metal particles by adjusting the CNT diameter and sintering.
Removal of the initial catalytic material could be an issue in this method. If this
is done by an acid treatment (to remove alumina) the original structure of the car-
bon nanotubes could be destroyed.
Process-wise CNT can be synthesized in a flow reactor (Fig. 2.100) at ~700 °C to
which a carbon-containing gas is fed along with a process gas (for example, nitro-
gen or hydrogen).
When plasma is generated by applying a strong electric field using plasma en-
hanced chemical vapor deposition, the nanotubes grow in the direction of the elec-
tric field.
Fluidized bed reactors are used industrially for CNT preparation. The fluidized
bed pilot plant of 10 t year–1 developed by Arkema (Fig. 2.101A) for the production
of Graphistrength™ MWNT allowed for the synthesis of materials as highly en-
tangled bundles (Fig. 2.101B) with a size of several hundreds μm.
Similarly a fluidized bed reactor was used for the production of Baytubes®
(Fig. 2.102) resulting in agglomerates of multiwall nanotubes, where each tube (diame-
ter ~15 nm) comprises several graphite layers. The process, with an announced annual
capacity of 3,000 t, is claimed to provide high carbon yields and high space-time.
 2.3 Preparation of catalytic materials 151

Quartz
tube
Gas inlet Gas outlet

Quartz Sample
boat
C2H2N2
Oven 720 °C

Fig. 2.100: Scheme of CNT synthesis in a flow reactor.

100 µm

(A) (B)

Fig. 2.101: Production of CNT at Arkema. (A) Pilot plant photo, (B) SEM image of MWNT bundles.
(From [54]).

Filter/incineration unit

Catalyst

Heating

Fig. 2.102: CNT production in a fluidized bed (redrawn from

Hydrocarbons/H2/inert gas [55]).
152 2 Engineering catalysts

Carbon nanofibers (CNFs) are another material made of graphene layers, which
are high surface area materials (~200 m2 g−1). They are synthesized in a similar way
to CNT using the catalytic growth methodology with iron, cobalt and nickel as the
catalytically active metals. Graphical layers of different orientation can be formed,
for example, the fishbone type (Fig. 2.103).

Fig. 2.103: Electron microscopy of fishbone nanofibers. (Reproduced with permission from [56]).
Copyright (2012) American Chemical Society.

The thickness of the fibers can be controlled by the metal particle size, while the
graphite plane orientation depends on the growth temperature and the metal type
(parallel or fishbone types are formed with iron or nickel, respectively). The growth
rate influences the strength of the fibers, for example, slow growth of thick fibers
leads to strong particles.

2.3.5.2 Deposition of metals on carbon and surface charge of supports

Carbon is essentially hydrophobic in nature. It has low affinity for polar solvents
such as water and high affinity for non-polar solvents (acetone). When a metal, for
example, platinum, is deposited from water solutions of hexachloroplatinic acid, it
will be, therefore, located primarily at the outer surface. Application of acetone as a
solvent allows more uniform distribution along the catalyst grain.
Another important factor is the presence of oxygen surface groups. Because of
these surface groups carbon supports (as well as other supports, such as silica or alu-
mina, which will be discussed in the subsequent sections below) develop a pH-
dependent surface charge (Fig. 2.104A) when placed in an aqueous solution.
Dissociation of the acidic groups on the surface of a particle will lead to a negatively
 2.3 Preparation of catalytic materials 153

– +
– – + +
– ++ + – –
+ + – –
– +
+ + – + –– – +
+ – – +
– + + + – –
– +
– +
+ H+A– + B+OH–
(A) Acid pH Basic pH

40
20
0
ζ (mv)

–20
–40 IEP
–60
–80
–100
0 2 4 6 8 10 12
(B) pH

Fig. 2.104: (A) Surface polarization as a function of the solution pH. (B) Dependence of zeta
potential on solution pH for Norit carbon.

charged surface. A decrease of pH in the case of a negatively charged surface will

reduce the surface charge to zero (isoelectric point or point of zero charge). The basic
groups will attract a positive charge from the solution and an increase of pH will
counterbalance it. As the surface of the support is charged depending on the pH it
can accept either cations or anions. Thus, a special care should be taken during cata-
lyst preparation regarding the charge of the surface and the charge of species to be
deposited.
Figure 2.104(B) illustrates how the zeta potential of activated carbon Norit de-
pends on the solution pH.
Because of the surface charge there are more ions of the opposite charge in the
interfacial region around the particle. The electric double layer consists of two
zones: an inner region or Stern layer where the ions are strongly bound and the
outer layer (diffuse) with less firmly bound ions. A boundary in the diffuse layer
separates ions (which move when a particle moves, for example, because of grav-
ity), from those ions that stay in the bulk. Zeta potential reflects the potential at this
boundary. High values of zeta potential typically mean that the particles will repel
each other. As clearly seen from Fig. 2.104B zeta potential displays strong depen-
dence on pH in aqueous solutions. Figure 2.104B also illustrates that the position of
154 2 Engineering catalysts

the isoelectric point (IEP) is close to pH = 4.5. Below this pH the surface is positively
charged and will accept anions.
The thickness of the double layer depends on the concentration of ions in so-
lution. Some inorganic ions (for instance contaminants in the solution) strongly
adsorb on the surface and can thus strongly influence the zeta potential even if
they are present in low concentrations.
Although the oxygen group’s presence is beneficial for better metal dispersion,
it worsens reducibility of metals in a carbon supported catalyst, where carbon-
active phase interactions are typically weak.

2.3.5.3 Transition metal oxides

Transition metal oxides (e.g., titania, chromia, zirconia and vanadia) exhibiting var-
iable oxidation states, many crystallographic forms and thermal stability are widely
used as catalysts and supports in different catalytic reactions, including selective
oxidation, hydrogenation, selective dehydrogenation, alkylation, aldol condensa-
tion, carbonylation, amination and ammoxidation reactions to name a few. Surface
acidity and basicity as well as cationic/anionic vacancies make transition metal
oxides attractive for utilization in catalysis. A particular interesting application is
related to photocatalysts when titania, which is a wide band gap semiconductor, is
often implied for various reactions including solar water splitting.
Titanium dioxide TiO2, or titania, as a naturally occurring oxide of titanium, is
used in the form of nanoparticles, as a pigment in paints and in sun protection
creams. It is typically prepared by aqueous hydrolysis of titanium salts or flame
hydrolysis of TiCl4. Two crystalline forms of titania – anatase or rutile – compris-
ing chains of TiO6 octahedra are obtained depending on the preparation condi-
tions. Both rutile and anatase have tetragonal crystalline structure containing,
respectively, 6 and 12 atoms per unit cell. Anatase is formed as a metastable
phase with high surface area and can be then slowly transformed into a thermody-
namically more stable rutile.
Flame hydrolysis of titanium chloride leads to anatase with a surface area
40–80 m2 g–1 and mean pore diameter close to 50 nm. This phase is more stable at
high T and thus more suitable for catalytic applications. Aqueous hydrolysis of tita-
nium salts gives a support with the surface area of 190–200 m2 g–1 and a pore diam-
eter of 15–20 nm. Contrary to silica and alumina, which are insulators, titania is a
semiconductor and thus can be used as a photocatalyst. Moreover because of the
possibility of ready reduction of Ti+4 to Ti3+ the partially reduced titanium oxide
species can decorate supported metal nanoparticles, significantly changing the cat-
alytic properties because of strong metal support interactions.
There is a long history of using an aerosol flame method for the production of
titania (spherical particles of predominantly anatase) for various photocatalytic
reactions (such as titania marketed by Degussa as P25). It is not only titania that
 2.3 Preparation of catalytic materials 155

can be produced by this method but also other supports (alumina or silica) with
non-porous structure and high external surface areas, which includes exposure to
high temperatures and cooling rates. The resulting materials have good thermal
stability.
Aerosol flame synthesis can be classified into vapor-fed and liquid-fed
variants. In the vapor-fed methods applied for the synthesis of fumed silica and tita-
nia, volatile precursors (such as chlorides) are evaporated and fed into the flame.
For synthesis of titania, the flame is needed to ignite the process, while a hydrogen/
oxygen flame can also assist the process as in fumed silica manufacturing. The
metal precursor is first converted into the metal oxide, which undergoes nucleation
from the gas-phase. The schematic diagram is presented in Fig. 2.105A. A drawback
of the method is associated with limited availability of a volatile precursor at a com-
petitive price. In flame or flame-assisted spray pyrolysis liquid precursors are intro-
duced as either non-combustible (FASP, Fig. 2.105B) or combustible (FSP,
Fig. 2.105C). The main advantage of this method is the formation of nanosized par-
ticles with possible size control through the precursor-solvent composition.

A B C

Vapor-fed Flame- Flame

aerosol assisted spray
flame spray pyrolysis
synthesis pyrolysis

Nozzle
HC or H2/O2 HC or H2/O2
or air or air Dispersion gas
Aqueous O2 or air
Precursor Evaporator precursor Atomizer
solution Organic
precursor
solution

Fig. 2.105: Various flame configurations. (Redrawn from [57]).

The schematic picture of the particles formation is illustrated in Fig. 2.106.

For other common catalyst supports, such as silica or alumina, the flame meth-
ods result in highly aggregated structures. In general, the final particle size of the
solid material depends on the droplet size, gas velocity of the spray or the residence
time in the reactor and quenching efficiency. The flame process is influenced by a
number of parameters, such as gas and liquid-feed rates, type and concentration of
the precursor and solvent, solvent-to-fuel ratio. For instance, nucleation and coagu-
lation rates are increased at high liquid-feed rates, which results in enhanced
156 2 Engineering catalysts

Hollow Large Small

particles particles particles

Nucleation
Intraparticle Intraparticle Precursor Product
conversion conversion vapor vapor
and collapse

conversion
Precursor
evaporation
Solvent and
precursor
Solvent evaporation and
precursor precipitation
Precursor
Droplet Droplet vapor

(Flame-assisted) Flame spray Vapour-fed aerosol

spray pyrolysis pyrolysis flame synthesis

Fig. 2.106: Schematics of the particle formation. (Redrawn from [57]).

particle growth. Conversely, a reduction of the overall particle growth is achieved at

lower flame temperatures and thus allows for faster cooling. This in turn is possible
at high dispersion gas-flow rates. However, because of the complexity of flame
methods, precise control over the size is not easy and the particles exhibit generally
a wide particle size distribution.
Moreover, complete crystallization is difficult to achieve in conditions of very
short residence times at high temperatures and fast cooling rates. This can result in
partially amorphous materials restricting the use of flame methods for synthesis of
catalyst supports when high crystallinity is needed.
As flame pyrolysis methods give materials with high thermal stability, in addi-
tion to titania such supports such as ceria are also produced by these methods.
Such thermo-stability is important for automotive exhaust applications.
Ceria along with zirconia are bifunctional catalysts and contain both acid and
base sites, which obviously make them attractive in bifunctional heterogeneous
catalytic reactions as well as in catalytic conversion of CO2 to value-added chem-
icals. In particular, appealing are the high oxygen mobility and redox property of
CeO2.
Zirconia is formed mainly in tetragonal and monoclinic crystallographic
phases. The former phase is metastable, possessing a high surface area, while the
monoclinic form is thermodynamically stable.
 2.3 Preparation of catalytic materials 157

2.3.5.4 Silica
The amorphous silica supports are typically used in applications that require rela-
tively low temperatures, as it has lower thermal stability when compared to alumina.
Moreover, under exposure to steam volatile hydroxides can be formed at elevated
temperatures. By its chemical nature, silica is a weak acid and is more resistant than
alumina to acidic media.
Two different methods are used for preparation of silica supports: flame hydro-
lysis giving fumed silica and sol-gel precipitation leading to silica gels.
Flamed or fumed silica is a very pure material with nanometer sized dense par-
ticles (40–50 nm) not possessing micropores. The surface area is up to 300 m2 g−1
and the diameter of pores is above 7 nm. This type of silica is manufactured by
flame hydrolysis of SiCl4, which is first formed by carbo chlorination of a macro-
scopic oxide:

SiO2 + C + 2Cl2 ! SiCl4 + CO2

Silicon tetrachloride then reacts in a flame reactor where (in addition to

SiCl4) hydrogen and oxygen (air) are fed, leading to the following reactions:
2H2 O + O2 ! 2H2 O, SiCl4 + 2H2 O ! SiO2 + 4HCl. Nanometer sized silica particles are
separated from HCl in a cyclone followed by deacidification of the silica surface with
steam and air in a fluidized bed downstream of the cyclone (Fig. 2.107). Formed silica
has a low density and needs to be compacted prior to forming operations in a com-
pactor, located downstream of a fluidized bed reactor and a hopper. Process parame-
ters such as flame temperature, content of the substrate, hydrogen to oxygen ratio
and the residence time allows control of the size and aggregation of silica.

Hydrogen To HCl absorption

Oxygen (air) Cyclone

Mixer

Burner
Vaporizer

Silicon tetrachloride Pyrogenic

Deacidi- silica
Cooler
fication

Fig. 2.107: A flow scheme of fumed silica production. (From [51].)

In the sol-gel route the alkaline solution of sodium or potassium silicate (for exam-
ple, the pH of water glass for sodium silicate is 12) is mixed with sulfuric acid,
158 2 Engineering catalysts

which triggers the formation of first Si(OH)4 monomer units, followed by polymeri-
zation to a colloidal solution (sol) where silanol (Si–OH) groups and Si–O–Si bonds
are formed. Minimization of –OH and maximization of Si–O–Si bonds is the driving
thermodynamic force for the formation of oligomers and cyclic species, small par-
ticles and eventually larger particles. Colloidal solutions are formed by micelles,
which are separated because of repelling electric charges on their surface and in
the solution, prohibiting coagulation. Sodium ions play a critical role in controlling
agglomeration of micelles. At some point the sol undergoes gelation – the formation
of a 3D hydrogel (Fig. 2.108). Gelation depends on the micelle concentration, tem-
perature, ionic strength and especially pH.

OH HO OH HO OH HO
OH
OH Si HO
HO Si Si OH
HO Si OH Si
O O O O O O Si
Sol O
O O
OH O

Gel

Fig. 2.108: Formation of a gel.

The key factors are the rates of hydrolysis –M–O–Na(R) + H2O → –MOH + Na(R) OH
condensation –MOH + XOM → M–OM + XOH, where X = H, R. R indicates that pre-
cursors in sol-gel synthesis could be not only salts such as metal silicates, but also
alkoxides. If hydrolysis is faster than condensation then the material is less
branched, while slower hydrolysis leads to a highly branched one.
The gelation rate is slow at low and high pH and can range from minutes to
days. The density of hydrogel, which is a metastable polymer, increases with an in-
crease of the initial salt concentration and gelation rate. The subsequent steps in
the preparation of silica are washing of the gel to remove sodium and then drying.
The drying is critical in producing a material that could be used as a catalyst sup-
port. Dried material can occupy 5% of the original hydrated gel volume. Drying at
moderate temperatures (150–200 °C) leads to the collapse of pores, drastic reduc-
tion of porosity and formation of xerogels, which lack macropores and the sufficient
strength for them to be used in fixed-bed reactors. These xerogels are first milled to
the desired size and mixed with binders prior to shaping. An alternative to this pro-
cedure is drying under supercritical conditions leading to aerogels. First a less
polar then water solvent (i.e., acetone) is used to wash away water and then ace-
tone itself is washed away with high-pressure liquid carbon dioxide. After heating
 2.3 Preparation of catalytic materials 159

Silicate Acid

Mixing

Washing
Drying

Classification Milling

Packaging

Fig. 2.109: Preparation of silica by sol-gel method. (From [51].)

beyond the critical point and gradually releasing pressure a dried product is
formed. The flow scheme of silica preparation is illustrated in Fig. 2.109.
After drying at a low temperature the concentration of silanol groups is high
(4–5 OH per nm2) and the material is hydrophilic. Heating at a high temperature
results in dehydroxylation, and thus a decrease in concentration leading to hydro-
phobic surfaces. The surface area of commercially available silica supports varies
between 400–700 m2 g−1 with an average pore diameter of 2.5–5 nm depending on
the preparation conditions.
The surface charge of silica depends on pH (Fig. 2.110). At pH equal to the point
zero charge (pH ~2) the surface is neutral with –SiOH groups, while at acidic and
basic pH the main surface groups are –Si–OH2+ and –SiO–, respectively.
In addition to colloidal and fumed silica, another type of silica – kieselguhr –
found an application as a support for vanadium pentoxide catalysts, which are ap-
plied in oxidation of SO2 to SO3. Diatomaceous earth (diatomite) kieselguhr is a
form of silica composed of the siliceous shells of unicellular aquatic plants of micro-
scopic size. This material is heat resistant and has been used as an insulator, as a
component in toothpaste and as an abrasive in metal polishes. Kieselguhr, with a
surface area of 20–40 m2 g−1 and a broad range of pore sizes, contains small
amounts of alumina and iron as part of the skeletal structure.
160 2 Engineering catalysts

10
0
–10
–20
ζ (mv)

IEP
–30
–40
–50
–60
–70
0 2 4 6 8 10 12
pH

Fig. 2.110: Zeta potential as function of pH for silica.

2.3.5.5 Alumina
Although a variety of alumina exists, distinguishable by XRD, only a few phases
are applied as catalyst supports, such as non-porous low surface area crystalline
α-alumina, and porous amorphous η and γ-alumina. Annual world production of
alumina is approximately 45 million tons, over 90% of which is used in the manu-
facture of aluminum metal.
Aluminum is extracted from bauxite, an ore containing aluminum hydroxide, sil-
ica and other oxides by treating it with sodium hydroxide. Formed sodium aluminate
Na2O·Al2O3 (or Na2Al2O4) undergoes crystallization that leads to different gels de-
pending on the pH. This is done by adding nitric acid to sodium aluminate. When
precipitation occurs at 8 < pH < 11 bayerite (Al(OH)3) is formed, which upon aging at
high pH gives gibbsite gel with the same chemical formula of aluminum hydroxide.
Precipitation at a lower pH (6 < pH < 8) results in crystalline boehmite gel AlOOH.
Different calcination procedures result in various phases of alumina (Fig. 2.111).
In cases when the presence of alkali might influence catalytic performance,
an alternative to the above mentioned route should be used. Alkali-free alu-
mina could be prepared from aluminum sulfates (Fig. 2.111) or aluminum alco-
holates. Both alkali and alkali-free methods result in high amounts of acids
and bases needed (2–4 t per ton of alumina), which are cumbersome to regen-
erate. γ-Alumina has a surface area in the range 50–300 m2 g−1 and pores between
5–15 nm. Rather high thermal and mechanical stability also makes alumina a popu-
lar support for various applications. The surface of alumina contains Brønsted acid
sites (H+ donors), Brønsted basic sites (acceptors of protons) and Lewis acid sites
(electron acceptors). After dehydroxylation the surface of alumina contains mainly
Lewis acidity,

O– OH O–
+ H+
O Al O Al O Al O Al
Lewis site Basic site Brønsted acid sites
 2.3 Preparation of catalytic materials 161

Al-sulfate solution Na-aluminate solution

Al2O3.3SO3 Na2O.Al2O3
pH < 2 pH > 12
+ +
Base Acid

At 3 < pH < 6 At 6 < pH < 8 At 8 < pH < 11

precipitation of micro precipitation of precipitation of
crystalline boehmite gel crystalline boehmite gel bayerite gel Gibbsite gel
AlO(OH) AlO(OH) Al(OH)3 Al(OH)3

Heating above 770 K Aging at pH 8 and 353 K Aging at high pH

Amorphous Al2O3 γ-Al2O3 η-Al2O3 𝜒-Al2O3

Spinel structure: cubic close packing
Heating at about 1,170–1,270 K

θ-Al2O3 δ-Al2O3 α-Al2O3

Heating above 1,370 K

α-Al2O3
Corund structure: hexagonal close packing

Fig. 2.111: Preparation of different alumina phases. (From [51].)

– – –
100 ++++ + + +
+
Al2O3
80
Zeta potential (mV)

60
40
20
0
0
2 4 6 8 10 12 14
–20
+ + +
–40 Isoelectric point
–––––––
(zero surface charge
–60 at pH = 7.5–9) Al2O3

Fig. 2.112: Zeta potential dependence on pH.

In water solutions, the surface charge depends on pH, with the isoelectric point
close to 7 (Fig. 2.112).
Heat treatment of γ-alumina at temperatures above 1,370 K results in the forma-
tion of first θ alumina and then α-alumina (corundum). Simultaneously, the pri-
mary particles of γ-alumina grow in size to 70 nm for α-alumina concomitant with
162 2 Engineering catalysts

the reduction of the surface area (few m2 g−1). The latter phase of alumina found its
niche in high temperature applications, such as steam reforming of natural gas,
which will be discussed in detail in Chapter 4. The use of additives such as oxides
of chromium, iron, molybdenum or HF allows for lowering the temperature of
phase transformations and even avoiding the formation of the θ alumina phase. In
practice ammonium fluoride is used instead of hydrofluoric acid, as it decomposes
to NH3 and HF. HF adsorbs on the surface of γ -alumina at ~700 °C followed by
polymorphic transformations of γ -alumina to α-alumina at 700–1,000 °C. Although
the surface area of the product is ~10 m2 g−1, the total porosity (~0.55 cm3 g−1) could
be preserved in this preparation procedure.

2.3.5.6 Zeolites
A special type of supports, which are widely used as catalysts, are solid acids called
zeolites. Zeolites are hydrated aluminosilicate minerals with a microporous struc-
ture. The term was originally coined in the 18th century by a Swedish mineralogist
A. F. Cronstedt who observed, upon rapidly heating a natural mineral that the
stones began to dance about as the water evaporated. Using the Greek words which
mean “stone that boils”, he called this material zeolite. Zeolites are crystalline alu-
minosilicates containing pores and cavities of molecular dimensions. There are
close to 50 natural zeolites and almost triple this amount of synthetic zeolites that
do not occur in nature. The synthetic zeolites are among the most widely used sorb-
ents, catalysts and ion exchange materials in the world. Zeolite crystals are porous
on a molecular scale, their structures revealing regular arrays of channels and cavi-
ties (~3–15 Å), creating a nanoscale labyrinth. The complicated structure of zeolites
is constructed from several secondary building blocks with different number of
units, possessing 6, 8, 10 or 12-membered rings of T-atoms (Si, Al). Properties of
some zeolites are presented in Tab. 2.5 and Fig. 2.113.
Channels contain water molecules and alkali metal cations (M I, MII), which
can be ion-exchanged. The general formula of zeolites is MI MII0.5[(AlO2)x*(SiO2)y*
(H2O)z]. In addition to Al, silicon in the T atom position could be replaced, for exam-
ple, by P, Ti, Ga, Ge, Hf or Zr. Ions should have a coordination number of four with
the respect to oxygen and an ionic radius that fits into the zeolitic framework. An
interesting example of molecular sieves is VPI-5, which is an aluminophosphate
with a pore aperture of 1.2 nm. Such materials are electrically neutral and thus
largely catalytically inactive.
The key properties of zeolites are size and shape selectivity, together with the
potential for strong acidity. In zeolites SiO4 and AlO4 tetrahedra linked through com-
mon oxygen atoms and cations (such as Na+) are needed to neutralize the charge dif-
ference between Al3+ and Si4+. Brønsted acidity appears when Si4+ is replaced by Al3+
and the charge is balanced by proton, localized in vicinity of Al. Zeolites could thus
have acidities compared to mineral acids. Acidity depends on the Si/Al ratio. With an
 2.3 Preparation of catalytic materials 163

Tab. 2.5: Properties of zeolites (From [29]).

Common Three-letter Channels Window or Pore Si/Al

name code channel volume
diameter (nm) (cm g−)

Zeolite A LTA D, Cages connected by windows, . . 

cubic
Zeolite X FAU D, Cages connected by windows, . . –.
tetragonal
Zeolite Y FAU D, Cages connected by windows, . . >.
tetragonal
Mordenite MOR D, two straight, parallel .; . . >
channels
Zeolite L LTL D, one straight channel
ZSM- MFI D, one straight, one sinusoidal, . . >
mutually perpendicular channels
Zeolite BEA D, two straight, one sinusoidal, .; . . >
Beta perpendicular channels

Fig. 2.113: Various types of zeolites with different numbers of T-atoms (copyright [58]).

increase of Si content, Lewis acidity decreases and is zero for silica. Brønsted acidity
(zero for alumina) starts to be seen at Si/Al ~0.3, has a maximum at Si/Al ~3, and is
zero for silica. An empirical Lowensteins’ rule determines the structure of linkages in
zeolites, stating that Al-O-Al linkages are forbidden and thus Si/Al ratio must be
above or equal to unity. In fact, Si/Al ratio varies from 1 (LTA) to infinity (pure SiO2).
164 2 Engineering catalysts

Zeolites with a low aluminum content are hydrophobic (and vice versa). Zeolites are
thermally stable and if deactivated could be regenerated by simply burning the car-
bon deposits. Stability of zeolites increases with Si/Al ratio.
Usually the as-synthesized forms of zeolites (typically with Na+ balancing the
charge) are not the most suitable for the catalytic purposes. Acidity can be intro-
duced in the as-synthesized zeolite by ion-exchanging sodium cations with NH4Cl
or NH4NO3 to achieve an ammonium form. The ammonium form of zeolite is dried
and calcined to obtain the proton form of zeolite. The exchange of extra-framework
species (for example, Na+ to NH4 and H+) is one of the ways that zeolites structure
modifications. Other methods include dealumination by steaming or acid treatment
in order to change Si/Al ratio. Removal of aluminum from the framework by, for
example, hydrothermal treatment with steam at 600–900 °C leads to an ultrastable
zeolite Y, industrially used as a catalyst for catalytic cracking.
Incorporation of transition metals into zeolites gives rise to bifunctional cataly-
sis, where both metal and acid centers are involved in a catalytic reaction.
Incorporation of Pt into zeolites tested in a ring opening of decalin (Fig. 2.114A), re-
sulted in a decrease of the amount of strong acid sites while the overall number of
acid sites was almost unchanged (Fig. 2.114B). As a consequence, the amount of
cracking products decreased with a simultaneous increase in selectivity towards de-
sired ring-opening products.

cis- and trans-decalin Cracking products

CH3 200
CH3
Concentration of BAS (μmol/g)

H3C 180
CH3 160
140
120
100
CH3 CH3
80
H3C H3C
CH3 60
CH3
40
CH3 CH3
CH3 20
H3C
0
Weak Medium Strong
Decalin isomers Ring-opening products Strength of Brønsted acid sites (BAS)
(A) (B)

Fig. 2.114: (A) Ring opening of decalin and (B) concentration of Brønsted acid sites after
incorporation of Pt into zeolites. (Data from [59]).

In general, non-uniform distribution of aluminum leads to materials with different

acid strength. Acidity of zeolites is probed by mainly non-aqueous spectroscopic
methods (IR spectroscopy of probe molecules) and temperature-programmed de-
sorption of bases, giving information on the total amount of acid sites and the acid
sites distribution.
 2.3 Preparation of catalytic materials 165

For water non-sensitive materials, for example, acidic amorphous oxides or

active carbons, acidity could be determined by titration of aqueous slurries of the
acidic solids with a standard base.
Although catalytic activity of zeolites is often related to Brønsted acidity (for ex-
ample, the presence of protons) they also contain Lewis sites (coordinatively unsat-
urated Al+3 ions).
Zeolites are microporous materials with pore diameters below 2 nm, which re-
sults in some interesting selectivity properties. Thus an important property of zeolites
is shape selectivity, which refers to reactants (Fig. 2.115A), product (Fig. 2.115A) or
transition state selectivity (Fig. 2.116). Fig. 2.115A illustrates that in the case of a zeo-
lite with narrow pores, reactants with bulky substituents cannot enter the pores. For
product selectivity (Fig. 2.115B) only p-xylene can diffuse out of the ZSM-5 channel
pores, while o-xylene and m-xylene cannot be formed. Some transition state inter-
mediates are too large to be accommodated within the pores/cavities of the zeolites,
even though diffusion of neither the reactants nor the products is restricted. This is
the case for transition state selectivity (Fig. 2.116). The diameter of the cage is impor-
tant for the reactant and transition state selectivity, while diameter and the pore
length are the critical parameters for the product selectivity. In some instances coking
to reduce the pore size can be a method to improve shape selectivity.

Ca/X
OH

+ H2O
OH

Active site
OH
+ H2O
OH

(A) Ca/A

CH3OH +

(B)

Fig. 2.115: (A) Reactant and (B) product selectivity in zeolites.

Aluminosilicate zeolites are formed by hydrothermal synthesis, typically under

mild conditions (350–525 K) and autogenous pressure (above 373 K). Zeolites are
synthesized by mixing a silica source (the primary building units of the framework,
166 2 Engineering catalysts

Fig. 2.116: Transition state selectivity.

i.e., SiO2), an alumina source (origin of the framework charge), organic templates,
mineralizing agents (OH– anion, NaOH) and water (as solvent) to form an alumino-
silicate gel. The process in illustrated in Fig. 2.117.

SiO2 source
Liquid-phase-
Colloidal SiO2 mediated Crystals
Waterglass (Na silicate) transformation
Pyrogenic SiO2
Si alkoxide (SiOEt)4
Nuclei

AIO2 source

AI + NaOH Hydrogel/
NaAIO2 (pseudoboehmite) Precursors
precipitate
gibbsite
AI alkoxides

SDA: structure-directing agent

Mineralizers (catalysts)
Cationic organics (tetraalkylamines)
Neutral organics (solvents, amines) Salts (alkali, alkaline earth)
OH-
F-

Fig. 2.117: Schematic representation of zeolitization. (Reproduced with permission from [60]).

After initial transformation of Si, Al sources and the structure directing agent (SDA)
through the liquid phase mediation into a hydrogel or precipitate, the latter is con-
verted into precursor species. Such species assemble into nuclei, which then grow in
crystal according to classical crystallization. SDAs act as topology-determining
 2.3 Preparation of catalytic materials 167

templates while transformations of the gel to precursor/nuclei are thought to be cata-

lyzed by mineralizing agents such as alkali (alkaline earth) salts and OH−. Figure 2.118
illustrates the crystal growth of zeolites exhibiting typical S-shaped crystallization
curves [61] with an induction period.

(a) Nucleation rate (b) Crystal growth rate

Gel diffusion Nucleation Crystal growth

to equilibrate process
composition

Time

(c)
Polymerization
and nucleation Crystal growth

Structure-directing
T (Si,Al,Ti,....) precursors Mineralizers (OH-,F-)
agents

Fig. 2.118: Schematic representation of the (a) nucleation rate and (b) crystal growth rate of
zeolites described with a typical S-shaped curve and (c) related rearrangements from amorphous
particles into crystalline zeolite during the synthesis (Reproduced with permission from [61]).

Templates of the NR4+ type are used, such as tetra-alkylammonium hydroxide.

Alkylphosphonium cations (R4P+) and organic complexes can be also applied.
Reactants are mixed according to a synthesis recipe. Usually, different solutions are
prepared under continuous stirring and mixed to obtain a unique thick milky white
gel. pH of the solutions is an important control parameter. Other critical parameters
are: the molar ratios (OH–/SiO2 and Si/Al); temperature; the sources of silica and
alumina; presence of contaminants; the order of the addition of the reactants;
aging and ripening prior to crystallization; synthesis time; the nature of the organic
templates; pre-treatment of the reactants; inclusion of special additives; presence
of seeds and stirring rate and so on.
168 2 Engineering catalysts

In laboratory-scale preparations, once the gel is ready it is transferred to

Teflon-lined autoclaves, where the nucleation and crystallization of the gel takes
place. Condensation reactions occur faster than hydrolysis which leads to formation
of nuclei and growth by the addition/condensation of monomers. The particles are
negatively charged, preventing the formation of aggregates.
The Teflon-lined autoclaves are put in an oven with a constant predetermined
temperature for a period of time according to the synthesis procedure. In a labora-
tory scale, crystallization can be performed in a static mode, under internal rotation
or under rotation of the autoclaves.
When the crystallinity is plotted as a function of time, S-shapes curved are usu-
ally obtained (Fig. 2.118, 2.119), where the fraction of the transformed material F can
be described by the Avrami-Erofeev equation (eq. 2.36). with n determined by data
fitting. S-shaped features are caused by autocatalytic behavior at low conversions
because of slow nucleation and decline of the crystal growth rate with an increase
in crystallinity.

2.5
Av. crystal size (µm)

1.5

0.5
20 µm
0
0 20 40 60 80 (B)
(A) Time (h)

5 µm

(C)

Fig. 2.119: Effect of synthesis time on ZSM-5 zeolite. (A) Average crystal size as a function of synthesis
time, (B) morphology after 6 h, (C) morphology after 72 h. (Reproduced with permission from [62]).

An increase of the synthesis temperature at constant alkalinity (pH) increases nu-

cleation and linear growth rate. The induction time in zeolite crystallization de-
creases rapidly with an increasing temperature, up to a certain point. Similarly, the
 2.3 Preparation of catalytic materials 169

induction time decreases and crystal growth is enhanced by an increase in pH at

constant temperature. Alkalinity also has an effect on the Si:Al ratio of the zeolite.
After crystallization, the autoclaves are quenched with cold water. The formed zeo-
lite crystals are filtered, washed and dried. Washing is required to reduce high alkalin-
ity created by the addition of NaOH. In order to remove the organic template, zeolites
are typically calcined after drying (100 °C). A calcination step is needed to remove
the SDA. It is performed at 500–600 °C for several hours, depending on the zeolite.
Different calcination techniques have an effect on the properties of the synthesized zeo-
lite. After calcination, the zeolite is in the sodium form (the as-synthesized form, de-
noted by Na at the beginning of the name, for example, Na-ZSM-5) because it contains
sodium ions to balance the framework charge induced by Al as T-atoms. Acidity can be
introduced in the as-synthesized zeolite by ion-exchanging sodium cations with NH4Cl
or NH4NO3 to achieve an ammonium form. The ammonium form of zeolite is dried and
calcined to obtain the proton form of zeolite.
A flow sheet for the preparation of a sodium form of zeolites is given in Fig. 2.120.

NaOH Al-source H2O Si-source

Vapor
Fresh-
water
Product to
drying and
Mixing vessels crushing
Belt conveyer

Slurry
Vapor

Slurry
Crystallization
Concentrated mother liquor

Fig. 2.120: Preparation of a sodium form of zeolites. (From [51]).

The catalyst in the final process always requires further fine-tuning by a number of
secondary treatments. It is worth considering as an example the production of fluid
catalytic cracking catalysts. Detailed description of these catalysts is given in
Chapter 4. Typical commercial cracking catalysts are a mixture of rare earth-Y zeo-
lite and SiO2–Al2O3 amorphous aluminosilicate as a binder providing the required
mechanical stability. Small pore ZSM-5 is applied as a co-catalyst with rare earths Y
zeolite to increase the octane number of gasoline. The sodium form of zeolites Y is
170 2 Engineering catalysts

first produced followed by an exchange of sodium with rare earth ions, which is an
example of a secondary treatment. Although there is a clear advantage to using
binders the drawbacks are associated with blocking of the zeolites pores by the
binder and longer diffusion paths. The flow scheme for the production of rare earth
zeolites Y is given in Fig. 2.121.

Al2(SO4)3
RE(NO3)2
NH4NO3

Na2SiO3
NaAlO2
H2O

H2O

H2O

H2O

H2O
NaOH Al2(SO4)3

1 2 3 4 5 Na2SiO3
6 7 8

NH4NO3
NaAlO2

H2O 9
H2O

H2O Zeolite
10 14 H2O
11 13 15 17
H2O H2O
16 REY
18

Fig. 2.121: Synthesis of rare earth zeolite. (From [52].) 1, 2, 3, 4, 5 = vessels for ammonium nitrate,
rare earth nitrate, NaAlO2, Al2(SO4)3 18 H2O and sodium silicate respectively; 6, reactor for sodium
aluminosilicate synthesis; 7,crystallizer; 8, crystallizer for synthesis of zeolite seeds; 9, filter; 10,
vessel for ion exchange of NH4+ into rare earths; 11, milling; 12, heat exchanger; 13, vessel of
additional ion exchange of NH4+ into rare earths; 14 cooler; 15, mixer; 16, filter; 17, vessel for
repulping of zeolite; 18, suspension storage. RE, a rare earth element.

Water solutions of ammonium nitrate, rare earth nitrate, NaAlO2, sodium silicate and
aluminum sulfate are prepared in vessels 1–5. Sodium aluminosilicate gel is obtained
in vessel 6 and a solution of NaAlO2 is added. After cooling, the mixture is routed to a
crystallizer in 7 to which a seed of zeolites Y is added from vessel 8 in the quantities
corresponding to 1–3 wt%. Mixing of the Na-Y zeolites seed with the reacting mixture
in vessel 7 lasts for ~1 h, thereafter hydrothermal synthesis is performed for ~24 h at
371 K followed by cooling to room temperature. After filtration (pos. 9) the mother
liquor is separated and the zeolites are washed with water and ammonium nitrate to
ion exchange sodium for ammonium. This is followed by primary ion exchange with
 2.3 Preparation of catalytic materials 171

rare earth (Ln3+) nitrates in vessel 10. After milling (pos. 11) and additional ion
exchange (pos. 13) at an elevated temperature (up to 473 K) the zeolite is cooled
(pos. 14), filtered (pos. 16) and transported to a vessel 17. From this vessel it is taken
as rare earth zeolite Y or routed to another vessel for storing in the form of suspension
(pos. 18) to be used in the synthesis of, for example, microspherical FCC catalysts.
The flow scheme of the microspherical catalyst synthesis is given in Fig. 2.122.

NaOH Sand H2O

H2O REY
6
1 2 4
9

Al(OH)3 H2SO4 H2SO4

H2O
3 2 5 H2O
11

14 15

6
7 12

H2O
Air
(NH4)2SO4
17
8 13
H 2O
10
Steam
Gas 16
H2O Air

Fig. 2.122: Microspherical catalyst synthesis. (From [52].) 1, vessel for water glass synthesis;
2, filter; 3, vessel for the preparation of aluminum sulfate solution; 4, vessel for the preparation
of water glass solution; 5, vessel for preparation of aluminum sulfate solution with sulfuric acid;
6, coolers; 7, mixer for preparation of the sol of zeolitic aluminosilica gel; 8, column for
preparation of microspherical gel particles; 9, vessel for rare earth zeolite suspension; 10, vessel
for water solution of ammonium sulfate; 11, vessel for gel syneresis; 12, vessel for gel activation;
13, vessel for washing out microspherical gel particles from sulfate and other ions; 14, drying
column; 15, column for calcination; 16, burner for flue gases generation; 17, storage.

In vessel 1 (Fig. 2.122) sand, water and sodium hydroxide are introduced fol-
lowed by the preparation of water glass at 98–110 °C, which is filtered to remove
172 2 Engineering catalysts

unreacted sand particles (pos.2) and transported to vessel 4 where the solution of a
needed concentration is prepared. Aluminum sulfate is first prepared in vessel 3,
then filtered and transported to vessel 5, where a working solution of aluminum sul-
fate is prepared by adding sulfuric acid. This is necessary as otherwise the solution
containing the required amount of alumina is too basic, which will result in the
generation of a zeolite with overly large pores and lower bulk density. Vessel 9
contains water suspension of rare earth zeolite Y. Solutions from vessels 4, 5 and
9 are routed to a mixer-sprayer 7, where water glass, aluminum sulfate and the
rare earth zeolite Y suspension are mixed at the ratio to assure pH 7.5–8.6.
Microdroplets of 2–5 nm leave vessel 7 and enter the column 8, which is filled
(~3 m) with transformer oil. The microdroplets pass through the column at ~
10–12 °C to form the gel. The large column diameter (1.5 m) is needed to prevent
sticking of the gel to the vessel walls. A droplet passes through the transformer
oil in 8–11 s, while 5–8 s are needed for coagulation. The coagulation rate in-
creases with temperature increase and concentration of the parent solutions and
depends on pH and presence of electrolytes. As mentioned above, an acid is
added to the mixture to ensure the required coagulation rate. Passing through
oil, a gel of microspherical particles is formed that contains 90% water and 10%
dry substance. Syneresis (expulsion of a liquid from a gel) is done in vessel 11 at
40–50 °C for 6–24 h. Replacement of sodium ions by ammonium occurs in vessel
12 at 50–80 °C for another 6–12 h. In order to prevent removal of Al3+ and rare
earth ions during washing (pos. 13) they are added in small amounts to the
washing step while sodium, ammonium and sulfate ions are washed out.
Washed gel particles can be treated with water solutions of surfactants to pre-
vent the collapse of the gel structure during drying (pos. 14). Washing is fol-
lowed by drying of the zeolite (pos. 14) and calcination (pos. 15). The flue gases
needed for drying and calcination are generated by burning natural gas in the
presence of steam and air (pos. 16).
During drying the flue gases enter the dryer from the bottom, while zeolite
suspension falls in the opposite direction. Temperature along the dryer changes
from ~315–325 °C at the top to 475–485 °C at the bottom. This process leads to a
material with ~8–10% water. Calcination can be performed in a fluidized bed
(Fig. 2.123). The vertical vessel for calcination has an extended upper part, while
the bottom part has an inlet for flue gases. The amount of flue gases is con-
trolled to allow fluidization, while at the same time preventing carryover of
smaller particles. Calcination starts at 600–650 °C for 10 h, followed by a de-
crease in temperature of the flue gases leaving the burner. At 250–300 °C the
fuel is shut off, while air is still added. The gradual decrease of the catalyst tem-
perature for 4–6 h to 80–90 °C allows the transfer of the catalyst to metallic
drums for storage.
 2.3 Preparation of catalytic materials 173

Flue gases

2 Flue gases

Catalyst

Catalyst
Flue gases

Fig. 2.123: Fluidized bed calcination unit. (From [48].)

2.3.5.7 Metal organic frameworks

These materials with a very high surface area (1,800–3,000 m2 g−1) have inorganic
building blocks interconnected with organic linkers. The organic units are typically
mono-, di-, tri-, or tetravalent ligands. The structure and properties of the metal or-
ganic framework (MOF) are influenced by the choice of metal and the linker.
Similar to zeolites, MOFs exhibit shape selectivity with more flexibility of chemical
composition.
MOFs are synthesized by hydrothermal or solvothermal methods, although,
contrary to zeolites, organic ligands remain intact. A variety of options for post-
synthetic modification of MOFs is thus available. One example of the use of MOFs
in catalysis are 2D, square-grid MOFs containing single Pd(II) ions as nodes and
2-hydroxypyrimidinolates as struts (Fig. 2.124). If all the Pd centers are catalytically
active in reactions such as Suzuki C–C coupling, when redox oscillations of the
metal nodes between Pd(+2) and Pd(0) intermediates are needed to accomplish the
reactions, these changes in the coordination number would lead to destabilization
and potential destruction of the original framework.
For hydrogenation reactions on MOFs it is also possible that reactions occur on
the surface of MOF-encapsulated palladium nanoparticles.
174 2 Engineering catalysts

Fig. 2.124: A palladium 2-hydroxypyrimidinolate metal organic framework (MOF) used to catalyze
efficient oxidation of allylic alcohols to aldehydes. (From [63].) blue, N; red, O; yellow, Pd; gray, C;
light blue, Cl.

Besides high costs, the drawbacks of MOFs are air/moisture-sensitivity and

scaling up issues associated with their low-temperature stability, which limits the
application of MOFs in various catalytic reactions.

2.3.5.8 Mesoporous materials

Because of the limitations caused by zeolites having pores below 1.1–1.5 nm
there is an interest in materials with characteristics similar to zeolites in terms
of acidity but with larger pores. Such materials could separate and can catalyze
reactions with larger molecules, which are otherwise not able to penetrate into
the pores of zeolites. From the point of view of synthesis the size of the templat-
ing molecule determines whether ordered mesoporous materials or zeolites are
formed. Isolated short alkyl chain quaternary ions lead to formation of micropo-
rous materials, while self-assembled surfactant structures with long alkyl chain
quaternary ions (such as C16H33(CH3)3NH3+ Br–) the direct formation of mesopo-
rous molecular sieves (Fig. 2.125). In solution, surfactants can self-assemble to
form micelles, rods, sheets and 3D structures. Synthesis of mesoporous materials
is similar to zeolites, as it is also conducted in basic media.
One of the most famous examples of mesoporous materials is MCM-41 silicate,
which is highly ordered and even gives a diffraction pattern although the glass-like
 2.3 Preparation of catalytic materials 175

Self-assembly

Condensation

Surfactant
removal

Fig. 2.125: Formation of mesoporous materials. (Modified from [64].)

silicate walls are not crystalline. The pore size distribution in amorphous MCM-41 is
usually quite narrow although not as tightly defined as for zeolites.
Acidity of MCM-41 is much less than of zeolites, which along with rather expen-
sive templates somehow hinders potential use of this material as a catalyst despite
hundreds of published scientific papers.
Hexagonal mesoporous silica SBA-15 and cubic cage SBA-16 are two other in-
teresting materials that belong to the group of mesoporous materials. Hexagonal
mesoporous silica SBA-15 is synthesized in acidic media (pH 1–2) using a non-
ionic tri-block copolymer, such as pluronic P123 (PEO)20(PPO)70(PEO)20 or HO
(CH2CH2O)20–(CH2CH(CH3)O)70)–(CH2CH2O)20H containing poly(oxyethylene) PEO
and poly(oxypropylene) PPO units. Cylindrical micelles are formed with silica par-
ticles assembling around the template with hydrophobic heads and hydrophilic
tails.
SBA-15 has thicker pore walls than MCM-41 and possesses monodisperse pores
of 5–20 nm, while microporosity can be also introduced. Thermal and hydrothermal
stability is also higher than for MCM-41. Synthesis of mesoporous SBA-15 proceeds
through dissolving tri-block copolymer, preparation of gel in acidic media, ripening
of the gel at 373 K for 24 h, filtration, washing with distilled water, drying at 373 K
and calcination at 813 K.

2.3.5.9 Layered compounds

Layered compounds can be categorized into three classes: (a) neutral ones (e.g., bru-
cite, phosphates and chalcogenides), (b) compounds containing negatively charged
layers with compensating cations in the interlayer space (e.g., montmorillonite) and
176 2 Engineering catalysts

(c) materials containing positively charged layers with compensating anions in the
interlayer space, such as layered double hydroxides (e.g., hydrotalcites).
A specific feature of cations or anions in the interlayer space of this relatively
high surface area layered materials is related to their ability to ion exchange.
Another feature is related to swelling properties.
Cationic clays, which may exhibit both Brønsted and Lewis acid sites, contain
two types of sheets comprising Si(O,OH)4 tetrahedron and M(O,OH)6 octahedron
(M = Al3+, Mg2+, Fe3+ or Fe2+) as building units. They are typically synthesized from
minerals possessing negatively charged aluminosilicate layers compensated with
interlayer cations. The Brønsted acidity is due to OH groups, while Lewis acidity is
associated with Al3+ substituting Si4+ in tetrahedral sheets. Another parameter
influencing hydrophilic–hydrophobic properties is the type of an exchangeable
cation.
Pillaring of clays to develop large pore and strongly acidic materials was initiated
by replacing the interlayer exchangeable cations first in smectites with tetraalkylam-
monium ions. Organic pillared clay minerals are, however, thermally unstable at
high temperatures (above 250 °C), leading to the interlayer collapse. Below the de-
composition temperature, organic pillared clays can be, however, applied in catalysis
and adsorbents. Intercalation of clay minerals with inorganic species allows to gener-
ate pillared clays, which retain their micro- and mesoporosity after heating above
300 °C.
Natural and modified clays have found applications in various acid-catalyzed
reactions, including cracking and alkylation of aromatics, as well as in bulk and
fine chemical synthesis.
Hydrotalcites or anionic clays (Fig. 2.126) are layered double hydroxides of gen-
eral formula [M2+(1–x)M3+x(OH)2]x+(An-x/n) yH2O, where M2+ is typically Mg (or Ni, Zn,
Cu), and M3+ is typically Al, but can be Mn, Cr, Co or Fe, while A is any anion (e.g.,
CO23 − , NO3− , F– and Cl–). They are formed in nature by weathering of basalts or from
precipitation in saline water. A typical example of a hydrotalcite is Mg6 Al2CO3
(OH)16·4(H2O). The main application of hydrotalcites is in base-catalyzed organic
reactions.

Anions M2+ or M3+

Water molecules

OH-

Fig. 2.126: Structure of hydrotalcite.

 2.3 Preparation of catalytic materials 177

2.3.5.10 Ceramic and metallic monoliths and foams as catalyst supports

Ceramic and metallic monoliths are typically used as catalyst supports [65].
Ceramic monoliths are made from cordierite 2MgO × 2Al2O3 × 5SiO2 or mullite
3Al2O3 × SiO2 by extrusion with subsequent firing or by corrugation. This is followed
by washcoating with γ-, θ- and δ-alumina, enhancing the original surface area of
just few m2/g to approximately 100 m2/g. After washcoating again firing at an
appropriate temperature is performed. The components, having a similar thermal
coefficient of expansion, are able to withstand high operation temperature
(1,200–1,300 °C) as well as heating and cooling cycles without disintegrating.
Typical cordierite honeycombs contain 400–600 channels per square inch with
the wall thickness of approximately 0.15 mm.
Metallic monoliths, made of Fe-Cr-Al alloy by corrugation, have thinner walls
(ca. 0.05 mm), better warm-up properties and are more mechanically strong than
cordierite. Corrugation is followed by rolling up or folding into monoliths of re-
quired size and shape. A suitable oxide layer should be formed to allow adhesion of
the alumina washcoat.
As already mentioned previously the surface area of cordierite is low and the
active metal cannot be deposited directly on the monolith. Therefore, there is a
need for applying wash-coat to increase the surface area and to deposit the active
component, which is typically silica or alumina, on the wash-coat. A wash-coat
thickness varies in the range of 15–100 μm providing 15–30 m2 g−1 surface area. A
colloidal solution, for example, of bohemite alumina, can be used for wash-coating.
In addition to alumina, the slurry contains stabilizers and binders. Stabilizers
such as ceria, zirconia and barium oxide are added to the washcoat to prevent
phase changes at the higher temperatures or to improve catalytic performance. In
three-way automotive exhaust catalysts, cerium oxide acts as an oxygen sink ab-
sorbing it during lean operation and promotes the water-gas shift reaction generat-
ing hydrogen and thereby improving catalyst performance under rich conditions.
The same function of oxygen absorption under lean conditions is provided by bar-
ium oxide. Zirconia in the washcoat composition stabilizes and serves in the three-
way catalysts as a support for rhodium.
To ensure chemical and physical binding of the washcoat of the honeycomb
surface and filling the large macropores of the ceramics, alumina in slurry should
be ground, providing a size distribution mainly in the range of 1–10 mm that is com-
patible with the pores of the ceramic wall.
The pH is usually controlled when preparing the colloidal solution and the ac-
tive components can be incorporated both directly in the slurry or after the wash-
coating. Typically, the slurry contains 10–20 wt% of the solid material depending
on the monolith cell density. The laboratory procedure of wash-coating comprises
first dipping the monolith into the wash-coat slurry with forced flow. The wash-coat
slurry contains typically more than the calculated desired amount of the wash-coat.
Dipping is followed by blowing the extra slurry out of the channels, drying and
178 2 Engineering catalysts

calcination of the monolith. If the desired amount of the wash-coat (usually

20 wt%) is not reached the procedure is repeated.
Often wash-coating is non-uniform along the channel unless appropriate drying
methods are applied (Fig. 2.127).

Fig. 2.127: Photograph of the cross section of monolithic catalysts showing the impregnation
profiles obtained after static drying (left), freeze drying for 1 h, followed by static drying (center),
and freeze drying for 24 h (right). (Reproduced with permission from [66]).

In general, it can be stated that detailed information on the preparation of mono-

liths at the commercial scale is typically not released by the catalyst manufacturers
keeping it proprietary.
Metallic or ceramic foam are special types of structured catalysts with the ad-
vantages, that a shape required to fit the reactor size can easily be produced using
a template (Fig. 2.128).

Fig. 2.128: Metallic foams. (From [67].)

Ceramic foams have several other important advantages besides the ability to
match the shape and size of the reactor, such as much lower than packed beds’
pressure drop and much larger external surfaces allowing higher effectiveness
 2.3 Preparation of catalytic materials 179

factors and better heat transfer. The latter is important in avoiding local overheat-
ing and hot spots. Ceramic foam catalysts can be thus advantageous in case of
strongly exothermic or endothermic reactions. The open pore structure (Fig. 2.128)
can also help to control selectivity.
Lower solid loading when compared with packed beds (i.e., 10−20% vs 45−55%,
respectively) inevitably gives a lower productivity per reactor volume. This should be
compensated by better mass transfer within the catalyst layer, larger loading or even
more active catalytic phase.
Because of a more complex structure and a laborious procedure to prepare
foams discussed below, the costs of manufacturing foams are apparently higher
than that of more conventional catalyst shapes.
Complications in utilization of ceramic foam catalysts are related to their brittle
character and thus a potential breakage during manufacture and loading into a re-
actor. Moreover, long lengths of preformed ceramic foam might be troublesome to
charge in not perfectly straight reactor tubes.
Ceramic foams have lower pressure drop compared to packed beds, and com-
pared to conventional monoliths the flow pattern in gas–liquid system is closer to
that of packed beds.
Metallic foams could be either closed-cell or open-cell materials. The most com-
mon way to prepare closed-cell foams is to create gas bubbles in molten metal by
gas injection. Stabilization agents are introduced into the melt to prevent pore coa-
lescence. The size of the pores, or cells, is usually 1–8 mm. Open-celled metal
foams are usually replicas that use open-celled polyurethane foams as a skeleton
when first physical vapor deposition of metal is made onto reticulated polyurethane
foam followed by burn-off from the polymer foam.
Metal foams could be based on aluminum, nickel, chromium, iron or alloys.
Because of the large surface area, high thermal conductivity, mechanical durability
and corrosion resistance metal foams can be used as catalyst supports offering in-
creased heat and mass transfer and reduced diffusion resistance. Even if these ma-
terials have a potential to replace some conventional pellets or packing materials
the introduction of metallic foams into industry requires more quantitative data on
heat transfer coefficient, temperature profiles, stability and especially catalytic data
to support more widespread use of these foams.
Ceramic and carbon forms are also produced commercially. Mechanical strength
of these foams is lower compared to metal and cordierite monoliths. Polyurethane
foam templates with different pore size (Fig. 2.129) can be used for the preparation of
ceramic foams enabling unlimited shape and size combinations.
In order to prepare a ceramic foam block (Fig. 2.9) a defined piece of template is
dipped into the initial water solution, containing, for example, alumina as the main
compound and additives such as magnesia, titania and polyvinylalcohol. Uniform in-
filtration of the polymer by the aqueous slurry, composed of the ceramic particles
with the 0.1−10 μm diameter particles along with wetting agents, dispersion
180 2 Engineering catalysts

Fig. 2.129: Ceramic foams: polyurethane templates.

stabilizers and viscosity modifiers, is challenging. For better adherence, the polymer
foam might be pretreated (etched). Improved wettability needed for better quality of
the ceramic foam can be achieved by applying mechanical pressure or by vacuum
processing. Shear thinning of the fluid (i.e., the non-Newtonian behavior with viscos-
ity decreases under shear strain) is necessary for polymer template coating.
Low-viscosity suspensions are mainly used in dip coating and excess slurry is
removed by blowing air through the foam or by squeezing and kneading. The wet
foam is then dried and calcined in air at temperatures of 1,000−1,700 °C. Repeated
slurry coating is carried out with the green form for adjustment of the strut thick-
ness or filling cracks for crack filling or for functionalization of ceramic foams.
Recoating might be carried out with the so-called green foam prior to calcination or
after an intermediate sintering of the ceramic foam.
The laboratory-scale preparation procedure thus includes dipping, drying,
re-dipping, squeezing, centrifuging and air blowing followed by calcination of
the impregnated template. During the latter stage parameters such as ramping,
temperature and time are important in preparing the desired material. During the
calcination the template is burned and a replica of the template structure remains
consisting of α-Al2O3 and additives. Firing is a delicate process and can be accom-
panied by a foam block bending. The simplified scheme of the replica method for
synthesis of foams is presented in (Fig. 2.130).
The final steps in the preparation of the support are first wash-coating the foam
with a high surface area support material and then deposition of the active compo-
nent, which will be discussed in Section 2.3.6. The wash-coating of foams is princi-
pally made in the same way as the wash-coating for cordierite monoliths.

2.3.6 Supported catalysts

In supported catalysts, catalytically active compounds are attached to a support

with a large specific surface area. Industrial preparation of catalysts is considered
to be strategic and thus details of preparation are typically not disclosed or are
 2.3 Preparation of catalytic materials 181

Dispersion Foam template

Replication

Recoating Green-body

Burn out

Fig. 2.130: Processing scheme for foam

manufacturing (Reproduced with permission from
Sintering
[68]).

protected by patents. Patents typically feature broad claims and thus the exact
preparation procedures applied in industry are not revealed. Deposition of the pre-
cursor on the support surface can be done by using aqueous solutions and “wet”
methods of ion exchange, adsorption, precipitation and impregnation.
Alternatively, dry methods (i.e., atomic layer deposition) involving the gas-solid in-
terface could be applied. Initial deposition is followed by transformation of the pre-
cursor into the required active phase (metal, oxide, sulfide). An overview of the
methods applied for preparation of supported metals is given in Tab. 2.6.

Tab. 2.6: An overview of the methods for preparation of supported metals (Reproduced with
permission from [58]).

Synthesis method Metal Metal Advantage Challenge

loading dispersion

Colloidal synthesis Low–medium Medium Controlled size/ Anchoring on support

shape
Chemical vapor Low Monolayer Monolayer control Special equipment
deposition needed
Ion adsorption Low High Atomic dispersion Gradients over
support
Deposition High High–medium Reproducible Limited applicability
precipitation
Coprecipitation Very high Medium Very high Catalyst activation
loadings
Impregnation and High Medium Widely applicable Limited
drying reproducibility
182 2 Engineering catalysts

Placement of a support in contact with a solution containing a precursor leads

to electrostatic interactions, chemical binding to the surface and formation of sur-
face compounds. In addition, the support could undergo dissolution, at least
partially.
An ion exchange method involves the replacement of an ion in electrostatic con-
tact with the support by another ion. Because of the equilibrium nature of ion ex-
change a single operation might not be sufficient and several successive operations
should be adopted. Alternatively, the exchange solution is fed continuously through
the solid bed.
Adsorption of an active precursor on charged surfaces is typically done from liq-
uid and the precursor should be selected to have a charge opposite to that of the sup-
port. The latter depends on the isoelectric point of the oxide, pH and ionic strength of
the solution. The distribution of the active precursor depends on the density of the
adsorption sites on the surface, which is affected by the pH level. These methods in-
volve a subsequent washing step, contrary to impregnation, when a certain volume
of a precursor containing solution is put in contact with the support.

M M OH
M
O O OH
M M OH
OH M
O
OH OH
M M M
OH
O + H2O O + H2O
OH
M M M
OH
O OH
OH
M M M
OH
O O
OH
M M M
OH
O O

Fig. 2.131: Reaction of water with support hydroxyl groups.

2.3.6.1 Adsorption
Reaction of water with catalyst supports leads to hydroxylated surfaces (Fig. 2.131)
and formation of M–O− or M–OH2+ species.
The surface is thus charged depending on the solution pH. Surface charge is
equal to zero (isoelectric point or point of zero charge) at some particular value of
pH depending on the support pH. An increase in the pH levels results in the hydro-
lysis reaction of the surface OH groups shifting equilibrium from left to the right
hand side on Fig. 2.131. At a pH larger than the pzc the surface is negatively
charged, welcoming the adsorption of cations. For example, amine complexes
[Cu(NH3)4]2+ are very suitable for adsorption on silica at an elevated pH, when the
surface is negatively charged.
 2.3 Preparation of catalytic materials 183

The pzc of γ-alumina is between 7 and 9, so it can adsorb cations and anions.
Special care should be taken during catalyst preparation to the charge of the sur-
face and the charge of species to be deposited. For example, at low pH when the
alumina surface is positively charged platinum can be adsorbed from hexachloro-
platinic acid H2[PtCl6] as anions [PtCl6]2−. Fixation of anions on the surface is typi-
cally fast and does not limit the overall kinetics. In the case of platinum deposition
on alumina from H2[PtCl6], because of strong interactions between the complex and
the surface, platinum tends to adsorb strongly and rapidly, leading to heteroge-
neous distribution of the metal on the support. In order to avoid this, a competitor
(HCl) is added in industrial impregnations to ensure homogeneous distribution of
platinum. The conventional interpretation is that effective concentration of the plat-
inum precursor in the solution is influenced by the following equilibrium with Cl–
playing the key role:

½PtCl6 2 − =support + 2Cl − solution $ ½PtCl6 2 − =solution + 2Cl − support

An alternative explanation [69] assumes that the protons in HCl are needed to
charge the surface sufficiently for adsorption to occur.
In addition, HCl acts as a stabilizing agent for the coordination sphere of plati-
num, as otherwise [PtCl6]2− can lose several chloride ligands.
A successful recipe for one support will not necessarily work for another one.
Thus at pH 8 platinum could be adsorbed on silica from the cation using [Pt(NH3)4]
(NO3)2, while the same precursor will be inefficient for alumina (Fig. 2.132). Isoelectric
points of some typical supports are given in Tab. 2.7.
Adsorbed [Pt(NH3)4]2+

0.3
(mmol g–1)

Silica gel
0.2

0.1
γ-Alumina

0
6 7 8 9 10
pH

Fig. 2.132: Adsorption of [Pt(NH3)4]2+ from the solution on the surface of silica gel and γ-Al2O3.
(Redrawn from [70].).

The discussion above was mainly concerned with hydroxyl groups on the oxide
supports, however, the same careful attitude towards adsorption is valid for car-
bons supports, where the surfaces are also charged because of the presence of vari-
ous functions groups, such as carboxylic ones (Fig. 2.133). Thus H2[PtCl6] is used at
184 2 Engineering catalysts

Tab. 2.7: PZC of supports.

Support SiO CeO TiO FeO ZrO AlO ZnO CoO MgO

pzc    .–. . .– .– . .–.

Acidic groups

Lactol
Basic groups
O
Carboxyl Lactone O OH
O OH O O
Phenol Chromene
O Ketones
OH O O

Fig. 2.133: Functional groups on carbon support.

low pH on positively charged surfaces, while [Pt(NH3)4](NO3)2 should be applied at

high pH.
Point of zero charge of the support and pH are thus important parameters influ-
encing the electrostatic interactions between the metal complex and the support.
Regulation of pzc could be done by changing temperature (pzc is increasing for alu-
mina and decreasing for silica with temperature rise) and/or selective doping of the
support surface. In selective doping the dopant should not deactivate the catalyst.
As an example of selective doping, in alumina modifications with Mg+2 or F– ions
change pzc in the opposite directions. Even if pH is an important parameter, it
should be noted that measurements and control of pH are difficult tasks. Changes
in pH as mentioned above can cause dissolution of support and in addition can
change metal speciation, which leads to adsorption of other types of species. Such
changes in chemistry of adsorption should be thus taken into account during cata-
lyst preparation both at the experimental level influencing post-treatment and at
the theoretical level of modeling.
The type of ligand, in particular its charge and bulkiness, influences the nature
and the strength of the metal complex-support interactions. For example, shielding a
bulky ligand can decrease the interactions between the metal complex and the charged
surface. Moreover, a hydrated complex could be too large to enter the support pores.
Several metal precursors are suitable for catalyst synthesis. It should be noted
that metal salts (Tab. 2.8) are typically more expensive than the bulk metals. As
chloride containing precursors could suppress catalyst activity, thus nitrate
 2.3 Preparation of catalytic materials 185

precursors are applied or alternatively efficient washing methods should be used to

wash away chloride.

Tab. 2.8: Metal cations and anions used in catalyst

preparation.

Metal Anion Cation

Cobalt Co(NH)++
Nickel Ni(NH)++
Copper Cu(NH)++
Ruthenium Ru(NH)Cl++
Rhodium RhCl− Rh(NH)Cl++
Palladium PdCl− Pd(NH)++
Silver Ag(NH)++
Iridium IrCl− Ir(NH)Cl++
Platinum PtCl− Pt(NH)++
Gold AuCl−

In general, formation of different species with different charges should be modeled

as a function of pH and other concentrations (Fig. 2.134). The surface ionization re-
actions are described by two constants (Fig. 2.135), which along with the surface
density of hydroxyls and the dielectric constant, characterize the charging proper-
ties of an oxide defined through pzc = [0.5 (pK1 + pK2)].

–
HO HO O
+
H C C C
O O O

pH < PZC < pH

Fig. 2.134: Surface charge dependence on pH for carbon supports with carboxylic groups.

In order to model adsorption behavior (Kads) a simple Langmuir type adsorption

model for the metal adsorption capacity is often applied

cp K
a = a∞ (2:37)
cp K + 1

where cp is equilibrium concentration, K is adsorption equilibrium constant and a∞

is maximum sorption capacity. As a typical example, adsorption isotherms of
H2PtCl6 on carbons are presented in Fig. 2.136.
Adsorption models available in the literature very seldom consider lateral inter-
actions between adsorbed ions, site heterogeneity and oxide dissolution. Such dis-
solution could occur when deposition of the active agent happens at a pH far from
186 2 Engineering catalysts

Kads, anion
OH2+ [PtCl6]2–
pH < PZC
K1
(pH shift)
PZC OH [H]+

K2
Kads, cation
pH > PZC O– [(NH3)4Pt]2+

Fig. 2.135: A schematic presentation of the mechanism of electrostatic adsorption of platinum.

(From [71].)

CG-48A
Active carbon

1
Sibunit
Ptads (mmol g–1)

Pyrocarbon

0.5

Mogul L
Coke
0
0 0.05 0.1
Equilibrium Pt concentration
in solution (mol L–1)

Fig. 2.136: Adsorption isotherms of hexachloroplatinic acids on carbon (From [50]).

the pzc of the support, i.e. the point of maximal stability of the solid because of
minimum solubility.
Dissolution of the support could also be beneficial to a certain extent. New
hydroxyl groups created because of silica dissolution in a basic medium act as ad-
sorption sites, resulting in an increase of cations adsorption on silica at pH higher
than 10.5.
It should be noted that not only the desired ion but also the associated counter-
ion could interact with the support. Moreover, the counter ions can determine the
precursor solubility.

2.3.6.2 Impregnation and drying

Impregnation, which is in a way similar to adsorption, could be done by contact-
ing a certain volume of solution with a precursor and the support (Fig. 2.76). The
 2.3 Preparation of catalytic materials 187

method is used to obtain higher loading than in simple adsorption and to apply
active precursors that do not adsorb easily on support. In this preparation method
the solution of the precursor is broken down into small discontinuous elements
presented in the pore of the support by gradually evaporating the solvent. The
wetting time could be modified by changing the temperature of impregnation,
which has an influence not only on pzc but also on the precursor solubility and
viscosity of the solution. Mass transfer of the impregnated compound within the
pores during impregnation and drying determines the concentration profiles.
Solubility of the precursor in the solution limits the maximum loading, thus suc-
cessive impregnations with drying and even calcinations between these impreg-
nations are applied to increase the loading, as will be discussed separately in
Section 2.3.6.3.
In wet impregnation an excess of solution is used and after a certain time the
solid is separated and the excess solvent is removed by drying. In this non-
exothermal method the pores are filled with the neat solvent prior to impregnation.
When this solvent saturated support is introduced into the solution containing the
precursor there is a concentration gradient between the external solution and the
pores driving the precursor salt in the pores.
In dry methods (pore volume impregnation or incipient wetness) the volume of
the solution containing the precursor corresponds exactly or approximately to the
volume of the pores. One of the disadvantages of the method is broad particle size
distribution. If pH is not adjusted and, for example, impregnation is done at condi-
tions when pH is close to pzc, only few hydroxyl groups are available and non-
adsorbing metal particles might agglomerate giving large crystallites.
In dry methods a replacement of the solid-gas interface by a solid liquid one
causes a significant decrease in the free enthalpy. This results in heat release,
which in general is of minor importance for impregnation. In some special cases
this heat release might influence impregnation by provoking unwanted precipita-
tion, limiting solubility or modifying the surface.
When solution is sucked into pores by capillary forces, air bubbles are trapped
and could burst the support and lower the mechanical strength. This is caused by
considerable forces exerted on the pore walls being in contact with these bubbles
during dry impregnation. The reason is in the high pressure (several MPa) inside
the compressed bubbles defined by the Young-Laplace law:
2γ
ΔP = (2:38)
r

where ΔP is the pressure difference inside and outside the bubbles, γ is the liquid-
gas interfacial tension and r is the liquid-gas meniscus radius.
In order to prevent or limit the bursting of the support during impregnation, a
surfactant could be added to the solution or impregnation could be done in vac-
uum, complicating the whole procedure.
188 2 Engineering catalysts

Time needed for the liquid to fill the pores is very short (a few seconds) in dry
impregnation, although in practice longer operation times are used. Contrary to
capillary impregnation in wet methods (diffusional impregnation) migration of the
precursor into the pores is much longer and distribution of the solute is governed
by diffusion into the pores (Fick law can be applied) and adsorption on the support
(typically Langmuir model is used). As already mentioned above, adsorption de-
pends on the adsorption capacity and the equilibrium constant of adsorption.
Phenomena during wet and dry impregnation are illustrated in Fig. 2.137. The
characteristic time t needed to attain equilibrium in wet impregnation is propor-
tional to the square of a pellet radius, thus is increasing with the pellet size and is
far from instantaneous but requires several hours.

Adsorption Adsorption

Air
Dissolution
Diffusion Diffusion
Capillary flow

Solute Solvent Support

Fig. 2.137: Phenomena of transport involved in (left) wet impregnation and (right) dry impregnation.
The solute migrates into the pore from the left to the right of the figures. (Redrawn from [72]).

In dry impregnation pressure-driven capillary flow of the solution inside the empty
pores is described by the Darcy’s law:

KKl, eff
Nl, s = − Cl, s ΔPl (2:39)
ηl

where Kl,eff is the relative permeability of the liquid-phase, η is its viscosity, K the
intrinsic permeability, and Pl is the liquid-phase pressure that is equal to the local
gas pressure minus the capillary pressure PC. Influence of viscosity and concentra-
tion on the diffusion is complex. High viscosity associated with high concentra-
tions prevents the precursor diffusion according to eq. (2.39), while at the same
time a high concentration favors the diffusion of the solute toward the center of
the pellet.
Quantification of the conditions leading to profiling during adsorption and im-
pregnation can be done by considering the characteristic time tC at which the solu-
tion completely fills the pore volume:
 2.3 Preparation of catalytic materials 189

8r2 η
tC = (2:40)
rC γ cos θ

where r is the grain radius, θ is the wetting angle, γ is the liquid-gas interfacial ten-
sion, η is the liquid viscosity, rc is the pore radius. Uniform distribution can be
achieved when the impregnation time is a priori longer than the time tc of capillary
impregnation. Otherwise distribution will be non-uniform.
For wet impregnation the characteristic impregnation time tD is given by:

r2 ð1 + PÞτ
tD = (2:41)
DM ε
where τ is the pore tortuosity, DM is the molecular diffusion coefficient of a compo-
nent in the solution, τ is the porosity and P is the ratio between adsorbed and non-
adsorbed amounts of the deposited compound in the support volume, i.e.,:
a
P= (2:42)
g−a

In eq. (2.42) a is the amount of a compound adsorbed and g is the overall amount of
this compound located in the pores (trapped or adsorbed). When P > 1 the precursor
is mainly adsorbed and the metal deposition can be defined as adsorption, while at
P < 1 the role of adsorption is much less prominent. In practice contribution of ad-
sorption and impregnation is comparable.
Uniform distribution is achieved when the impregnation time is substantially
larger than the characteristic impregnation time tD, which as follows from eq. (2.41)
depends at P << 1 on the grain radius, pore structure and diffusion coefficient.
When P >> 1 the overall process might be limited by adsorption. For strongly ad-
sorbing compounds the characteristic impregnation time could be so high that uni-
form distribution cannot be reached.
Otherwise, for weakly adsorbing compounds long impregnation times will lead
to uniform distribution (Fig. 2.138). Such distribution is needed for slow reactions
occurring in the kinetic regime without diffusional limitations.

Impregnation time increases

Active phase Support

Fig. 2.138: Influence of impregnation time on active phase profile.

190 2 Engineering catalysts

Increase of the precursor concentration in principle increases the penetration

depth and leads to more uniform distribution. This approach is not, however, very
efficient as it gives low P values and requires unnecessary high quantities of the
active compound.
Another option to improve uniformity of the active phase profiling is to dimin-
ish the grain size, which obviously has technical limitations because of other re-
quirements such as pressure drop in the reactor. Uniform precursor distribution is
also achieved when precursors and competitors (HCl or acetic acid) are equally in-
teracting with the surface.
Finally, temperature of impregnation is usually of minor importance regarding
precursor distribution.
Egg-shell distribution is applied for diffusional limited reactions when the inter-
nal part of a grain is not utilized in catalysis. In general, in order to get an egg-shell
distribution, one of the options is to use a precursor, which should be strongly ad-
sorbed during impregnation. Impregnation with a very viscous solution can be rec-
ommended as an alternative technique.
An interesting option is egg-yolk distribution, when the active phase is concen-
trated in the center of the grain. Such distribution is obtained when there is a com-
petitor interacting more strongly with the surface than the precursor. It could be
useful for slow reactions when the particles undergo severe attrition. For example,
synthesis of platinum on alumina catalysts could be considered with hexachloro-
platinic acid as a precursor. In the presence of oxalic or citric acids, which adsorb
stronger than the platinum precursor, adsorption of H2[PtCl6] at the outer layer is
prevented, forcing it to move further towards the grain center.
Finally an egg-white distribution with the location of the active phase away
from the outer surface and the center can be used when the center of the grain is
not used because of diffusional limitations, while the outer surface without the
active phase is needed to protect the catalyst because of attrition or poisoning.
Such distribution could be achieved by, for example, using oxalic acid prior to im-
pregnation and then regulating the concentration of competitor and impregnation
time.
Elimination of the solvent after impregnation is done by drying, or in other
words heating the solvent up to the boiling point under static conditions. Uniformity
of distribution depends not only on impregnation but also on drying. When adsorp-
tion is weak or moderate it is intuitively expected that the metal profile in the pellet
can change during drying. To prevent migration of the weakly interacting precursor,
drying can be done at room temperature. An alternative method is to dry a concen-
trated, viscous solution obtained by addition of cellulose derivatives.
Although drying chemistry is typically not considered in modeling of impregna-
tion, it might be important in influencing the after-drying steps. For example, plati-
num complexes adsorbed on the oxide surface undergo further transformations at
the drying stage. When high loaded [PtCl6]2− on alumina is dried at room
 2.3 Preparation of catalytic materials 191

temperature two chloro-ligands are replaced by the two hydroxyl groups of the sup-
port, changing the complex-support bonding. When dried at higher temperatures
the hydrolyzed platinum complexes could lead to polynuclear hydroxo complexes,
which require reduction at higher temperatures.
Changes in the heating rates also govern drying and can lead to non-uniform dis-
tribution. The slow and fast drying regimes are presented in Fig. 2.139A. Compared
with slow drying, fast drying, usually associated with high heating rate and drying
temperature, tends to block all movements of the solute toward the outside of the
pellet.

Micro-
Adsorption convection Adsorption
Back-diffusion Evaporation
Back-diffusion Evaporation
Convection Convection

Solute Solvent Support

(A)

100
Uniform
DI,i × 1010 (m2 s–1)

Egg-shell
10

Pronounced egg-shell

1
50 100 150 200 250 300
(B) T (°C)

Fig. 2.139: (A) Phenomena of transport involved in (left) slow and (right) fast drying. The solvent
migrates from the left to the right of the figures. Microconvection is not mentioned on the right
side for the sake of clarity. (Redrawn from [73].) (B) Contour map of the final metal profile in the
diffusion coefficient – drying temperature plane. (Reproduced with permission from [73]).

During drying the reactions between the precursor and the support could be non-
uniform because of different solvent (water) amounts and non-isothermicity along
the pore leading to a position-dependent particle size distribution.
It is apparently clear from the discussion above that impregnation is a complex
process depending on a number of interrelated parameters (concentrations, pH, viscos-
ity, presence of counterions and competitors, solvent). As operational parameters, pH
adjustment, introduction of ligands or organic additives and impregnation duration
could be used. In the drying step temperature, heating rate, gas flow and pressure, as
well as drying time could be also altered to influence the resulting supported catalyst.
192 2 Engineering catalysts

Because of the complexity of impregnation, which should be considered to-

gether with drying, only a few studies have been conducted with a detailed engi-
neering analysis. As an example of an engineering approach to study impregnation
and drying, the work on parametric investigation of impregnation and drying of
supported catalysts by Liu et al. [73] should be mentioned. For drying, the mass bal-
ances of the drying medium (air) and the solvent (water) were considered in addi-
tion to metal dissolved in the liquid and adsorbed on the support. Simulations for
the case of drying with uniform initial conditions demonstrated that the effective
diffusion coefficient and the drying temperature have the most significant impact
on the metal redistribution during drying. The most pronounced egg-shell profile
can be observed at the point of the highest drying temperature and the smallest ef-
fective diffusion coefficient (Fig. 2.139B).
The rate of drying also influences the profile of the active phase (Fig. 2.140).
Thus egg-shell profiles can be enhanced by increasing the drying rate.

Increasing rate of drying

Catalyst
particle

Egg-yolk Homogeneous Egg-shell

Fig. 2.140: Influence of the drying rate on the active phase profile. (Redrawn from [74]).

2.3.6.3 Multiple adsorption and impregnation

As the required amount of the active precursor cannot necessarily be dissolved in a
liquid volume used in pore volume impregnation, high metal loadings could not be
always achieved and multiple impregnations are needed.
Let us first analyze the mass balance for impregnation following the treatment
presented by Pakhomov [50]. The initial amount of the active component precursor
G0 can be calculated as:

G0 = V0 c0 (2:43)

where V0 is the initial volume of the solution with concentration c0.

After the impregnation time τ a certain amount of the precursor g is trapped in
the pores of the volume VΣ and the concentration of the precursor in the solution
changes to cτ . Thus:

Gτ = g + ðV0 − VΣ Þcτ (2:44)

 2.3 Preparation of catalytic materials 193

For wet impregnation in the diffusional regime and combining eqns. (2.43) and
(2.44) one gets an expression for g:
n
g = VΣ c0 (2:45)
n+1

where n = V0 =VΣ is the excess of the impregnation solution compared to the pore
volume. For dry impregnation:

g = VΣ c0 (2:46)

The maximal amount of the active compound that could be loaded on the support
during one impregnation is determined by pore volume VΣ and the solubility cs of
the precursor:

g = VΣ cs (2:47)

Multiple impregnations are usually needed when the desired amount of the active
compound is rather high (above 20 wt%).
It should be noted that since the amount of the available pore volume is re-
lated to the volume of the precursor (bulky organic salts or hydrates), this vol-
ume could be already filled after the first impregnation, while the amount of the
active compound is below the requirement. In such cases thermal treatment is
needed between subsequent impregnation steps to form an intermediate (for ex-
ample, oxides from metal salts). Obviously, as multiple impregnations compli-
cate catalyst preparation, the number of impregnations should be as small as
possible.
Volume of either the intermediate compound or the metal after the first impreg-
nation (followed by thermal treatment) is:

g1M g1A M
ΔV = = = g1 α (2:48)
ρ ρ A

ρ is density, A/M is metal content in intermediate, A is metal mass, M is mass of the

intermediate, g1M or g1A is mass of intermediate or metal after first impregnation
and α = M/Ap.
Free volume of pores after the first impregnation is given by

V1 = V0 − ΔV = V0 − g1 α = V0 ð1 − c0 αÞ (2:49)

Free volume after i impregnation for uniform distribution and no pore blockage can
be expressed in the following way

Vi = V0 ð1 − c0 αÞi = V0 − gα (2:50)

The value of α could be easily calculated for the intermediate compounds when
density is known or experimentally determined through estimation of the pore
194 2 Engineering catalysts

volumes after impregnations. When α is known all the required parameters to per-
form multiple impregnations could be calculated, such as the number of impregna-
tions i needed to introduce g grams of metal per gram of support at c0:

gα
lg 1 − V0
i= (2:51)
lgð1 − c0 αÞ

the amount of metal g loaded in i impregnations:

ð1 − αc0 Þi − 1
g = V0 − (2:52)
α
the volume of solution needed for i impregnations:
X
i−1
Vp = V0 ð1 − c0 aÞi − 1 (2:53)
1

and the minimum volume of pores needed to introduce the needed amount of metal g:
gα
V0, min = (2:54)
1 − ð1 − c0 αÞi

2.3.6.4 Industrial implementation of impregnation

Industrial-scale impregnation is done in both common type reactors and in dedi-
cated equipment. The conveyor type of wet impregnation equipment (Fig. 2.141) has
been used for large-scale impregnation. Conceptually, in this method the catalyst is
loaded in buckets that are moving down into the impregnation basin. The impreg-
nation time depends on the conveyor speed and the length of the basin. The con-
centration of the impregnation solution is kept constant.
For dry impregnation the impregnating solution is sprayed over the support
(Fig. 2.142), which is rotating.

(A) (B)

Fig. 2.141: Conveyor wet impregnation. (A), from [75], (B) from [48]. Explanations are given in the
text.
 2.3 Preparation of catalytic materials 195

Impregnating solution

Spray header

Support to be
impregnated

Rotating drum

Fig. 2.142: Incipient wetness impregnation. (Redrawn from [38].)

There are also reactors operating batch-wise, which allow impregnation and
drying and calcination in the same reactor. The reactor with the volume of 0.25–10
m3 can have a capacity of 100–2,000 kg/day, operating with air as a drying and
calcination agent at, respectively, 80–180 °C and up to 600 °C. An example of syn-
thesis of zinc acetate on a carbon support is illustrated in Fig. 2.143. Activated car-
bon is charged through the pipe pos. 1. For a reactor diameter of 2 m, the height of
the support bed is approximately 0.6–0.7 m. Prior to impregnation, the support is
dried by air through the distribution grid 4. The bed is fluidized by adding

Steam air mixture

Support

2 Impregnating solution
6

3
4

Steam 5 Impregnated support

Air

Fig. 2.143: Reactor combining impregnation, drying and calcination [48]. 1, pneumatic tube for
loading the support; 2, heating elements; 3, fluidized bed of the support; 4, distribution grid;
5, discharging outlet; 6, spraying.
196 2 Engineering catalysts

approximately 20–30% more air than the fluidization threshold. After drying of the
support, the solution of zinc acetate is sprayed through distributer (pos. 6).
Intensive mixing guarantees efficient impregnation. Air steam mixture containing
dust is taken from the reactor top. Reactor is heated by steam through spiral ele-
ments pos. 2. After impregnation in the same reactor, the catalyst is matured, dried
in air at 70–110 °C and unloaded through pos. 5.
This reactor operates batch-wise. To enhance throughput, two reactors for im-
pregnation can be employed (Fig. 2.144), one undergoing impregnation while the
support is unloaded from the second one.

Fig. 2.144: An installation with two reactors for impregnation [76].

As an example, for catalyst preparation by impregnation, a supported silver catalyst

for synthesis of formaldehyde by oxidation of methanol will be discussed. The syn-
thesis of formaldehyde can be performed in a fixed-bed reactor with silver catalysts.
The global production of formaldehyde is divided by this process and alternative
oxidative dehydrogenation at a lower temperature on Fe–Mo oxide. More details
are provided in chapter 4. Because the methanol and air mixture is flammable, a
safe process should operate either below or above the flammability limits (6.
7–36.5% methanol in dry air at 100 °C). Oxidative dehydrogenation of methanol on
silver is organized above the higher limit at 600–700 °C, allowing to achieve the
yields of formaldehyde exceeding 85%.
The classical method of catalyst preparation by impregnation developed in the
1950s is still used, although at the beginning of the twenty-first century an
 2.3 Preparation of catalytic materials 197

alternative electrolytic preparation method was commercialized based on the dis-

covery that the crystal growth could be controlled by the creation of a silver-
diamine complex during the electrochemical synthesis step.
In the classical method silver (30–35 wt%) on pumice is prepared by impregna-
tion of pumice by silver nitrate (Fig. 2.145).

HNO3

AgNO3 NO, NO2

1
2 3 8
Catalyst
2
4
5 6 7

Fig. 2.145: Preparation of silver catalyst by impregnation, 1 – grinding, 2 – sieving, 3 and 6 reactors
with impeller and heating jackets. 4 – filter, 5 – dryer, 7 – calcination. (Redrawn from [48].)

The large grains of the support are ground (pos.1) and sieved (pos.2). The smaller and
larger fractions are disposed and returned to grinding, respectively, while the me-
dium fraction (2–5 mm) is introduced in the reactor with a heating jacket (pos. 3).
Nitric acid is added into the reactor 3 at 60–70 °C for 7–8 h order to remove iron con-
taining impurities

Fe2 O3 + 6HNO3 = FeðNO3 Þ3 + 3H2 O

FeO + 6HNO3 = FeðNO3 Þ2 + H2 O

that otherwise lead to cracking and coking.

After filtering out and washing with water the support is dried at 100–110 °C and
then impregnated in another reactor (pos. 6) with a solution of silver nitrate, with si-
multaneous evaporation of water at 100 °C. After impregnation that results in deposi-
tion of silver mainly in the pore mouth and on the external surface, the catalyst is first
thermally treated at 400–500 °C for 1–2 h to remove NOx followed by calcination at
approximately 650 for 2–3 h. The level of NOx emissions is approximately 0.4 ton per 1
ton of catalyst, while an approximate feedstock consumption for production of 1 ton
of silver on pumice catalyst is as follows: pumice – 630 kg, nitric acid (47%) – 800 kg,
hydrochloric acid – 475 kg, silver nitrate – 740 kg. The example above has mainly ed-
ucational purposes as in industrial practice mainly silver gauze’s are applied.

2.3.6.5 Precipitation
This method has two main variants. The first one is used when cheap catalysts
are produced and the aim is to optimize the activity per volume unit. It relies on
198 2 Engineering catalysts

co-precipitation of the support and the active component which gives a material
that is then dried, calcined and if needed reduced. Details of this method are
covered in a section devoted to the precipitation of bulk oxides (Section 2.3.3),
as the CuZnOA2O3 (Fig. 2.95) catalyst described in that section can be also con-
sidered as copper and zinc oxide supported on alumina. Precipitation is caused
by adding a precipitation agent such as sodium hydroxide or bicarbonate.
A short description of the nucleation phenomena was provided in Section 2.3.3.
Thermodynamic analysis of deposition-precipitation should also consider the inter-
facial energy with the support. Because of the presence of such interactions the
total free energy also includes enthalpy of the support-nanoparticles (NP) interface:

ΔGtotal = ΔHNP + ΔH NP − TΔS (2:55)

sup

and can be significantly lowered. This in turn results in a lower concentration of the
active precursor at which the crystallization rate can be measured. Solubility plots in-
cluding the solubility and supersolubility curves are presented in Fig. 2.92. The solubil-
ity curve (saturation concentration) separates liquid from the solid, indicating that
solubility increases with temperature. There is a concentration (supersolubility, or nu-
cleation threshold) where the precipitation becomes abruptly measurable. Between
these two lines of solubility and supersolubility precipitation does not occur within the
bulk and happens exclusively on the support. Keeping the concentration of the catalyt-
ically active precursor in this range is the basis of the deposition-precipitation method.
This second method of precipitation, originally developed for high loading of
the active phase, employs pre-existing supports and is preferred when the active
phase precursors are expensive and deposition of a nm size active phase is needed.
Obviously, precipitation in the bulk should be avoided. As mentioned above, metal
hydroxides or carbonates precipitate on the support after the addition of a base to
increase pH.
In order to achieve homogeneous precipitation, pH should be homogeneous
(difficult to achieve during scaling up) and the concentration of the active precursor
should be below the supersolubility curves.
Such homogeneous solution can be obtained in the case of deposition-
precipitation with urea that slowly hydrolyzes at 70–90 °C according to the follow-
ing reactions:

ðNH2 Þ2 CO = NH4+ + CNO − ;

CNO − + 3H2 O = NH4+ + HCO − + OH − ;

CNO − + 2H2 O = NH4+ + CO2 + 2OH − .

The rate of urea hydrolysis does not depend on pH as it is of first-order towards

urea. The amount of urea is typically 1.5–2.5 times above stoichiometric because of
ammonia volatilization.
 2.3 Preparation of catalytic materials 199

It should be also mentioned that although urea hydrolysis takes a relatively

long time there is a possibility to thoroughly mix the support suspension at a tem-
perature where there is no reaction. This allows for preparing highly concentrated
solutions. Large amounts of highly viscous suspensions of support can therefore
be used with urea, separating the mixing of the ingredients for catalyst production
and the generation of precipitating agents. This is especially important for the
scaling up of catalyst preparation as efficient stirring is difficult to achieve in
large vessels.
Deposition-precipitation with urea gives a constant level of supersaturation by
maintaining a uniform concentration of OH– in both the bulk and pore solutions.
The precipitated phase is a hydroxide of the active metal. Supported Mb, V, Fe, Ni,
Cu and gold catalysts could be prepared by deposition-precipitation with urea.
The drawbacks of the method are the possible formations of cyanate and nitrite
in the waste water and release of NO.
Obviously, properties of the support are important in the deposition-precipitation
process. Thus, large particles could be formed in the case of neutral support surfaces
close to the support pzc, neutral metal complexes, lack of hydroxyl groups on the
support surface at low pH or dissolution of the support at low pH. In addition, the
following lead to large particles: the repulsions occurring between the anion and
the negatively charged, deprotonated support surface at pH above the pzc of the sup-
port, the presence of chloride causing agglomeration of metal particles, low stability
and slow adsorption rates of some metal complexes, such as Al(OH)4–.
Besides the strength of interactions between the metal precursor and the sup-
port, maturation time and redispersion also have an impact on the final nanoparti-
cle size.
Changing the pH level of a solution of an active precursor, changing the va-
lency of a dissolved active precursor, or decreasing the concentration of a com-
pound forming soluble complexes with the active precursor are procedures that can
be used in deposition-precipitation. Many deposition-precipitation procedures in-
volve an increase in the pH level of a suspension of the support in a solution of the
active precursor. As a significant number of catalytically important metal com-
pounds dissolve as oxy-anions, raising the pH level cannot be used to bring about
the precipitation of such metal ions, and consequently a procedure has been devel-
oped to reduce the metal in the oxy-anions to a lower valency. The metal ions of a
lower valency exhibit a basic behavior and are thus soluble at low pH levels and
precipitate at high pH levels.
For the supports with high pzc the use of the deposition-precipitation (DP)
method is challenging. The DP method is also not very easy to perform for hydropho-
bic carbon. As the nature of adsorption sites for oxides and for carbon is different,
the same method of pH adjustment cannot always be applied. Zero potential of silica
does not change when increasing the pH from 3 to 7, whereas these changes are sig-
nificant for carbon in the range of pH 2–5. Carbon surface is thus oxidized to
200 2 Engineering catalysts

efficiently increase the number of carboxylic groups and subsequently the loading. A
particular case is preparation of carbon supported gold catalysts, which could be eas-
ily reduced during deposition, leading to large clusters of gold. In order to avoid it,
gold colloidal sols are prepared with polyvinyl alcohol (PVA) followed by a reduction
with NaBH4, drying, and finally deposition onto the carbon support.
A simple thermodynamic model [77] that takes into account the number of hy-
droxyl groups on oxide surfaces during the preparation of supported catalyst par-
ticles by deposition-precipitation with urea assumes that the interfacial energy can
be represented by two terms:

ðpH − pzcÞ
γss = γ0SS + γelectrostatic
SS = γ0SS + γ0SS λ (2:56)
pzc

one γ0SS describing pH independent non-electrostatic interfacial energy, while the

other γelectrostatic
SS denotes electrostatic energy between a nanocluster and the charged
surface.
The latter is proportional (with λ being the cluster size independent propor-
tionality constant) to the relative approach of pH to the point zero charge
Δz = ðpH − pzcÞ=pzc. The latter parameter could be either positive or negative (for
example, for supports with high pzc, such as magnesia). This simple approach al-
lows for predicting dependence of the cluster size on the pH at which precipita-
tion was performed. The enthalpy for nanoparticles is a function of the number of
nanoparticles multiplied by the surface free energy and the atomic volume of bulk
metal. The number of nanoparticles for a hemispherical cluster can be written as
a cubical equation with respect to the particle radius r. The minimal total energy
with respect to this radius can be derived by taking into account the above men-
ΔGtotal
tioned correlations via taking a derivative = 0 giving finally:
dr
p2 p2
r3 − r− =0 (2:57)
Δz Δz

With:

ðβRT − γ0SS Þ 4αγΩ

p1 = 3 , p2 = − 2 0 (2:58)
γ0SS πλ γSS πλ

where α is a dimensionless coefficient of an order of unity (α can be equal to 1 only

for spherical particles), Ω is the atomic volume of the bulk metal, R is the gas con-
stant, β is a dimensionless constant (β≈ 5–8). Despite the apparent simplicity of
eq. (2.57), modeling of the cluster size dependence as a function of Δz (Fig. 2.146)
shows clearly that the proposed model can be used as a tool for the estimation of
the metal particle size on a given support surface.
 2.3 Preparation of catalytic materials 201

1.8
1.7
Cluster radius (nm)

1.6
1.5
1.4
1.3
1.2
1.1
1.0
0.0 0.1 0.2
Δz

Fig. 2.146: Dependence of cluster radius on Δz for gold catalysts on titania prepared by deposition-
precipitation with urea at different pH. (Redrawn from [77].)

2.3.6.6 Atomic layer deposition

Dry methods of catalyst preparation, such as chemical vapour deposition (CVD)
and atomic layer deposition (ALD), being inherently more expensive than wet meth-
ods, have their advantages resulting in catalysts with controlled metal dispersion.
In fact the dispersion could be very high at significant loadings. In CVD, volatile
precursors are deposited in a “cold zone” directly from the gas-phase and thin films
are formed on the support surface by thermal decomposition or reactions of gas
compounds.
ALD uses alternating cycles of self-limiting reactions and deposition occurs in a
layer-by-layer fashion. Sequential exposure of the support with surface hydroxyl
groups to, for example, trimethyl aluminum Al(CH3)3 and water, results first in the
formation of O-Al linkages and the generation of methane in the gas-phase
(Fig. 2.147). When no surface hydroxyl groups are available the reaction terminates.
The surface methyl groups can react, however, with water producing methane and
another set of O-Al linkages.
Typically a constant stream of an inert gas (N2) is flowing through a chamber at
a pressure of ~0.1–10 Torr and the precursor vapors are injected into this carrier
gas. A purge period is introduced between each precursor dose to avoid mixing of
the chemicals, which can give non-self-limited growth. The film thickness is very
precisely controlled by the number of precursor cycles rather than the deposition
time, as is the case for conventional CVD processes.
For growing thin films, reaction cycles with the metal and the non-metal reactants
require just seconds being performed at the same temperature. For porous materials,
longer reactant exposures (hours) are needed. Moreover, longer times allow different
reaction temperatures to be used for the two reactants. Organometallic compounds
and metal halides are most commonly applied as metal reactants. The deposition of
noble metals by ALD requires a long time because of the difficulties in removing the
202 2 Engineering catalysts

Gas in
A CH4 Coil
CH3 H3C CH3 power
Al(CH3)3 Al Al
OH OH OH O O O
Inert gas
Substrate Substrate valve
Precursor
in
B CH4
CH3 H3C CH3 OH HO OH
Al Al H2O Al Al
O O O O O O

Substrate Substrate
Table
(A) (B) Pumps power

Fig. 2.147: Illustration of surface chemistry for alumina ALD. (A) (From [78]), (B) Figure courtesy of [79].

ligands from the noble metal precursor bound to an oxide surface. Poor nucleation and
the ability of noble metal atoms to agglomerate results in the formation of highly dis-
persed metal clusters of nanometer size.
Desirable characteristics for ALD precursors include high volatility, good ther-
mal stability at the ALD growth temperatures, and high reactivity with the other
compound used for the film growth.

2.3.7 Catalyst forming operations

Several post-treatment and forming operations, which are performed during cata-
lyst preparation, were briefly discussed in Section 2.3.1.2. Catalyst forming, which is
seldom considered in academic research, is, however, extremely important for the
preparation of catalysts influencing the final performance. Information about cata-
lyst forming is typically proprietary and primarily empirical. Only a few reviews on
catalyst forming are available, such as Schüth and Hesse [80]. At the same time, it
is apparently clear that solid engineering approaches are needed in this field, other-
wise unit operations in catalyst preparation will resemble those used centuries ago
(Fig. 2.148).
Spray-drying, extrusion, pelleting and granulation are the most important cata-
lyst forming methods to give different catalyst shapes, which define the pressure
drop in a reactor. The properties of a primary solid in terms of activity and selectiv-
ity should be maintained. In addition to the primary porosity, compaction creates
additional porosity that is important for mass transfer. Porogenes are added during
forming, and are burned when the catalyst is calcined. These porogens that help to
 2.3 Preparation of catalytic materials 203

Fig. 2.148: Unit operations. (A) Georg Bauer (Agricola), “De Re Metallica” (1566). (B, C) Catalyst
preparation operations. Photo courtesy of Dr. I. Simakova.

improve porosity could have adverse effects. The distribution of particle size in the
powder that is used for compaction can be beneficial, as the fines fill the voids and,
after agglomeration, improve the strength of the large particles. Mixing the particles
with different sizes could be challenging as they tend to be accumulated in different
parts of a reactor (smaller at the top). Moreover, in some cases excessive fine pow-
der leads to pilling problems.
Forces of varying strength ranging from weak and medium (van der Waals,
electrostatic, capillary forces) to strong bonds (covalent, siloxane bridges) are in-
volved in the agglomeration of solid particles. Even the weak van der Waals forces
of very small particles could be of importance when dispersion forces are prominent
for many points of contacts between particles.

2.3.7.1 Spray-drying
Spray drying is a method for producing attrition-resistant microspherical catalyst
particles. In the spray dryer, the slurry comprising the catalyst particles, suspended
in an aqueous sol or hydrogel of a binder (e.g., silica–alumina, alumina, clays), is
sprayed into a hot gas flow. The binders are used if shaping is otherwise not possi-
ble in the spray-drying process.
Spray-drying can be used to generate the catalysts in the final shape or
make powders that will be further processed with another method. Spray-drying
process results in 10–100 μm of almost identical shape particles, which could be
used in processes operating in fluidized bed reactors, such as fluidized catalytic
cracking.
During spray-drying a powder slurry in water suspension (10–30 wt%), prepared
by milling (for example, the filter cake to the size of several hundred nanometers), is
fed through a nozzle that sprays small droplets into hot air. The formation of micro-
spheres occurs when the spray falls downward in the dryer, the binder particles
204 2 Engineering catalysts

undergo gelation and the solvent evaporates. This results in embedding of the cata-
lyst particles into a matrix of cross-linked particles.
In high gas temperatures calcination can even occur. This process can be re-
peated, that is, spray-dried particles can be processed the second time to furnish
them with a protective shell.
The nozzle (or atomizer) generating the spray is very important in spray-drying,
defining the diameter and size distribution of the droplets. The choice of the spray
nozzle construction material is important, especially in the presence of abrasive
components in the slurry.
Rotary atomizers with a hollow wheel rotate at high speed (up to 20,000
revolutions min−1) to spray the slurry mechanically which leads to fine to medium
coarse powders. Another parameter is the flow direction, which could be cocurrent,
counter-current, or a mixed flow. The air distribution is needed to ensure homoge-
neous flow across the chamber. In a cocurrent operation mode the hottest air con-
tacts the droplets at a stage when the majority of the water is still present, giving
the most desired profile.
The particle size distribution is adjusted through control of the atomizer wheel
speed. Coarse grained powders can be produced using a low speed atomizer wheel
operation.
For conditions when the inlet drying temperature is restricted to below 550 °C,
roof air dispersers are used (Fig. 2.149). When high outlet temperatures are needed,
part of the exhaust air can be recycled to preheat the drying air passing to the dry-
ing chamber, improving the overall heat efficiency.
Catalyst formulations containing organic solvents are dried in an inert gas
closed cycle plants.

Feed

Drying air

Product

Fig. 2.149: Spray dryer with a rotary atomizer and a roof air disperser. (From [81].)
 2.3 Preparation of catalytic materials 205

There are several advantages of rotary atomizers, such as rather narrow par-
ticle size distributions, requirements of high-pressure pumps only when very
thick pastes are sprayed, resistance to plugging, ability to handle relatively abra-
sive slurries, and high capacities. The design is, however, more mechanically de-
manding and the size of the catalyst particles is tightly bound to the diameter of
the drying chamber. Dependence of the mean particle size as a function of the
chamber diameter is discussed by Roberie et al. [82] and demonstrates that,
for example, FCC catalysts (65–80 μm) require a particular chamber diameter
(2.8–4 m).

2.3.7.2 Size reduction and granulation

Both types of operation (size reduction and size enlargement) are used in catalyst
preparation.
Ball mills and agitated media mills can be used to produce powders of very
high fineness from dry or wetted feedstock even down to the submicron range. For
wet grinding, the powder is suspended in a liquid upstream milling and then fed in
such form to the mill. Ball mills (Fig. 2.150) are equipped with grinding media. The
bed of balls or grinding beads is put into motion by rotation of the grinding cham-
ber or by an agitator. The feed material is comminuted between the grinding media
because of impact and shear forces. Ball milling can be organized in a continuous

Dust collection
Feed material

Feed material

Grinding balls

Adjustable
discharge slots Finishaed product

Fig. 2.150: A scheme of ball milling [83].

206 2 Engineering catalysts

mode, when the product is pumped to the mill, flows through the agitated bed of
grinding beads and finally exits the mill through a suitable discharge. The volume
of the grinding beads can be approximately 70–80% of the mill volume. The size of
the mills used in catalyst manufacturing is approximately 1–3 m3.
Size enlargement by granulation, which is conceptually similar to a snowball
formation, could be used for large-scale production of support spheres, giving par-
ticles of irregular shape and size. Moreover, particles are less dense and strong com-
pared to tableting and extrusion. During granulation, size enlargement occurs by
wet tumbling-growth agglomeration. Adhesion between wetted particles originates
from capillary forces or can be additionally stabilized by cross-linking if the sprayed
liquid contains binders, which can either undergo hydrolysis and condensation re-
actions or gelation. Better mechanical strength provided by the inorganic binders is
achieved at the expense of the pore size. Alternatively, application of organic bind-
ers necessitates their removal by calcination, leading thereby to lower mechanical
strength.
Granulation can be done by placing the powder, which could be pre-wetted, on
a tilted (45–55° against the floor) rotating pan (Fig. 2.151). The particle growth and
agglomeration occur while the pan is rotating. After reaching a certain size the par-
ticles fall over the rim, which could be higher than 1 m for pans with large diame-
ters of several meters. In addition to growth, several other processes proceed during
granulation, such as wetting and nucleation prior to agglomeration; and also attri-
tion and breakage, which decrease the size of the formed granules. Rotational
speed is adjusted to be 25–40% less than the critical rotation speed at which the
centrifugal force is compensating for the weight. A granulation pan with 3 m diame-
ter tilted to 50° operates between 12 and 15 rpm.
Another method of granulation is to use a cylindrical tilted rotated drum (diam-
eter 0.5–1 m, length up to 3 m) with the granulation mass fed on one side and then
transported through the drum. This mode of operation, where tilting is used solely
for transportation, gives a broader granule size distribution than is obtained with a
granulation pan, requiring downstream granulation.

2.3.7.3 Extrusion
Compared to other preparation methods, the extrusion process affords high
throughput at relatively low costs and gives a variety of possible extrudate shapes.
The downside of the method is a non-uniform shape of extrudates and lower abra-
sion resistance compared to pellets. Moreover, not all catalysts even in the presence
of binders can be shaped by extrusion.
Extrusion of catalysts and supports is different from extrusion of polymers,
when properties of a homogeneous polymer melt can be regulated by the extruder
temperature. Because pastes for catalyst extrusion are highly concentrated disper-
sions, their behavior is determined by rheological characteristics. The pore
 2.3 Preparation of catalytic materials 207

Fig. 2.151: A granulation pan.

structure and mechanical stability of extrudates are determined by the properties of

the paste and extrusion conditions. For catalytic applications, a certain pore struc-
ture is needed to allow transport of the reactants to the active sites. Typically, extru-
dates contain large transport pores of 300–600 nm in addition to mesopores (10–25
nm) being different from materials prepared by tabletting as the latter have mainly
monomodal distribution of mesopores.
During extrusion a wet paste from a hopper at the top is forced through a die, and
the emerging ribbon passing through holes in the die plate is cut to the desired length
using a suitable device. The pressure that is developed in the screw extruder as the
paste moves toward the die is affected by the screw geometry and the paste rheology.
After extrusion, the formed extrudates are either broken because of gravity giv-
ing a non-uniform sized material or cut-off, for example, by a rotating blade. The
cut-off leads to very similar extrudates.
The effect of the extrusion pressure on the compaction in extrusion is, however,
smaller than in pelletizing, because in the former case the particles are suspended
in a liquid. Adjustment of the liquid-to-solid ratio in the extrusion paste is thus
208 2 Engineering catalysts

used not only to influence the apparent viscosity of the paste, which obviously de-
creases when there is more liquid in the paste, but also to vary the porosity of extru-
dates. On the other hand, too high liquid content will prevent the formation of
extrudates after leaving the die. Even if plasticity can be improved by addition of
plasticizers, flow of suspensions is characterized by strong non-Newtonian flow.
While understanding of suspension flow is not yet sufficient to predict extrusion
and control the final properties of catalyst extrudates; nevertheless, rheology is im-
portant in analyzing extrusion processes and will be briefly discussed below.
Flow of suspensions can be described by a single property, namely, the shear
viscosity. Shear is an action of stress when forces are applied causing two contigu-
ous parts of a body to slide relative to each other in a direction parallel to their
plane of contact. Figure 2.152 illustrates deformations produced between parallel
plates when the upper plate of surface area A is moved in response to force F. The
displacement of a plane layer (dx) over the separation between layers (dy) is termed
the shear (dx/dy) acting on the fluid.

Moving plate

Fluid

dx

dy

Stationaty plate

Fig. 2.152: Illustration of deformations between parallel plates.

Depending on the material, there are differences in the behavior when the shear
stress is removed. While a solid will tend to return to its original shape, a plastic
material may show only partial recovery. The force per area in the plane termed the
shear stress (τ) is expressed as

τ = F=A (2:59)

According to the Newton’s law of viscosity, the shear stress is proportional to the
.
derivative dV/dy termed the shear rate γ:
 2.3 Preparation of catalytic materials 209

.
F=A ∝ dV=dy ∝ γ (2:60)

giving dependence of shear stress on shear rate

.
τ=ηγ (2:61)

where η is the coefficient of viscosity having units of mass·length−1·time−1. In SI

.
units, τ is in Pa, γ in s−1; thus, η is expressed in Pa·s. The coefficient of viscosity,
being the ratio of the shear stress to the shear rate, represents the internal resis-
tance to flow of the fluid. Equation (2.61) is applied to Newtonian fluids when vis-
cosity (temperature dependent) does not change with the relative flow velocity. For
the large class of fluids relevant for chemical product technology, this does not
hold and viscosity changes depending on the shear rate
. .
τ = ηðγÞ γ (2:62)

The rheograms for non-Newtonian fluid can often be described by an empirical

power-law function called the Ostwald-de Waele equation
 n
dγ
τ=k (2:63)
dt

where k is the consistency index. Materials such as concentrated dispersions used

in preparation of catalysts often exhibit the so-called shear thinning flow, which is
distinctly different from the Newtonian behavior (Fig. 2.153). For shear thinning
(pseudoplastic) n < 1, while for the Newtonian flow n = 1. Shear thickening (dilatant)
fluid has n exceeding unity (n > 1).
Figure 2.153A illustrates a parabolic profile for the Newtonian liquid. Deviations
for shear thinning liquids from the ideal behavior are attributed to a low viscosity
near the wall, where the shear rate is high and a high viscosity in the middle. The
velocity profile is more flat resembling a plug flow. The opposite behavior with a
sharp profile is observed for shear thickening fluids. Dependence of viscosity on the
shear rate is shown in Fig. 2.153B. Shear viscosity decreases or increases with increas-
ing shear rate, leading to, respectively, shear thinning and shear thickening.
The entire flow curve for pseudoplastic behavior can often be fitted using the
Reiner–Philippoff equation:
 . 
. . η0 − η∞
τ = − γ η∞ + (2:64)
1 + τ2 =A
. .
where A is an adjustable parameter, η0 is viscosity at γ →0, η∞ is viscosity at γ →∞.
An often used model of Herschel–Bulkley is the power-law type of equation
with the addition of a yield stress
210 2 Engineering catalysts

Bingham plastic
Dilatant
Plug flow profile
Shear thinning n<1

Visoosity
Parabolic profile
Newtonian n=1

Newtonian
Sharp profile
Shear thickening n>1
Pseudoplastic

Shear rate

Fig. 2.153: Behavior of suspensions. (A) Flow profiles. Coefficient n appears in eq. (2.63) and
(B) typical curves of viscosity versus shear rate [84].

.
τ = τ0 + kðγÞn (2:65)
.
where τ, τ0 and γ are shear stress, shear yield stress and shear rate; k is the consis-
tency constant and n is the flow index. In this model, viscosity is expressed as
τ0 . n−1
η= . + kjγj (2:66)
jγj

Binders can also influence the flow behavior, altering the velocity profile in the
paste, thereby deteriorating extrudated properties due to inadequate adhesion
during extrusion. The desired product during extrusion (high shear rates) is
achieved for pseudoplastic pastes with low viscosity, since after leaving the die
at a low shear rate the viscosity is increased leading to stable extrudates. Most
catalyst pastes are, however, prone to shear deformation. The flow behavior of
molding masses is well described by Ostwald–de Waele equation (eq. (2.63)).
The flow behavior index n in eq. (2.63) indicates pseudoplastic fluids being for
simple shapes at most 0.7, while for preparation of honeycomb monoliths it is
below 0.3.
Because extrusion of catalysts and supports is in essence defined by the flow of a
suspension inside an extruder, the ability to make an extrudate is determined by the
geometry of the extruder, chemical properties of the screw, wall and die materials,
which should ensure minimal adhesion of the paste, most importantly by the proper-
ties of the paste.
The value of the flow behavior index is related to the velocity profile. There is a
radial profile of flow velocity in the die channel as the flow velocity decreases with
an increasing distance from the flow axis. If the velocity gradient is significant, the
defects will be generated after a granule leaves a die channel because of the
 2.3 Preparation of catalytic materials 211

velocity equalization. Subsequently, efficient molding will not be possible. A small

flow behavior index corresponds to less significant velocity gradients and an ideal
plug flow is achieved when the index is approaching zero.
Molding pastes have non-Newtonian flow behavior and during their movement
through dye channels, different velocity profiles are observed (Fig. 2.154), determin-
ing the tensile strength of shaped catalysts. Obviously, in case of substantial velocity
profiles in the radial direction, produced grains will have macrodefects, originating
from improper “layering” of the material.

Fig. 2.154: Flow deformation profile during extrusion (From [85]).

Pastes that are too viscous can block the extruder, while the opposite leads to un-
stable extrudates. Therefore, an important characteristic of suspensions used in cat-
alyst extrusion is viscosity, which can be expressed by the Einstein equation as a
function of the solid content at low concentrations:

ηr = 1 + 2.5’ (2:67)

where φ is the solid fraction. At higher concentration, the data for relative viscosity
appear to follow more of an exponential function of concentration according to the
Mooney equation:
 
ξϕ
ηr = exp (2:68)
1 − kϕ

where ζ is a coefficient determined by fitting, constant k is the self-crowding factor,

being in the range between 1.35 and 1.91. The self-crowding factor is inversely pro-
portional to the maximum concentration. The latter has a limit because of the expo-
nential character of the curve. The self-crowding factor at maximum concentration
does not allow the movement of neighboring particles as the suspension acts like a
solid.
Krieger–Dougherty semi-empirical equation describes the concentration depen-
dence of the viscosity:
  − K’max
ϕ
ηr = 1 − (2:69)
ϕmax

where φmax is the maximum packing fraction or the volume fraction at which the
zero shear viscosity diverges. Parameter K is equal to 2.5 for spherical particles,
212 2 Engineering catalysts

while deviation from this value is related to shape changes of the particles.
Equation (2.69) reduces to the Einstein relation at low particle concentration.
An alternative approach of Brodnyan–Mooney was developed for ellipsoidal
particles
" #
2.5ϕ + 0.399ðp − 1Þ1.48 ϕ
ηr = ηo exp (2:70)
1 − kϕ

where k is the self-crowding factor.

The treatment above considered suspensions of hard spheres without colloidal or
thermodynamic interactions. Repulsive surface forces stabilize dispersions and pre-
vent aggregation, as the particles are kept away at a certain distance. Electrostatic or
steric repulsion results thus in an excluded volume inaccessible to other particles,
which can be expressed as follows:
.
aeff 3
’eff = ’ (2:71)
a
where aeff is the effective particle radius, which is defined as half the distance to
which centers of two particles can approach each other under colloidal forces.
Rheology of colloids in many cases is similar to hard sphere dispersions being
quantitatively described as a hard sphere system with the concept of effective ra-
dius. Equation (2.71) implies that particle size is a parameter that influences φeff. An
increase in the particle radius a at constant φ and a constant range of the repulsive
colloidal interactions corresponds to a decrease in φeff. This means that dispersions
with the same φ but different a may exist in different phases having a strong impact
on the shear rate-dependent viscosity.
Not only the size but also the particle size distribution can influence the
phase behavior of colloidal dispersions. A significant viscosity decrease is typi-
cally observed by mixing particles with different sizes when the particle volume
fraction φ exceeds 0.5. This effect is more pronounced with increasing φ and the
ratio between large and small particles χ. When repulsive colloidal interactions
become relevant, the viscosity passes through a minimum increasing again
when χ is increased at a constant total particle concentration and a fixed fraction
of small particles.
Attractive particle interactions can lead to large compact aggregates and rapid
phase separation. Alternatively, loose aggregates attract water and increase an ef-
fective volume fraction φeff. This results in an increase in the zero-shear viscosity.
Upon an increase in the shear rate, there is a gradual floc breakdown and/or their
alignment in the flow direction, and as a result a decrease in viscosity.
The extrusion process is influenced by the paste composition: water content
(typically between 20 and 40 wt%); presence of peptizing agents (to adjust the col-
loid-chemical properties of the paste); binders; plasticizers; lubricants and
 2.3 Preparation of catalytic materials 213

porogenes. It is also important that no large solid particles are present for the extru-
sion of complex shapes. The operating variables also include mixing time, aging
and extrusion temperature.
While elevation of temperature is positive, improving relative deformations
from the viewpoint of technology, the process becomes much more energy intensive
and unnecessarily complex. Moreover, temperature fluctuations can influence me-
chanical and rheological properties of the suspension, primarily the optimal water
content. Therefore, extrusion of highly concentrated suspensions is carried out at
ambient temperature.
The paste should be plastic enough to be able to pass through the die and ob-
tain the desired shape. After extrusion, the extrudate should not have any visible or
microdefects having adequate properties for further processing (transport, drying,
calcination, etc.). The coagulation structure of the paste, on the one hand, should
be strong enough ensuring adequate rheological properties and retention of the
shape after leaving the die, and on the other hand, too strong coagulation structure
implies higher energy needed for extrusion.
As already mentioned an important parameter influencing properties of the ex-
trudates is the size of the particles in the suspension, which should be small
enough enhancing plasticity. An energy-intensive operation, ball milling, is typi-
cally applied to reduce the size of the solid phase.
When the amount of water is much lower than the solid pore volume, it might
be difficult to process efficiently such dry pastes. Settling of the pastes after
extruder can be, on the contrary, troublesome for a significant water content
(Fig. 2.155). The amount of water depends on the surface lyophilic and adsorption
properties of the solid phase and its specific surface area. For example, the opti-
mum amount of water in mass% for an alumina support using 20% nitric acid solu-
tion as a peptizing agent is 22–23%, while somewhat higher amount of water
(28–29 mass%) is required when PVA is used instead of nitric acid. A catalyst for

Fig. 2.155: Extrusion of too wet paste (courtesy of Dr. Z. Vajglova).

214 2 Engineering catalysts

oxidation of SO2 to SO3 based on vanadia on silica when processed with sulfuric
acid as a peptizing agent needs as high as 37–38 mass% of water.
Porogenes (carbon black, starch and sawdust) are added to the paste to create
porosity, which is generated during calcination. Besides porogenes, plasticizers and
lubricants are also burned away during calcination, contributing to porosity genera-
tion. Such porosity while beneficial for mass transfer can deteriorate mechanical
stability.
Modification of shaping masses, when the particle sizes are in the (sub)microm-
eter range, from the colloidal viewpoint is controlled by zeta-potential. These sys-
tems that are used in extrusion could be modified, for instance, by the addition of
peptizing agents (e.g., nitric, formic or acetic acids) influencing colloid–chemical
interactions between particles and assuring deviation of pH from the point zero
charge. Otherwise , the liquid phase might start to agglomerate. Corrosivity of pep-
tizing agents (e.g., nitric acid) should be considered when selecting them. In addi-
tion, utilization of nitric acid during extrusion might influence acidity and can give
NOx during calcination. To mitigate these emissions, a selective catalytic reduction
of NOx should be installed, obviously increasing the catalyst production costs. As a
consequence, application of electrolytes as peptizing agents is justified when other
methods to regulate rheological properties of suspension pastes are not successful.
Plasticizers or their mixtures applied to improve the paste rheological behavior
are selected from a wide variety of plasticizers, including clays, starch, sugars, cel-
lulose derivatives, polyethylene glycol and PVA. The amount of plasticizers should
not be too high (up to 3–5 mass%), corresponding to covering the available surface
of the solid and allowing the desired changes in rheological properties. Higher con-
centration might not be beneficial from the rheological viewpoint. Selection of rhe-
ology improvers should also consider potential adhesion of the suspension paste to
the extruder material, which should be minimized.
Formation of shaped catalysts implies cross-linking between particles in the
suspension, which can be achieved either by condensation of surface hydroxyls, re-
sulting in oxygen bridges, or by sintering. In the absence of suitable surface hy-
droxyl groups in many catalysts or catalyst precursors, addition of inorganic
binders allows interparticle cross-linking. Binders, such as hydroxide forms of alu-
mina, silica–alumina, clays or silica sols applied together with the catalytic phase
result in cross-linking at sufficiently low temperatures avoiding thereby damage of
the catalytically active material. Binders required during extrusion make the extru-
dates strong enough, preventing a potential collapse.
Typically, inorganic binders are utilized because organic ones will be burned
away at the calcination step at 500–600 °C, which also results in the burnout of
organic additives like plasticizers.
In many cases, the catalytic pastes cannot be shaped without adding binders.
When alumina is added in the form of boehmite or pseudoboehmite, it is trans-
formed into transition alumina during calcination. Gelation of silica sols and
 2.3 Preparation of catalytic materials 215

agglomeration of delaminated fragments of clays are responsible for the stability of

extrudates when these binders are used.
Introduction of binders can influence catalytic activity sometimes in a negative
way. In such a case, a compromise should be made between the rheological proper-
ties and mechanical stability on one hand, and catalytic behavior, on the other. For
example, introduction of aluminosilicates for extrusion of monolithic blocks based
on titania improves the rheological properties of the suspension paste at the ex-
pense of the catalytic properties.
A special care should be taken on the surface properties of binders, which can
be themselves catalytically active. In production of traditional ceramics, this is a
rare case. In production of catalysts by extrusion, there are, however, several ex-
amples when the main active component and the binder influence each other,
which has a consequence in terms of physicochemical properties of the final cata-
lyst. Figure 2.156 shows temperature-programmed desorption of ammonia from
fresh forms of H-ZSM-5 zeolite and the material with aluminum phosphate as a
AlPO binder.

Pure zeolite powder

Phosphate-bound zeolite
NH3 des. (a.u.)

0 100 200 300 400 500

Temperature/ oC

Figure 2.156: Ammonia TPD of H-ZSM-5 powder after calcination and AlPO4-bound extrudate
containing the same zeolite. Reproduced with permission from [86].

There are evident changes in acidity after addition of a binder resulting in another
distribution of sites in terms of acidic strength. Such an example of the binder influ-
ences on the acidic strength, and thus catalytic properties can be interpreted by the
phosphor migration from the binder to the zeolite crystal structure.
Application of binders can result in nonuniformity of the active component dis-
tribution. It implies that there could be zones with a high concentration of the
216 2 Engineering catalysts

active phase, and consequently, higher rate, maybe local overheating and appear-
ance of zones which are controlled by mass transfer rather than kinetics.
Special shapes (trilobates, rings, hollow cylinders, monoliths or honeycombs)
can be obtained using proper dies (Fig. 2.157).

Fig. 2.157: Shapes of extruded catalysts.

Internal cavities can be formed when, for instance, a two-piece aluminum extrusion
die set is used. In a set with separated parts, the piece at the right of Fig. 2.158 is
used for forming the internal cavity.

Fig. 2.158: A two-piece aluminum extrusion die set to form extrudates with an internal cavity [87].

The quality of the extrudates depends not only on extrusion but also on down-
stream drying and calcination. These steps require special attention in the case of
larger structures such as extruded monoliths, which are applied in SCR of NOx at
power and waste incineration plants or the destruction of dioxins generated during
incinerating municipal wastes. The active phase V2O5 is deposited on titania (ana-
tase) support. To improve the mechanical strength, silicoaluminates or glass fibers
are added as mechanical promoters (3.5 wt%). The suspension for extrusion con-
tains clays (ca 6.5 wt%), water and small amounts of methyl hydroxyethyl cellulose
and polyethylene glycol as lubricants. Drying of the extruded monolith must be
slow enough to prevent ruptures and cracks. Calcination is performed at approxi-
mately 500 ± 600 °C.
 2.3 Preparation of catalytic materials 217

Screw extruders with short screw length to barrel diameter are used for cata-
lysts requiring less viscous pastes. For more viscous pastes, press extruders with a
rotating pressing cylinder are utilized due to shorter distance at which the pressing
force is applied. Figure 2.159 shows that this distance is just between two cylinders,
the pressing and the screening ones. Because of lower shear deformation, the flow
behavior of the paste is less critical than in screw extruders. The latter allow, how-
ever, production of extrudates with more complex shapes.

Pressing
cylinder

Screening cylinder Fig. 2.159: Extruder press.

As an example of mechanical mixing and extrusion during catalyst preparation, the

synthesis of sulfuric acid catalyst of the SVD type (on diatomite) is illustrated in
Fig. 2.160.

Diatomite

2
KHSO4 K2SO4
H2SO4
5
Hot air 7
V2O5 3 9
Plasticizer
4
Flue gases

10

Fig. 2.160: Production of sulfuric acid catalyst. (Redrawn from [48].) 1, dryer; 2, bunker; 3, grinder;
4, vibration milling; 5, ball milling; 6, extruder; 7, reactor for potassium bisulfate; 8, dryer; 9,
calcination; 10, sieve.
218 2 Engineering catalysts

The active component in this catalyst is vanadium pentoxide promoted with

sulfates of alkali metals, potassium and sodium. The support is diatomite, which is
silica with some minor presence of other oxides such as Fe2O3 and Al2O3. The cata-
lyst performance depends very much on the type of diatomite.
The support is dried in a spray dryer at 120–150 °C up to a humidity of 5–7%. The
technical quality vanadium pentoxide (80% of V2O5) is ground up, first to 2–3 mm
(pos. 3) and then to ~200 μm (pos. 4). After mixing and ball milling (pos. 5) the solid
phase, containing diatomite and vanadium pentoxide, is extruded with the addition
of potassium hydrosulfate (promoter), stearic acid (plasticizer) and graphite (poro-
genes) directly to the extruder. Potassium bisulfate is prepared in the reactor (pos. 7)
by contacting potassium sulfate with sulfuric acid. The extrudates contain 26–30%
water and are thus dried at 120 °C (pos. 8) and calcined (pos. 9). During calcination
the following reactions occur

2KHSO4 = K2 S2 O7 + H2 O; K2 S2 O7 + V2 O5 = K2 S2 O7 · V2 O5

In this example the supported catalyst was produced by a dry method of mechani-
cal mixing with subsequent calcination. This procedure leads to a catalyst when va-
nadium is present as a complex sulfated salt mixture, which is in the molten state
during actual catalysis.

2.3.7.4 Tableting
Pellets or tablets of high shape accuracy, much better than in extrusion, are ob-
tained by dry tableting when a dry powder is pressed between two punches in a
press. The powder flows through a feeder mechanism, and is filled into a rotating
die, where it is compressed by pistons (Fig. 2.161). The maximum pressure and the
rate of the pressure are the parameters influencing pellet hardness and integrity.
High mechanical stress during tableting has the downside of potential crushing of
crystals, dislocations of crystal planes and low porosity, typically not exceeding

Fig. 2.161: A tableting machine cycle.

 2.3 Preparation of catalytic materials 219

30%, thereby the formulation can contain porosity improvers if porosity is impor-
tant for the final application.
At the start of the cycle, the lower piston is in the upper position. The pellet
made in the previous cycle is pushed out as the die passes under the feeder mecha-
nism. The upper piston moves down by a fixed length, which defines the dimension
of the pellet. Although usually simple cylinders with almost the same height and
diameter are produced, it is also possible to form ring-shape tablets with special
dies and punches.
The important parameter in tableting is plastic deformation, which depends on
the applied stress, overall time and rate of compression, and the time for the mate-
rial to be under maximum stress. The elastic component of the response to the ap-
plied pressure stored in the material is relaxed after the pressure release. A too high
elastic component gives unstable tablets.
Small amounts of liquid or solid additives may be used to prevent dust formation
and improve compaction by lubrication. Because in tableting dry feeds are used, thus
lubricants or pilling aids such graphite, PVA, polyethylene, talc, silicates, alumi-
nates, to name a few, are added to enable the powder to flow. One of the most fre-
quently used additives is graphite at concentrations between 0.5 and 1%. If for some
reasons application of graphite can deteriorate catalyst performance, it should be re-
moved by careful oxidation.
In general, the feed for tableting should have flow properties of sand. Mechanical
stability after compression can be improved by adding binders. Moreover, high me-
chanical stability is obtained when the feed consists of mixed particle sizes. If the feed
is not sufficiently dense prior to tableting, pre-compacting is applied.
Different geometries can be achieved during tabletting including tablets per se
when the height is low than the diameter or cylinders when the opposite holds
(height > diameter). Moreover, cylinders with holes or ribs on the external surface
as well as their combination can be manufactured.
An example of nickel-molybdenum alumina catalyst preparation is shown in
Fig. 2.162. Suspension of aluminum hydroxide and water solutions of NH4MoO4 and
nitric acid are prepared respectively in vessels (pos. 1–3) and then mixed in a reac-
tor (pos. 4). Thereafter the tablets are produced from this mixture (pos. 6). After dry-
ing at room temperature (pos. 7) for 24 h the tablets are first dried at 383–423 K for
12 h and then calcined at 803 K for 8 h. After grinding the tablets to the powder
size, the powder is impregnated with Ni(NO3)2 for 1 h in the reactor (pos. 4).
Subsequent tableting, drying and calcination give the final catalyst that contains
13–15 wt% MoO3, 3–4 wt% NiO, 0.05–0.1 wt% Na2O, 0.2–0.3 wt% Fe2O3 and alu-
mina (the rest). Approximately 30 kg of NOx is released per ton of catalyst.
220 2 Engineering catalysts

(NH4)MoO4
Al(OH)3 H2O H2O HNO3 H2O

Ni(NO3)2
1 2 3 5 NaOH
12

Steam
13 14
8
4 9
6 7

Gas 11
10
Air
Air

Fig. 2.162: Production of nickel-molybdenum alumina catalyst. (From [48].) 1, 2, 3, 5, vessels for
Al(OH)3 suspension and water solutions of NH4MoO4, HNO3, Ni(NO3)2; 4, reactor; 6, tableting
machine; 7, oven; 8, lift; 9, calcination unit; 10, burner for flue gases generation; 11, heat
exchanger; 12, absorber with caustic solution; 13, milling; 14, intermediate vessel.

2.3.7.5 3D printing in catalytic technology

The 3D printing process is a promising manufacturing method, which can in the fu-
ture be applied to many different catalytic materials, allowing careful design of tar-
geted structures. The method is suitable for large-scale manufacturing as well as for
catalyst development being applicable to lab-scale operations. 3D printing allows ac-
curate control of various sophisticated structures, including monolithic catalysts with
a complex preparation procedure. Not surprisingly, additive manufacturing tools
(e.g., 3D printing) have been mainly explored for shaping 3D catalyst bodies such as
structured monoliths with improved mass and heat transfer characteristics.
Figure 2.163 illustrates a comparison between an extruded monolith and a robocast
structure with FCC (Fig. 2.163C) lattice. In robocasting, highly concentrated suspen-
sions of the material and a solvent are deposited in a layer-wise fashion through a
syringe-type nozzle. The minimum printable feature size in robocasting is typically
30–500 μm.
Besides fabricating structured monoliths, geometrically complex reaction vessels
and microreactors with integrated catalysts can be manufactured using 3D printing,
 2.3 Preparation of catalytic materials 221

20 mm

(a) (b) (c)

Fig. 2.163: Monoliths of similar outer dimensions and surface areas: (a) extruded honeycomb, (b)
robocast material with (c) FCC lattice (Reproduced from [88]).

when the structure of the target material can be first designed through computer-
aided design.
The most suitable methods for catalyst preparation include (Fig. 2.164) fused
deposition modeling (FDM), direct ink writing (DIW), stereolithography (SLA) and
selective laser sintering (SLS).
In SLS a laser is used as the power source to sinter a powdered material, while
in SLA layer-by-layer deposition is done using photochemical processes. DIW (robo-
casting) is an extrusion-based printing process similar to FDM with the difference
that contrary to the latter process in DIW, the material is extruded directly without
melting or solidification.
Similar to extrusion processes described earlier, a low viscosity of the printing
ink is required for maintaining fluidity while passing through the nozzle. Thereafter,
the material should maintain the shape on the printing structure, which implies high
viscosity. Pseudoplastic fluids used in robocasting typically require utilization of a
solvent, necessitating post treatment drying and sintering. In addition to solvents,
mixtures for robocasting of more complex materials, such as zeolites, should contain
different additives aiding extrusion (e.g., clays and plasticizing organic binders).
Future advances in 3D printing of catalysts require [89]:
– better control of the micro/mesostructures of printed catalytic materials by im-
proving the accuracy;
– improvement of the feedstock materials to diminish processing costs;
– structural optimization of 3D-printed materials, including pore size, pore struc-
ture and active phase distribution;
– better preservation of the 3D-printed structures after post-treatment processes.
222 2 Engineering catalysts

(a) Fused deposition modeling (b) Stereolithography

Lenses
Material filament
Scanning mirror
Step motor Laser
Liquifier Build platform Part Elevator
Extrusion nozzle
Part Sweeper

Vat
Build platform

Liquid photopolymer
Material spool
(c) Direct ink writing (d) Selective laser Sintering

Lenses
Scanning mirror
Extruder
Laser
Powder supply
Extrusion nozzle Powder bed Part
Part

Leveling roller
Build platform
Build piston

Figure 2.164: Some 3D printing processes that are useful for catalyst preparation: (a) fused
deposition modeling; (b) stereolithography; (c) direct ink writing; and (d) selective laser sintering.
(From [89]).

2.3.7.6 Scaling-up catalyst preparation

The typical strategy in academic work on catalyst design is related to the use of cat-
alyst particles of such a size that diffusion limitations are suppressed. Of less con-
cern are the preparation methods and selection of precursors, which are feasible for
use on an industrial scale. For example, metal chlorides could be used for impreg-
nation on a laboratory scale, while nitrates are preferred industrially. Catalyst re-
duction in academia is usually carried out by molecular hydrogen, while some
catalyst manufactures prefer reduction with chemical reagents. There is therefore a
difference in catalytic materials prepared on a small scale and those used for pilot-
ing and eventual production. For example, the type of the support or the
 2.3 Preparation of catalytic materials 223

preparation method should not necessarily be the same. Commercial raw materials
are used for catalyst preparation in a large scale instead of pure chemicals on a lab-
oratory scale. Water is the most used solvent, while the use of organic solvents is
uncommon.
In lab scale, costs of chemicals are rarely considered, while cost-effective raw
materials should be selected for industrial production. Moreover, the corrosive na-
ture of raw materials, possibility of emissions and their mitigation, efficient dis-
posal of effluents, wastewater purification along with process water recovery
should be carefully considered, limiting catalyst preparation technology. Low slurry
concentration during precipitation can be, for example, applicable in lab-scale rec-
ipes but will be challenging from the production technology viewpoint.
Preparation of catalysts for larger than lab scale is not straightforward as pa-
rameters such as stirring rate, heating and cooling rates, heat transfer, flow pattern,
drying and calcination conditions can be different. In particular, heating and cool-
ing rates are much slower when the scale is increasing. This influences chemical
and physical process in catalyst preparation, for instance, crystallization by hydro-
thermal synthesis.
In addition, large reactors have more prominent spatial inhomogeneity in tem-
perature, pH and concentration. As the size increases, significant variations in the
local temperature profiles are possible during drying and calcination. Proper equip-
ment design is thus essential to minimize different spatial variations. Larger pro-
duction volume also requires equipment capable of handling such volumes.
Preparation methods used in industry are mainly kept as trade secrets or cover
a broad range of parameters, making it difficult to reveal the exact preparations de-
tails. Very few flow schemes for catalyst preparations can be found in the open
literature.
In general, scaling up the catalyst preparation procedure is not an easy task. In
pilot scale and industrial fixed-bed reactors, usually larger catalyst particles are
used than in the laboratory scale where the aim is to avoid problems with the pres-
sure drop. This can result in changes not only in the reactant conversion but also in
the product selectivity, which is known to be influenced by internal and external
diffusion.
Lab-scale academic research, aimed primarily at publications in high-impact
factor journals, does not typically consider a necessity of a high crush strength re-
quired for applications in fixed bed reactors, when also a minimally allowed pres-
sure drop should be achieved. This requirement implies higher compaction during
forming at the expense of porosity as well as elevated heat and mass transfer within
the catalyst particle. The latter can obviously influence activity and selectivity. Not
only the value of the crushing strength is of importance, but also the distribution in
strength, in particular percentage of catalyst particles with low strength should be
minimal, as otherwise the particles will break giving high pressure drop.
224 2 Engineering catalysts

Another important issue is the presence of a binder, for example, for extrusion.
Introduction of a binder can influence the catalyst performance, thus such differen-
ces should be recognized when progressing from laboratory to pilot and industrial
scale with kinetic and diffusion modeling.
Catalyst stability is an important issue that is often not in the focus of experi-
mental and theoretical studies at the laboratory scale. The necessity to develop ma-
terials and structures resistant to deactivation also requires some alterations on the
material level, resulting in a difference between catalysts tested in a lab and on an
industrial scale. In conclusion the scale-up of the catalyst preparation methods
from a laboratory scale to pilot and industrial scales is challenging, which calls for
more widespread involvement of the theories of physical chemistry and chemical
engineering in catalyst preparation.

References
[1] Rouquerol, F., Rouquerol, J., Sing, K.S.W. (1999) Adsorption by Powders and Porous Solids.
San Diego, CA: Academic Press, pp. 1–25.
[2] Leofanti, G., Padovan, M., Tozzola, G., Venturelli, B. (1998) Surface area and pore texture of
catalysts. Catal. Today 41: 207.
[3] http://www.vermeer.net/pub/communication/downloads/future-perspectives-in-cata.pdf.
[4] https://pubs.acs.org/doi/10.1021/acscatal.8b01708
[5] Catalysts: Combinatorial Catalysis, 2005, DOI: 10.1016/B0-12-369401-9/01128-1,
Encyclopedia of Condensed Matter Physics, A. Hagemeyer, A. Volpe
[6] Rodemerck, U., Linke, D. (2009) High throughput experimentation, in. Synthesis of solid
catalysts, ed. K.P. de Jong., Wiley, Weinheim.
[7] Li, H., Zhang, Z., Liu, Z. (2017); Application of Artificial Neural Networks for catalysis: A
Review, Catalysts 7: 306; doi:10.3390/catal7100306. https://onlinelibrary.wiley.com/doi/
full/10.1002/aic.16198
[8] Medford, A.J. Ross Kunz, M., Sarah M. Ewing, S.M., Borders, T., Fushimi, R. (2018),
Extracting knowledge from data through catalysis informatics, ACS Catal. 8: 7403-7429.
https://pubs.acs.org/doi/10.1021/acscatal.8b01708
[9] Rothenberg, G. (2008) Data mining in catalysis: Separating knowledge from garbage,
Catal. Today, 138:2-10. https://www.sciencedirect.com/science/article/pii/
S092058610800062X
[10] Goldsmith, B.R., Esterhuizen, J., Liu, J.‐X., Bartel, C. J., Sutton, C. (2018) Machine learning for
heterogeneous catalyst design and discovery, AIChe J, 64: 2311-2323. https://pubs.acs.org/
doi/10.1021/acscatal.8b01708
[11] Condon, J.B. (2006) Surface Area and Porosity Determination from Physisorption,
Measurements and Theory. Elsevier.
[12] Dubinin, M.M. (1960) The potential theory of adsorption of gases and vapors for adsorbents
with energetically nonuniform surfaces. Chem. Rev. 60: 235–241.
[13] Kalantar Neyestanaki, A., Mäki-Arvela, P., Backman, H., Karhu, H., Salmi, T., Väyrynen, J.,
Murzin, D.Yu. (2003) Kinetics and stereoselectivity of o-xylene hydrogenation over Pd/Al2O3.
Mol. Catal. A. Chem. 193: 237–250.
 References 225

[14] Bayraktar, O., Kugler, E.L. (2002) Characterization of coke on equilibrium fluid catalytic
cracking catalysts by temperature-programmed oxidation. Appl. Catal. A. Gen. 233: 197–213.
[15] Sahin, S., Mäki-Arvela, P., Tessonier, J.P., Villa, A., Reiche, S., Wrabetz, S., Su, D.S., Schlögl,
R., Salmi, T., Murzin, D.Yu. (2011) Palladium catalysts supported on N-functionalized hollow
vapour-grown carbon nanofibers: the effect of the basic support and catalyst reduction
temperature. Appl. Catal. A. Gen. 408: 137–147.
[16] Li, L., Lin, J., Li, X., Wang, A., Wang, X., Zhang, T (2016) Adsorption/reaction energetics
measured by microcalorimetry and correlated with reactivity on supported catalysts: A
review. Chin. J. Catal. 37: 2039–2052.
[17] Simakova, I. (2010) Catalytic transformations of fatty acid derivatives for food, oleochemicals
and fuels over carbon supported platinum group metals. PhD thesis. Åbo Akademi University,
Turku, Finland.
[18] Sharma, R., Bisen, D.P., Shukla, U., Sharma, B.G. (2012) X-ray diffraction: a powerful method
of characterizing nanomaterials. Recent Res. Sci. Techn. 4: 77–79.
[19] Davis, S.M. (1990) Effect of surface roughness on particle size predictions from
photoemission results. J. Catal. 122: 240–246.
[20] Eränen, K., Klingstedt, F., Arve, K., Lindfors, L.-E., Murzin, D.Yu. (2004) On the mechanism of
the selective catalytic reduction of NO with higher hydrocarbons over a silver/alumina
catalyst. J. Catal. 227: 328–343.
[21] Villegas Perdomo, J.I. (2006) Engineering bifunctional catalysts for hydrocarbon valorisation,
PhD thesis. Åbo Akademi University, Turku, Finland.
[22] Guerrero-Perez, M.O., Banares, M.A. (2006) From conventional in situ to operando studies in
Raman spectroscopy. Catal. Today 113: 48–57.
[23] Aho, A., Kumar, N., Lashkul, A., Eränen, K., Ziolek, M., Decyk, P., Salmi, T., Holmbom, B.,
Hupa, M., Murzin, D.Yu. (2010) Catalytic upgrading of woody biomass derived pyrolysis
vapors over iron modified zeolites in a dual-fluidized bed reactor. Fuel 89: 1992–2000.
[24] Arve, K., Popov, E.A., Rönnholm, M., Klingstedt, F., Eloranta, J., Eränen, K., Murzin, D.Yu.
(2004) From a fixed-bed Ag-alumina catalyst to a modified reactor design: how to enhance
the crucial heterogeneous-homogeneous reactions in HC-SCR. Chem. Eng. Sci. 59:
5277–5282.
[25] US Patent US 8,017,086 B2 (2011) to Volvo Technology Corporation, Catalyst unit for
reduction of NOx compounds.
[26] Fierro, G., Moretti, G., Ferraris, G., Andreozzi, G.B. (2011) A Mossbauer and structural
investigation of Fe-ZSM-5 catalysts: Influence of Fe oxide nanoparticles size on the catalytic
behaviour for the NO-SCR by C3H8. Appl. Catal. B-Environ. 102: 215.
[27] Nieminen, V., Karhu, H., Kumar, N., Heinmaa, I., Ek, P., Samoson, I., Salmi, T., Murzin, D.Yu.
(2004) Physico-chemical and catalytic properties of Zr- and Cu-Zr ion-exchanged H-MCM-41.
Phys. Chem. Chem. Phys. 6: 4062–4069.
[28] Niemantsverdriet, J.W. (2007) Spectroscopy in Catalysis. Weinheim, Germany: Wiley-VCH.
[29] Chorkendorff, I., Niemanstverdriet, J.W. (2003) Concepts of modern catalysis and kinetics.
Weinheim, Germany: Wiley-VCH.
[30] Lauritsen, J.V., Besenbacher, F. (2006) Model catalyst surfaces investigated by scanning
tunneling microscopy. Adv. Catal. 50: 97–147.
[31] Vang, R.T., Lauritsen, J.V., Lægsgaard, E., Besenbacher, F. (2008) Scanning tunneling
microscopy as a tool to study catalytically relevant model systems. Chem. Soc. Rev. 37:
2191–2203.
[32] Gladden, F., Mantle, M.D., Sederman, A.J. (2006) Magnetic resonance Imaging of catalysts
and catalytic processes. Adv. Catal. 50: 1–75.
226 2 Engineering catalysts

[33] Karreman, M.A., Buurmans, I.L.C., Geus, J.W., Agronskaia, A.V., Ruiz-Martínez, J., et al. (2012)
Integrated laser and electron microscopy correlates structure of fluid catalytic cracking
particles to Brønsted acidity. Angew. Chem. Intern. Ed. 51: 1428–1431.
[34] Piccolo, L., Nassreddine, S., Toussaint, G., Geantet, C. (2010) Discussion on “A comprehensive
two-dimensional gas chromatography coupled with quadrupole mass spectrometry approach
for identification of C10 derivatives from decalin” by C. Flego, N. Gigantiello, W.O. Parker, Jr., V.
Calemma [J. Chromatogr. A 1216 (2009) 2891].
[35] Bengaard, H.S., Norskov, J.K., Sehested, J.S., Clausen, B.S., Nielsen, L.P., et al. (2002) Steam
reforming and graphite formation on Ni catalysts. J. Catal. 209: 365.
[36] Rostrup-Nielsen, J.R. (2012) Perspective of Industry on Modeling Catalysis. In:
Deutschmann, O., editor. Modeling and Simulation of Heterogeneous Catalytic Reactions:
From the Molecular Process to the Technical System. Weinheim, Germany: Wiley-VCH,
pp. 283–301.
[37] Norskov, J.K., Abild-Pedersen, F., Studt, F., Bligaard, T. (2011) Density functional theory in
surface chemistry and catalysis. Proc. Nat. Acad. Sci. 108: 937–943.
[38] Campanati, M., Fornasari, G., Vaccari, A. (2003) Fundamentals in the preparation of
heterogeneous catalysts. Catal. Today 77: 299–314.
[39] Fokin, S. (1913) Catalytic oxidation reaction at high temperatures, Preliminary communication,
J. Russ. Phys. Chem. Soc. 45: 286-288.
[40] Xu, Q., Pearce, G.K., Field, R.W. (2017), Pressure driven inside feed (PDI) hollow fibre
filtration: Optimizing the geometry and operating parameters, J. Membrane Sci., 537:
323–336. https://www.sciencedirect.com/science/article/pii/S0376738817306403
[41] https://en.wikipedia.org/wiki/Rotary_vacuum-drum_filter#/media/File:Rotary_vacuum-drum
_filter.svg
[42] https://www.tsk-g.co.jp/en/tech/industry/horizontal-belt-filter.html
[43] http://trends.directindustry.com/comber-process-technology-srl/project-34050-131310.html
[44] http://www.solidliquid-separation.com/pressurefilters/nutsche/nutsche.htm
[45] http://www.gusu.com.cn/index_e.asp
[46] An, K., Alayoglu, S., Ewers, T., Somorja, G.A. (2012) Colloid chemistry of nanocatalysts: A
molecular view. J. Colloid. Interf. Sci. 373: 1.
[47] Jin, R. (2012) The impacts of nanotechnology on catalysis by precious metal nanoparticles.
Nanotechnol. Rev. 1: 31–56.
[48] Mukhlenov, I.P., Dobkina, E.I., Derjuzhkina, V.I. (1989) Catalysts Technology. Leningrad:
Khimia.
[49] Rusanov, A.I. (2012) The development of the fundamental concepts of surface
thermodynamics, Colloid J. 74: 136–153.
[50] Pakhomov, N.A. (2011) Scientific Basis for Catalyst Preparation. Novosibirsk: Institute of
Catalysis.
[51] Moulijn, J.A. http://www.tnw.tudelft.nl/fileadmin/Faculteit/TNW/Over_de_faculteit/
Afdelingen/Chemical_Engineering/Research/Catalysis_Engineering/Education/
Miscellaneous/doc/catcat03.ppt.
[52] Kolesnikov, I.M. (2004) Catalysis and Production of Catalysts. Moscow, Russia: Teknika.
[53] Bathgate, G., Iyuke, S., Kavishe, F. (2012) Comparison of straight and helical nanotube
production in a swirled fluid CVD reactor. ISRN Nanotechnology. Article ID 985834,
doi:10.5402/2012/985834. http://www.hindawi.com/isrn/nanotechnology/2012/985834/
[54] Bordere, S., Corpart, J.M., Bounia, El., Gaillard, N.E., Passade-Boupat, P.N., Piccione, P.M.,
Plee, D., Industrial production and applications of carbon nanotubes. http://www.graphis
trength.com/export/sites/graphistrength/.content/medias/downloads/literature/General-in
formation-on-carbon-nanotubes.pdf
 References 227

[55] www.bayerbms.com
[56] Jiménez, V., Ramírez-Lucas, A., Díaz, J.A., Sánchez, P., Romero, A. (2012) CO2 capture in
different carbon materials, Environ. Sci. Technol. 46: 7407–7414.
[57] Strobel, R., Baiker, A., Pratsinis, S.E. (2006) Aerosol flame synthesis of catalysts. Adv.
Powder Technol. 17: 457–480.
[58] Hanefeld, U, Lefferts, L. (Ed). (2018) Catalysis. An integrated textbook for students, Wiley,
Weinheim.
[59] Kubicka, D., Kumar, N., Venäläinen, T., Karhu, H. Kubickova, I., et al. (2006) The metal-
support interactions in zeolite-supported noble metals: the influence of metal crystallites on
the support acidity. J. Phys. Chem. B. 110: 4937–4946.
[60] Bulut, M., Jacobs, P.A. Concepts for preparation of zeolite-based catalysts, in Synthesis of
Solid Catalysts (ed. K.P. De Jong), Wiley, Weinheim, 2009, 243–376.
[61] Grand, J., Awala, H., Mintova, S. (2016) Mechanism of zeolites crystal growth: new findings
and open questions, CrystEngCom, 18: 650–664.
[62] Kumar, N., Nieminen, V., Demirkan, K., Salmi, T., Murzin, D.Yu. et al. (2002) Effect of
synthesis time and mode of stirring on physico-chemical and catalytic properties of ZSM-5
zeolite catalysts. Appl. Catal. A-Gen. 235: 113.
[63] Fraile, J.M., Herrerías, C.I. (2011) CAFC9: 9th congress on catalysis applied to fine chemicals.
Platinum Met. Rev. 55: 13-19.
[64] http://www.chem.strath.ac.uk/people/academic/lorraine_gibson/research/sorbents.
[65] Twigg, M.V., Richardson, J.T. (2002) Theory and applications of ceramic foam catalysts,
Chem. Eng. Res. Des. 80: 183–189.
[66] Vergunst, T., Kapteijn, F., Moulijn, J.A. (2001) Monolithic catalysts - non-uniform active phase
distribution by impregnation. Appl. Cat. A: Gen. 213: 179–187.
[67] http://www.selee.com/Selee_Corporation_Metal_Foam.php.
[68] Fey, T., Betke, U., Rannabauer, S., Scheffle, M. (2017) Reticulated replica ceramic foams:
processing, functionalization, and characterization, Adv. Eng. Mat. 19: 170036.
[69] Spieker, W.A., Regalbuto, J.R. (2001) A fundamental model of platinum impregnation onto
alumina. Chem. Eng. Sci. 56: 3491–3504.
[70] Geus, J.W., van Veen, J.R. (1999) Preparation of supported catalysts, Chapter 10, In: Moulijn,
J.A., van Leeuwen, P.W.N.M., van Santen, R.A., editors. Catalysis, An Integrated Approach to
Homogeneous, Heterogeneous and Industrial Catalysis. Amsterdam, The Netherlands:
Elsevier, p. 467.
[71] Jiao, L., Regalbuto, J.R. (2008) The synthesis of highly dispersed noble and base metals on
silica via strong electrostatic adsorption: I. Amorphous silica. J. Catal. 260: 329–341.
[72] Marceau, E., Carrier, X., Che M. (2009) Impregnation and drying. In: de Jong, K.P., editor.
Synthesis of Solid Catalysts. Weinheim, Germany: Wiley-VCH.
[73] Liu, X., Khinast, J.G., Glasser, B.J. (2008) Parametric investigation of impregnation and drying
of supported catalysts. Chem. Eng. Sci. 63: 4517–4530.
[74] Hagen, J. (2006) Industrial Catalysis. Weinheim, Germany: Wiley-VCH.
[75] Stillwell, W.D. (1957) Preformed catalysts and techniques of tableting. Ind. Eng. Chem. 49:
245–249.
[76] https://prodselmash.ru/specializirovannoe-oborudovanie/ustanovki-dla-vypuska-
vysokotehnologicnoj-produkcii/61
[77] Murzin, D.Yu., Simakova, O.A., Simakova, I.L., Parmon, V.N. (2011) Thermodynamic analysis
of the cluster size evolution in catalyst preparation by deposition-precipitation. React. Kinet.
Catal. Let. 104: 259–266.
[78] Lu, J., Lei, Y., Elam J.W. Atomic layer deposition of noble metals – new developments in
nanostructured catalysts. Noble Metals. doi: 10.5772/33082.
228 2 Engineering catalysts

[79] www.oxford-instruments.com.
[80] Schüth, F., Hesse, M. (2008) Catalyst forming. In: Ertl, G., Knözinger, H., Schüth, F.,
Weitkamp, J., editors. Handbook of Heterogeneous Catalysis. Weinheim, Germany: Wiley-
VCH. 676–699.
[81] http://www.niroinc.com/food_chemical/spray_dryer_selection.asp.
[82] Roberie, T., Hildebrandt, D., Creighton, J., Gilson, J.P. (2002) Preparation of zeolite catalysts,
Chapter 3, in Zeolites for Cleaner Technologies. Guisnet, M.J., Gilson, P. Editors. London, UK:
Imperial College Press, 57-73. Available from http://www.granmix.ru/equip/.
[83] https://media.licdn.com/dms/image/C5612AQFeVY3UvI9UJA/article-inline_image-shrink
_1500_2232/0?e=1568851200&v=beta&t=YNcarijTqHhqPeqqFZf3Ci4JGqS4bUPVvYs8DSIkk_c
[84] http://www.thermopedia.com/content/5408/762NNFFig1.gif
[85] Devyatkov, S. Yu., Zinnurova, A.A., Aho, A., Kronlund, D., Peltonen, J., Kuzichkin, N. V.,
Lisitsyn, N. V., Murzin, D.Yu. (2016) Shaping of sulfated zirconia catalysts by extrusion:
understanding the role of binders, Ind. Eng. Chem. Res., 55: 6595.
[86] Freiding, J., Patcas, F.-C., Kraushaar-Czarnetzki, B. (2007) Extrusion of zeolites: Properties of
catalysts with a novel aluminium phosphate sinter matrix, Appl. Catal. A Gen. 328: 210.
[87] http://upload.wikimedia.org/wikipedia/commons/thumb/2/21/Al_extrusion_die_set.jpg/
440px-Al_extrusion_die_set.jpg.
[88] Stuecker, J. N., Miller, J. E., Ferrizz, R. E., Mudd, J.E., Cesarano, J. (2004) Advanced support
structures for enhanced catalytic activity, Ind. Eng. Chem. Res., 431: 51–55.
[89] Zhou, X., Liu, C.-J., (2017) Three-dimensional printing for catalytic applications: Current
status and perspectives, Adv. Funct. Mater. 27, 1701134. https://onlinelibrary.wiley.com/doi/
10.1002/adfm.201701134.
3 Engineering reactions

3.1 Introduction

In order to understand a process it is important to have a model describing the fac-

tors that control the process outcome. Knowledge of reaction thermodynamics and
kinetics is thus required and physico-chemical understanding of catalytic processes
is essential for proper reactor and process design. According to the very definition
of catalysis, it is a kinetic phenomenon, and thus reliable kinetic models, which de-
scribe the rates and equilibria of catalytic reactions, are of vital importance for solv-
ing the problems in mathematical modeling, design and intensification of chemical
processes.
Moreover, the performance of catalytic processes depends on how the reaction
is conducted and the physical state of reactants and equipment (for example, the
reactor). Understanding of hydrodynamics, mass transfer and heat transfer along
with reaction kinetics is needed in order to properly select conditions, equipment
and operation mode (batch, semi-batch or continuous). It should also be remem-
bered that while the intrinsic rate constants do not change with scale and equipment,
rates of physical processes are scale- and equipment-dependent thus severely influ-
encing the outcome of a particular heterogeneous catalyst reaction in terms of activ-
ity, conversion or selectivity.
As heterogeneous catalytic reactions inherently require several phases, inter-
phase mass transfer is always present, increasing in complexity when the reaction
system is multiphase, which comprises not only the solid catalysts but reactants in
both gas and liquid (or several liquid) phases. In addition to influencing the reac-
tion rate, the mass transfer can influence selectivity in complex reactions, as men-
tioned above and addressed in detail below. The heat transfer depends on the
reaction rates, which could be concentration dependent.
Proper engineering on catalytic reactions needs understanding of the process
chemistry and reaction mechanism but also data on physico-chemical properties of
reactants, phase behavior, reaction thermodynamics and kinetics, catalyst deactiva-
tion, mass transfer and heat transfer.

3.2 Thermodynamics

Evaluation of thermodynamics is the starting point in reaction engineering, as it ad-

dresses the overall feasibility of the reaction and defines the composition of the re-
action mixture at equilibrium and thus the maximum conversion and selectivity. In
addition, the reaction rate depends on the approach to equilibrium, or the distance

https://doi.org/10.1515/9783110614435-003
230 3 Engineering reactions

of the reaction mixture composition to the equilibrium one, as at equilibrium the

rate of the forward and reverse reactions is equal to each other.
Thermodynamic conditions for equilibrium system:

aA + bB = cC + dD (3:1)

can be expressed through the Gibbs energy, which must be stationary. This means
that the derivative of G with respect to the extent of reaction, ξ, is zero, or that the
sum of chemical potentials µ (partial molar Gibbs energy) of the products and reac-
tants are equal to each other:

aμA + bμB = cμC + dμD (3:2)

where chemical potentials of reagents depend on activity:

μi + μ0i = RT ln a. (3:3)

In eq. (3.3) μi0 is the standard chemical potential. If a mixture is not at equilibrium,
the driving force for the reaction to approach equilibrium, leading to subsequent
changes in composition, is the minimization of the excess Gibbs energy (or Helmholtz
energy at constant volume reactions). The equilibrium constant for eq. (3.1):

½Aa ½Bb
K= (3:4)
½Cc ½Dd

is related to the standard Gibbs energy change for the reaction by the equation:

ΔG0T = − RT ln Keq (3:5)

where R is the universal gas constant and T the temperature.

The Gibbs energy is defined through reaction enthalpy ΔHT0 and entropy ΔS0T

ΔG0T = ΔHT0 − TΔS0T (3:6)

The temperature dependence of the standard Gibbs energy change of the reaction is
given by:

∂ΔG0 ΔG0 − ΔH 0
= (3:7)
∂T T
After rearrangement and division by RT, we arrive at an expression relating reaction
enthalpy and equilibrium constant:
 0
ΔG
  ∂
ΔH 0 1 ∂ΔG0 ΔG0 RT ∂ðln KÞ
=− − =− = (3:8)
RT 2 RT ∂T RT 2 ∂T ∂T
 3.2 Thermodynamics 231

The standard enthalpy changes are given by:

ðT !
X X X
Ni
ΔHT0 = vi ΔH298K
0
ðprodÞ − vi ΔH298K
0
ðreactÞ + vi cp, i dT (3:9)
i=1
298K

where cp,i is the temperature-dependent heat capacity, which for gases is given by:

cp, i = ai ′ + bi ′T + ci ′T 2 + di ′T − 2 (3:10)

In eq. (3.10) a, b, c and d are coefficients. The rates of forward and reverse reactions
are related through thermodynamics in the following way. Expressing the rates of
the forward and the reverse reactions, respectively:
Y α ,i Y
r+ = k+ CI + , r − = k − CIα − , i (3:11)
I I

the ratio of rates is:

r + k + Y α +, i − α − , i Y α ,i−α ,i
= CI =K CI + −
(3:12)
r− k− I I

with the equilibrium constant defined as K= k+/k_. In eq. (3.11) α+, i stands for reac-
tion order in compound I in the forward direction. Equation (3.12) can be rearranged
to the De Donder equation:
r+
= eA=RT (3:13)
r−

where A is the reaction affinity:

Y
A = RT ln K CIα + , i − α − , i (3:14)
I

which is defined as the derivative of the Gibbs free energy with respect to the extent
of the reaction:
 
∂G
A= − (3:15)
dξ T, p

The relationship between the overall process rate and the rate in the forward reac-
tions is also expressed through reaction affinity:
 
−A ðCÞ
r = r + − r − = r + 1 − e RT = r + 1 − φ (3:16)
K

where r is the overall rate, r+ is the rate of the forward reaction, φ(C) is a function of
only reactants concentrations and K is the equilibrium constant. For a reversible
232 3 Engineering reactions

reaction in eq. (3.1), assuming that stoichiometry corresponds to molecularity, the

rate is given by:
! !
k − CCc CDd 1 CCc CDd
r = k + CA CB − k − C C CD = k + CA C B 1 −
a b c d a b
= r+ 1 − (3:17)
k + CAa CBb K CAa CBb

where CA, etc. are partial pressures (in the case of gas-phase reactions) or concen-
trations (for liquid-phase reactions).

3.3 Kinetics

3.3.1 Definitions

Chemical kinetics studies the rates of chemical reactions and deals with experimen-
tal measurements in batch, semi-batch or continuous reactors. It addresses how the
reaction rates depend on concentrations, temperature, nature of catalyst, pH and
solvents, to mention a few reaction parameters.
Experimental kinetic investigations are one of the powerful ways to reveal the re-
action mechanisms. The following problems can be solved by using a kinetic model:
– choosing the catalyst and comparing the selectivity and activity of various cata-
lysts and their performance under optimum conditions for each catalyst;
– determination of the main and by-products formed during the process;
– determination of the optimum sizes and structures of catalyst grains and the
necessary amount of the catalyst to achieve the specified values of conversion
and selectivity;
– elucidation of the dynamics of the process and the determination of the behav-
ior of the reactor under unsteady state conditions;
– revealing the influence of mass and heat transfer processes on the chemical re-
action rates and product selectivity as well as the determination of the kinetic
region of the process;
– selecting the type of a reactor and structure of the contact unit that provides
the best approach to the optimum conditions.

Kinetic data thus forms the basis of catalysis, allowing the design and modeling of
catalytic reactors. In addition to these practical aspects, kinetic analysis is also im-
portant in the elucidation of catalytic reaction mechanisms.
Formulation of the reaction mechanism needed for kinetic modeling is difficult
and should be based on all available information about the reaction chemistry, ki-
netic experiments in steady and non-steady-state conditions, physical measurements,
surface science data and theory.
 3.3 Kinetics 233

Accurate kinetic measurements with high precision are needed as large devia-
tions in the values of the experimentally measured rates pose serious obstacles for
quantitative considerations. Such deviations could be caused by variations in the
feedstock, inhomogeneity of catalyst samples, alterations of reaction parameters dur-
ing a reaction, influence of heat and mass transfer. Another necessary feature is the
ability to generate maximum kinetic information in the minimum time. The analysis
of products as well as reactor lay-out should preferably be as easy as possible.
It should be noted that on many occasions kinetic calculations are based on erro-
neous data and interpretations. In Fig. 3.1, conversion after a certain period of time
reaches 100% independent of the catalyst applied. For example, reported values of
100% conversion after 24 h and corresponding turnover frequency (TOF) values de-
fined by eq . (1.3) are thus often misleading, because catalysts have very different activ-
ity. There is also an apparent danger in using the values of initial rates, as the reaction
rates are not obtained directly from measurements. As an illustration, in Fig. 3.1 a case
is presented when the calculation of the initial rate based on the first measured experi-
mental point for the most active catalyst obviously gives an overestimation.

100

80
Conversion

60

40

20

0
0 5 10 15
Time (h)

Fig. 3.1: Concentration vs time dependence for difference catalysts (see explanations in the text).

For simple kinetics it is possible to apply algebraic equations for batch reactors.
Differential equations representing mass balances can be also analytically integrated
rather easily for simple cases.
In heterogeneous catalysis, turnover frequency is often reported as an indicator
of catalyst activity. In organometallic catalysis, turnover number (TON) is used,
which has a different meaning, that is, the number of moles of substrate that a mole
of catalyst can convert until it is deactivated completely. Reporting TON as the
number of transformations during a certain period of time without proving that the
catalytic activity is completely lost would be thus misleading.
234 3 Engineering reactions

Another common mistake regarding interpretation of deactivation is illustrated

in Fig. 3.2. It is sometimes considered that when conversion is 100% in fixed-bed
reactors and the activity decline is visible only after a certain time-on-stream, there
is no deactivation prior to that point. In fact, deactivation is also present even when
conversion is 100%, but the whole catalyst bed is not required to reach the full con-
version and the catalyst could be much more active initially. Stability of catalysts
should be thus analyzed at intermediate conversion levels, which can be done by
decreasing the residence time (i.e., higher flow or less catalyst).

Conversion

100%

Time

Fig. 3.2: Experimental data in a fixed-bed reactor showing an initial 100% conversion that declines
after some time-on-stream.

For reactions performed in the batch mode, evaluation of catalyst stability is done by
catalyst recycling. In many occasions, the authors are reporting conversion of 100%
after a certain period of time as an evidence of catalyst stability, which is misleading
as elaborated earlier. In batch experiments, the reaction times or catalyst amounts
should be thus selected to avoid complete conversion.

The author is reviewing a significant number of articles on catalysis for different journals report-
ing kinetic measurements mainly in batch and continuous reactors. There are several details of
experimental procedures, which are very often not reported. Few of them are listed below and
can be even used by the readers as a checklist, while preparing their next papers. I hope that it
will help in smoothening the reviewing process as the reviewers, including myself, will be less
frustrated asking the same questions over and over again.
One of the key questions is related to a potential impact of mass and heat transfer limitations.
Differences between different catalysts could be masked if the reaction is performed under a signifi-
cant influence of mass transfer. Simple diagnostic tests or more elaborative calculations can be done
to evaluate mass transfer limitations. In particular, to assess the internal mass transfer (i.e., inside
the catalyst grain), knowing the size of the catalyst particles is critical. For batch slurry reactors, con-
trary to fixed-bed reactors, the size of catalyst grains (in the micron range) is, however, frequently
missing in scientific publications. Seldom calculations of the catalyst effectiveness factor (will be dis-
cussed in Section 3.5) are done. The liquid load in a batchwise-operated slurry reactor should be ap-
proximately 0.3–0.5 of the total reactor volume to allow efficient energy dissipation. In many
instances, the load is much lower, making the reviewers wondering if the impeller is really in the
 3.3 Kinetics 235

liquid, but not above it. It is not enough to report the stirring speed to assess the external mass trans-
fer limitations, as it is the relative ratio between the mass transfer to the external catalyst surface to
reaction kinetics which is important, rather than the absolute values of the mass transfer coefficient.
In many instances, the catalyst and the reactant are introduced simultaneously and then the mix-
ture is heated to the reaction temperature. Apparently, things are happening during heating and
should be properly accounted for. A better option is to heat the reactant and catalyst separately.
When sampling of the reaction mixture is done, the volume of samples and details on the pres-
ence of the catalyst in the samples are rarely reported. Depending on how the samples are taken,
the catalyst bulk density can, however, change, influencing interpretation of the kinetic data.
Often three-phase reactions are performed under hydrogen or oxygen. Only the overall pres-
sure is often reported without specifying if this was an overpressure (manometer reading) or
the total pressure and what was the vapor pressure of the solvent at the reaction temperature.
It thus remains a mystery what was the pressure of the reacting gas, which is important for
evaluation of the reaction order and reaction kinetics.
In papers describing experiments in a continuous mode quite often, it is not clear at which
temperature the flow rates or residence times are reported (i.e., room or reaction temperature)
and how gas expansion was taken into account in calculation of selectivity.
Another typical mistake is not to report the mass balance (or carbon balance) closure.

An essential feature for catalytic reactions is the readiness in reduction/activation

of heterogeneous catalysts and the possibility to use them in the required geometri-
cal form. As activity deterioration takes place in many catalytic reactions (contrary
to the definition of catalysis) there should be always activity control in measure-
ments of catalytic kinetics.
A measure of catalytic activity is the reaction rate, which is defined through the
extent of the reaction. The change in the extent of reaction (number of chemical
transformations divided by the Avogadro number) is given by dξ = dnB/vB, where vB
is the stoichiometric number of any reaction entity B (reactant or product) and nB is
the corresponding amount.
This extensive property dξ/dt is measured in moles and cannot be considered
as reaction rate, since it is proportional to the reactor size.
For a homogeneous reaction, when the rate changes with time and is not uniform
over a reactor volume v, the reaction rate is:

∂2 ξ
r= (3:18)
∂t∂v
while for the constant reactor volume it is defined as:

1 dCi
ri = (3:19)
vi dt

where i is the reactant or product with a corresponding stoichiometric coefficient vi.

For a heterogeneous reaction occurring over a reaction space S (catalyst sur-
face, volume weight, or number of active sites) the rate expression is given by:
236 3 Engineering reactions

∂2 ξ
r= (3:20)
∂t∂S
leading to further simplifications when the rate is uniform across the surface:

1 ∂ξ
r= (3:21)
S ∂t

Rate laws express how the rate depends on concentration and rarely follows the
overall stoichiometry as given in eq. (3.1). In fact, reaction molecularity (the num-
ber of species that must collide to produce the reaction) determines the form of a
rate equation. Elementary reactions occur when the rate law can be written from
its molecularity and when the kinetics depend only on the number of reactant
molecules in that step. For elementary reactions the reaction orders have integral
values typically equal to 1 and 2 (Fig. 3.3), or occasionally 3 for tri-molecular
reactions.

Rate r = k (A), first order

r = k, zero order

(A) Fig. 3.3: Reaction kinetics of first and zero order.

Reaction orders, which for a particular reaction can be fractional (rA = –kcAn) indicating
a complex reaction mechanism, could be determined using logarithmic plots:
 
− dcA
ln = ln k + n ln cA (3:22)
dt

with the reaction order corresponding to the slope.

The temperature dependence of the reaction constants is expressed by the
Arrhenius equation:
− Ea
k = Ae RT (3:23)

where the pre-exponential factor A weakly depends on temperature, which is often

neglected. The Arrhenius equation should be applied for elementary reactions, and
thus typically apparent values of activation energies are reported for multistep het-
erogeneous catalytic reactions. Such values could be easily determined by taking
the natural logarithms of both sides of the eq. (3.23):
 3.3 Kinetics 237

 
Ea 1
ln k = ln A − (3:24)
R T
1 Ea
A plot of lnk vs gives a straight line with a slope of − and an intercept of lnA.
T R

3.3.2 Reaction mechanism

A mechanism of a catalytic reaction comprises several elementary steps. A linear

mechanism contains steps that contain only one adsorbed species on both sides of
these elementary steps, while non-linear steps might have more than one species.
A proposal for a reaction mechanism should correspond to the principle of mi-
croscopic reversibility, which implies that reactions proceed in both directions
through the same set of elementary steps and intermediates on the same type of cat-
alytically active centers. A rate-limiting step reaction, if any, should be the same in
both directions.
Derivation of kinetic equations can be substantially simplified using the steady-
state approximation originally developed by Bodenstein. In general, a reaction is con-
sidered to be at steady-state if the concentrations of all species in each element of the
reaction space (surface coverage of reactive intermediates) do not change in time
(Fig. 3.4).

Rate of producing the

intermediate is the
Intermediate

same as its rate of

consumption

Fig. 3.4: Concentration of a reaction intermediate as a

function of reaction time.

According to Bodenstein there are intermediates in chemical reactions that are present
in “inferior” amounts, for example, in much less concentrations than the major species
in the mechanism. Under these conditions, the rate of concentration changes of an in-
termediate can be considered negligible. The application of the steady-state approxi-
mation to heterogeneous catalysis does not require that the surface concentrations are
low, but implies that they do not change with time:

dθ
=0 (3:25)
dt
238 3 Engineering reactions

Another formulation of the steady-state approximation resulting from eq. (3.25) is

that all the steps in a catalytic reaction, occurring at steady-state, are equal to each
other.
To illustrate the steady-state approximation, an isomerization of A to B will be
considered:

1. A + * , A*
2. A* ! B + * (3:26)
A=B

where * is the surface site and the second step is irreversible. The rates for elemen-
tary reactions 1 and 2 in (3.26) are given by:

r1 = r + 1 − r − 1 = k + 1 CA θv − k − 1 θA , r2 = r + 2 = k + 2 θA (3:27)

where θv and θA is the fraction (coverage) of vacant sites and adsorbed A, respec-
tively. The overall rate of the product formation is equal to the rate of the second
step in the reaction mechanism. In the steady-state hypothesis the rate is:

dθA
rA* = = r+1 − r−1 − r+2 ≈ 0 (3:28)
dt

leading to the dependence of the coverage of adsorbed A on the fraction of vacant

sites:

k + 1 CA
θA = θv (3:29)
k−1 + k+2

Taking into account the balance equation θA + θv = 1 which implies that the surface
sites are either vacant or occupied by adsorbed A, one gets:

k + 1 CA
k − 1 + k+2 
θA =  (3:30)
k + 1 CA
1+
k−1 + k+2

With this balance equation it is implicitly considered that the rate is proportional to
the amount of catalytic sites.
The rate of product B formation in the steady-state is equal to the rate of each
step, which could be taken as the rate of the second step for simplicity of derivation:

k + 2 k + 1 CA
k + k+2 
− rA = rB = k + 2 θA =  − 1 (3:31)
k + 1 CA
1+
k−1 + k+2
 3.3 Kinetics 239

Quite often, adsorption and desorption [for example, step 1 in eq. (3.26)] are con-
sidered to be much faster than the reaction step. It implies that k–1 > k+2 and
subsequently:

k + 2 k + 1 CA
k−1 k + 2 KA CA
r= = (3:32)
k + 1 CA 1 + KA CA
1+
k−1
 
ð − ΔHads Þ
with k + 1 =k.1 = K1 = KA = KA, 0 exp , where (–△Hads) is the heat of ad-
RT
sorption of a molecule, A, on the active catalyst surface. The exothermic nature of
adsorption (△H < 0) implies that KA decreases with temperature.
When the rates of a step in both directions are significantly higher than the
rates of some other steps, such step is considered to be at quasi-equilibrium
(Fig. 3.5). This is because r+/r+ ≈ 1. In heterogeneous catalysis, adsorption and de-
sorption are usually considered to be much faster than the steps of surface reac-
tions, making it possible to apply adsorption isotherms (for example Langmuir
isotherm) directly to these steps.

Quasi-equilibrium

Rate determining step

Fig. 3.5: Illustration of quasi-equilibrium and rate-determining step concept.

Two extreme limits are possible for eq. (3.32), the first when concentrations are low
and KACA ≪ 1, then the denominator is close to unity, that is, (1 + KACA)~1 The rate
is then proportional to the reactant concentration: r~k+ 2 KACA, as a first-order reac-
tion with respect to concentration of A with an apparent first-order rate constant, k’
= k+2KA (Fig. 3.6). This is the low concentration or pressure (or weak binding i.e.,
small K) limit: under these conditions the steady-state surface coverage, θ, of the
reactant molecule is very small.
The other limit is at high concentrations, KACA ≫ 1; then (1 + KACA)~KACA and the
rate is r~k+2, giving a zero order reaction with respect to the concentration of A. This is
the high concentration (or strong binding, i.e., large K) limit. Under these conditions
the steady-state surface coverage, θ, of the reactant molecule is almost unity. The rate
shows the same concentration variation as surface coverage in the Langmuir adsorp-
tion isotherm.
240 3 Engineering reactions

Rate High concentration

pressure limit: rate ~ k+2

Low concentration/pressure limit:

rate ~ k+2KA C (k+2KA P)

Concentration/pressure

Fig. 3.6: Concentration/Pressure dependence of reaction rate (eq. 3.32).

The law of mass action applied to processes on surfaces implies that rate ex-
pressions and adsorption isotherms contain concentrations. For gas-phase reac-
tions, it is tempting to utilize partial pressures rather than the concentrations as
shown in Fig. 3.6. Strictly speaking from the ideal gas law PV = nRT and thus C = P/
RT giving instead of eq. (3.32)

k + 2 KA PA =RT k + 2 K ′A PA
r= = = (3:33)
1 + KA PA =RT 1 + K ′A PA

The adsorption equilibrium constant defined through pressures KA ′ is thus different

from the adsorption constant expressed through concentrations.
A gas-phase reaction can proceed according to the Langmuir-Hinshelwood
mechanism:

Agas + * = A*ads ðquasi−equilibriumÞ

Bgas + * = B*ads ðquasi−equilibriumÞ
(3:34)
A*ads + B*ads ) Cgas + 2*

A+B=C

In this mechanism the surface reaction between the two adsorbed species competing
on the surface is the rate-determining step. The reaction rate can be easily written as:

KA CA K B CB kKA CA KB CB
r = kθA θB = k = (3:35)
1 + KA CA + KB CB 1 + KA CA + KB CB ð1 + KA CA + KB CB Þ2

When KACA ≪ 1 and KB CB ≪ 1 both coverage θA and θB are very low, and the rate is:

r = kKA CA KB CB (3:36)
 3.3 Kinetics 241

meaning that the reaction is first-order in both reactants. If KACA≪1 ≪KBCB, the
coverage of A is low, while the surface is predominantly covered by B (θA→0, θB→1),
and the rate is becoming first-order in A, but negative first-order in B:

kKA CA
r= (3:37)
KB CB

Depending upon the concentration (partial pressure) and binding strength of the
reactants, a given model for the reaction scheme can give rise to a variety of appar-
ent kinetics. This highlights the dangers inherent in the reverse process – namely
trying to use kinetic data to obtain information about the reaction mechanism.
If the product C adsorbs on the catalyst surface the rate equation is becoming
slightly more complicated, as the coverage of C should be included in the denominator:

kKA CA KB CB
r= (3:38)
ð1 + KA CA + KB CB + KC CC Þ2

If the reaction is reversible, with the rate determining step A*ads + B*ads , C*ads + *
then:

k + KA CA KB CB k − KC CC
r = k + θA θB − k − θC θ* = 2
−
ð1 + KA CA + KB CB + KC CC Þ ð1 + KA CA + KB CB + KC CC Þ2
(3:39)
k + KA CA KB CB 1 − k − KC CC  
k + KA CA KB CB 1 CC
= = r + 1 −
ð1 + KA CA + KB CB + KC CC Þ2 K CA CB

Another possibility is that a molecule from the gas-phase reacts with an adsorbed
molecule without adsorbing itself on the surface (for example, the Eley-Rideal pro-
cess). For the mechanism:

AðgÞ + * = A*ðadsÞ ðquasi−equilibriumÞ

BðgÞ + *AðadsÞ ) C*ðadsÞ

(3:40)
C*ðadsÞ = CðgÞ + *ðquasi−equilibriumÞ

A+B=C

the reaction rate for an irreversible reaction is given by:

k + KA CA CB k + K ′A PA PB
r = k + θ A CB = = (3:41)
1 + KA CA + KC CC 1 + K ′A PA + K ′C PC

For low pressures K’APA≪1 and the rate is first-order in both reactants, similar to
the Langmuir-Hinshelwood mechanism. However, the rate cannot have negative
order dependence as followed from Langmuir-Hinshelwood treatment.
242 3 Engineering reactions

Monomolecular isomerization in the adsorbed layer is a simplified case of the

mechanism (3. 40)

AðgÞ + * = A*ðadsÞ ðquasi−equilibriumÞ

*AðadsÞ ) C*ðadsÞ
(3:42)
C*ðadsÞ = CðgÞ + *ðquasi−equilibriumÞ

A+B=C

With the corresponding rate equation

k + KA CA
r = k + θA = (3:43)
1 + KA CA + KC CC

The energy diagram for mechanism (3.42) is shown in Fig. 3.7, illustrating that the
observed (apparent activation energy) is different from the activation energy of the
isomerization in the adsorbed state per se. The concept of activation energy will be
elaborated in section 3.4.4.

A#g

Ea, obs = Ea+ΔHR

A#*

Ea
A(g)+*
ΔHA ΔHR
A*
C(g)+*
ΔHB

C*

Fig. 3.7: Energy diagram for the reaction of A to C. For illustration purposes, a noncatalytic reaction
is also shown.

As mentioned earlier, a standard choice in kinetic modeling is to use concentrations,

although for gas-phase kinetics it is common to apply partial pressures, often neglect-
ing a more complex behavior of the adsorption constant on temperature. This ap-
proach is straightforward for ideal gases. In the case of high pressures and deviations
from ideal gas law it is, however, recommended that gas-phase fugacities in rate
 3.3 Kinetics 243

equations are used instead of partial pressures. For liquid-phase systems concentra-
tions are usually used, although for strongly non-ideal solutions they should be re-
placed by activities, which could be rather complicated.
At this point, some generalization of kinetic models is possible. For a reaction
A + B = C + D the reaction rate takes a form:

kinetic factor* driving force

r= (3:44)
ð adsorption termÞn

where the adsorption term includes KACA, etc. for molecular adsorption or
(KACA)1/2 for adsorption with dissociation and the power in the denominator cor-
responds to the number of species in the rate-determining steps, while the driv-

ing force is ð1 − PC PD =Keq PA PB Þ or 1 − CC CD =Keq CA CB .
The classical application of Langmuir kinetics to uniform surfaces can be extended
to the reactions that occur on two types of sites in close proximity. The reaction pro-
ceeds via interactions between species adsorbed on such distinct sites. Examples of
such reactions are various oxidation reactions on metal oxides, which contain two
types of sites, metal atoms sites and oxygen atom sites, each having a specific function.
Another example is hydrogenolysis of haloarenes where spillover hydrogen reacts with
an organic molecule adsorbed on the metal surface. The mechanism is

AðgÞ + * = A*ðadsÞ ðquasi−equilibriumÞ

BðgÞ + *′ = B*′ðadsÞ ðquasi−equilibriumÞ

(3:45)
A*ðadsÞ + B*′ðadsÞ ) CðgÞ + * + *′rate−determining step

A+B=C

Similarly, it can be written for the reaction rate:

KA CA KB CB
r = kθA θ′B = k (3:46)
1 + KA CA 1 + KB CB

Dual site formalism can be easily extended to other cases, such as dissociative ad-
sorption of for example B2 with stepwise addition where the first addition step is
rate limiting:
pffiffiffiffiffiffiffiffiffiffiffiffi
K A CA KB CB2
r=k pffiffiffiffiffiffiffiffiffiffiffiffi (3:47)
1 + KA CA 1 + KB CB2

Models with semi-competitive adsorption could be also applied. The principle of

semi-competitive adsorption is that the large organic molecules always leave some
empty space in between them even at complete coverage, i.e., vacant interstitial
sites, where the smaller molecule (typically hydrogen and oxygen) can adsorb.
Compared to classical models, the semi-competitive adsorption model includes one
244 3 Engineering reactions

additional parameter, namely the maximum coverage of the organic molecule. Some
examples of rate equations for various catalytic mechanisms are given in Tab. 3.1.

Tab 3.1: Rate expressions for catalytic gas-phase reactions with various mechanisms (rds – rate-
determining step). In the rate expressions, pressures are used instead of concentrations.

Surface reactions Rate expression

1.A + * = A* k2 K1 PA −
PB
Keq
2. A* ) B* r.d.s. rA = −
1 + K1 PA + K3 1 PB
3. B* = B + *

1. A + * ) A* r.d.s. k1 PA −
PB
Keq
2. A* = B* rA =
1 + ð1 + 1=K2 ÞK3− 1 PB
3. B* = B + *

1.A + * = A* k3 K1 K2 PA −
PB
Keq
2.A* = B* rA =
1 + ð1 + K2 ÞK1 PA
3.B* ) B + * r.d.s.

1.A + * = A k3 K1 PA K2 PB
rA = 2
2.B + * = B ð1 + K1 PA + K2 PB + K4− 1 PC Þ
3.A* + B* ) C* + * r.d.s.
4.C* = C + *

1.A + * = A k2 K1 PA PB
rA =
2.A* + B(g) ) C* + D(g) r.d.s. 1 + K1 PA + K3− 1 PC
3.C* = C + *
pffiffiffiffiffiffiffiffiffiffiffiffi
1.A2 + *’ = 2A*’ k3 K1 PA2 K2 PB
2.B + * = B* rA = pffiffiffiffiffiffiffiffiffiffiffiffi 2
ð1 + K1 PA2 + K2 PB + K5− 1 PC Þ
3.A*’ + B* ) AB* + *’ r.d.s.
4.A*’ + AB* ) C* + *’ (fast)
5.C* = C + *

1.A + * = A* k2 K1 PA
rA = 
2.A* + * ) B* + * r.d.s. 1 + K1 PA + K3− 1 PB
2

3.B* = B + *

The two-step sequence is a frequent case in heterogeneous catalysis and can be pre-
sented in the form:
1. * + A1 $ *I + B1
2. *I + A2 $ * + B2 (3:48)
A1 + A2 $ B1 + B2

where *I is the surface intermediate. The mechanism assumes binding of the reac-
tant A1 in the first step, giving an adsorbed intermediate I and releasing the product
 3.3 Kinetics 245

B1. In the second step the surface intermediate reacts with the reactant A2 leading
to a vacant site and the product B2. The rate of the steps is given by:

r1 = k + 1 ½* CA1 , r − 1 = k − 1 ½*ICB1 , r2 = k + 2 ½*ICA2 , r − 2 = k − 2 ½* CB2 (3:49)

The steady-state approximation gives the relationship between * and *I:

k + 2 CA2 + k − 1 CB1
*
*I = (3:50)
k + 2 CA2 + k − 1 CB1

The conservation equation for the catalytic sites is:

½* + ½*I = ½* = 1 (3:51)

Applying the steady-state principle for mechanism (3.48) and eliminating the cover-
age of the intermediate I, it can be demonstrated that the reaction rate is equal to:

k + 1 CA1 k + 2 CA2 − k − 1 CB1 k − 2 CB2 ω + 1ω + 2 − ω − 1ω − 2

r= = (3:52)
k + 1 CA1 + k + 2 CA2 + k − 1 CB1 + k − 2 CB2 ω + 1 + ω + 2 + ω − 1 + ω − 2

where ω+1, etc. are frequencies of steps (ω+1 = k+1CA1). In eq. (3.52) the rate is calcu-
lated per catalyst mass. Otherwise the rate expression contains the catalyst
concentration:

k + 1 CA1 k + 2 CA2 − k − 1 CB1 k − 2 CB2

r = Ccat (3:53)
k + 1 CA1 + k + 2 CA2 + k − 1 CB1 + k − 2 CB2

Equation (3.53) contains liquid-phase concentrations, while for gas-phase reactions,

partial pressures are often used. The water-gas shift (WGS) reaction can be, for
example, represented by the two-step sequence:

* + H2 O $ *O + H2
*O + CO $ * + CO2 (3:54)
H2 O + CO = H2 + CO2

The rate expression for WGS reaction can be written:

k + 1 PH2 O k + 2 PCO − k − 1 PH2 k − 2 PCO2

r= (3:55)
k + 1 PH2 O + k + 2 PCO + k − 1 PH2 + k − 2 PCO2

In the treatment above it was assumed that the number of adsorption sites on
which a molecule is adsorbed is equal to unity, which is not necessarily the case,
especially in organic catalysis (Fig. 3.8).
246 3 Engineering reactions

O
H3CO
O
HO
HO

OCH3
HO

Fig. 3.8: The lignin hydroxymatairesinol (HMR) depending on the adsorption structure covers ~
10–20 Pd surface atoms.

As an example, the hydrogenolysis reaction of a hydrocarbon (HC) will be con-

sidered here. For a uniform surface (where all the sites are equal and there is no inter-
action between adsorbed species), the following mechanism could be proposed:

1. HC + z*Ð HCz*
2. H2 + 2*Ð 2H* (3:56)
3. HCz* + H* ! Products

In (3.56) z is the number of free neighbor potential sites required for hydrocarbon
adsorption, steps 1 and 2 represent adsorption steps, while step 3 reflects carbon-
carbon bonds rupture, which could be followed by other fast steps leading to final
reaction products. Quasi-equilibria for the adsorption steps of hydrocarbon and hy-
drogen give:

θHC θ2H
K1 = z , K2 = (3:57)
PHC θ* PH2 θ2*

where θH denotes the coverage of hydrogen, θHC the coverage of hydrocarbon, θ*

the fraction of vacant sites and PHC and PH2 are the partial pressures of hydrocarbon
and hydrogen, respectively. The rate of step 3 is expressed by:

r1 = k3 θHC θH (3:58)

which when combined with expressions for adsorption gives:

r1 = k3 K1 K20.5 PHC PH0.52 θ*z + 1 (3:59)

The site balance equation:

1 = θHC + θH + θ* = K1 PHC θz* + K20.5 PH0.52 θ* + θ* (3:60)

should be solved numerically (except for z = 1) together with eq. (3.59) during pa-
rameter estimation and comparison with experimental data.
 3.4 Kinetics of complex reactions 247

Although the steady-state regime can only be realized in open systems, the same
steady-state approximation is usually applied for closed systems (batch reactors), as
relaxation times (time to reach steady-state) are shorter than times of one turnover.

3.4 Kinetics of complex reactions

3.4.1 Theory of complex reactions kinetics

Catalytic reactions are complex, which means that there is a set of different elementary
reactions occurring jointly. These reactions are related to each other, with some partici-
pating species in common. For example, hydrogenation of pentene involving the for-
mation of complexes with a catalyst (where z is the surface site, very often used along
with other signs to indicate surface sites, such as *) is described by the reaction:

1. H2 + 2z $ 2zH 1
2. C5 H10 + z $ zC5 H10 1
3. zC5 H10 + zH $ zC5 H11 + z 1
(3:61)
4. zC5 H11 + zH $ zC5 H12 + z 1
5. zC5 H12 $ z + C5 H12 1

C5 H10 + H2 = C5 H12

The latter equation is the overall chemical equation, which includes only reactants and
reaction products, while the chemical equations of elementary reactions grouped into
steps in (3.61) include surface intermediates that do not appear in the overall equations.
For reversible reactions steps consist of a pair (forward and reverse) of elementary reac-
tions, while only one elementary step is needed for an irreversible reaction. If the rates of
some elementary reactions are sufficiently large compared to the rate of the complex re-
action as a whole, several elementary reactions are grouped into one more complex step.
Overall reaction equations are linear combinations of chemical equations of steps.
They are obtained by the addition of chemical equations of steps multiplied by stoi-
chiometric numbers (positive, negative or zero). The numbers, introduced by Horiuti,
must be chosen in such a way that the overall equations contain no intermediates.
Stoichiometric (Horiuti) numbers could be also fractional, for example in SO2
oxidation:

2z + O2 $ 2zO 1 2z + O2 $ 2zO 0.5

zO + SO2 $ z + SO3 2 zO + SO2 $ z + SO3 1

O2 + 2SO2 = 2SO3 1=2 O2 + SO2 = SO3

248 3 Engineering reactions

The overall equation of the scheme on the right side is obtained from the overall
equation of the left one when multiplied by 0.5. Such an operation cannot be per-
formed for the equations of steps, because reaction z + 1/2O2 ↔ zO cannot occur.
The overall chemical reaction 1/2O2 + SO2 = SO3 only describes stoichiometry, but
not the reaction mechanism. The reaction route after Horiuti is defined as a set of
stoichiometric numbers of the steps that produces an overall reaction equation.
Routes must be essentially different and it is impossible to obtain one route from
another through multiplication by a number, although their respective overall
equations can be identical.
For an example of this, the synthesis of water from elements will be considered:

Nð1Þ Nð2Þ
1. O2 + 2z ! 2zO 1 1
2. H2 + 2z $ 2zH 0 2
3. zO + zH ! zOH + z 0 2 (3:62)
4. zOH + zH ! H2 O + 2z 0 2
5. zO + H2 ! H2 O + z 2 0

Nð1Þ , Nð2Þ :O2 + H2 = 2H2 O

The mechanism consists of two routes, the first once corresponding to an Eley-
Rideal mechanism, when oxygen is adsorbed with dissociation and reacts with di-
hydrogen from the fluid phase (steps 1 and 5, respectively). The second route com-
prises dissociative adsorption of dioxygen and dihydrogen (steps 1 and 2) and the
subsequent two consecutive surface reaction steps (steps 3 and 4). Such reactions
are usually considered to be of a Langmuir-Hinshelwood type.

The research proposal I wrote in the late 1980s contained the term “Langmuir-Hinshelwood kinet-
ics”. Prof. M. Temkin asked if I had ever read the original publications of Irving Langmuir and
advised me to do so when the reply was negative. It then became apparent to me that Langmuir
in fact considered both cases (for example for CO oxidation): the co-adsorption of both reactants
(called nowadays Langmuir-Hinshelwood mechanism) and the reaction of the second compound
from the fluid phase, while the first is adsorbed (the Eley-Rideal mechanism).

The route N(2) in eq. (3.62) cannot be obtained by multiplication of N(1) with a cer-
tain number, which means that the routes N(1) are N(2) different (as apparent from
their different mechanism), although their overall equations are identical.
The stoichiometric number of a step s in a route N(3) VS(3) could be defined as:
ð3Þ ð1Þ ð2Þ
vS = C1 vS + C2 vS (3:63)
 3.4 Kinetics of complex reactions 249

where C1 and C2 are some arbitrary numbers. Any linear combination of routes of a
given reaction (infinite combinations) will also be a route of the reaction:
ð3Þ ð1Þ ð2Þ
NS = C1 NS + C2 NS (3:64)

Reaction routes thus form the vector space, in terms of linear algebra. A set of such
routes, when no route can be represented as a linear combination of others, is called
the basis of routes, which are linearly independent. Although the basis of routes can
be chosen in different ways the number of basis routes for a given reaction mecha-
nism is determined in a unique way, being the dimension of the space of the routes.
There are also relationships between intermediates that give balance or conserva-
tion equations. For heterogeneous catalytic reactions, a balance equation relates the
surface coverage of adsorbed species and vacant sites. If there are two types of sites
on the surface there would be two balance equations. Another example of the bal-
ance (conservation or link) equation could be overall neutrality of the surface if ad-
sorbed intermediates are charged species.
Following Horiuti the number of basic routes P is determined by:

P=S−I (3:65)

where S is the number of stages and I is the number of independent intermediates

in the sense used in the Gibbs phase rule. A similar expression was proposed by
Temkin:

P=S+W −J (3:66)

where P is the number of routes, W is the number of balance equations, J is the num-
ber of intermediates (including vacant sites) and S is the number of steps.
The rate along the basic route N(p) is denoted as r(p). According to its definition
there are vðpÞ
s r
ðpÞ
runs of the sth stage per unit time per unit reaction space. Here vðpÞ
s
is the stoichiometric number of the sth step along the route N(p). For a one-route
mechanism for a steady-state reaction the overall reaction rate is expressed as:
r+s − r−s r+1 − r−1 r+2 − r−2 r+3 − r−3 r+4 − r−4
r= = = = = = ... (3:67)
vs v1 v2 v3 v4

For the mechanism (3.62) the derivation could be simplified assuming equilibrium
adsorption of dioxygen and dihydrogen with dissociation, giving:
qffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffi
θo = K1 PO2 θv , θH = K2 PH2 θv (3:68)

where θv is the fraction of vacant sites. From the steady-state for zOH (r3 = r4):

k3 k3 qffiffiffiffiffiffiffiffiffiffiffiffi
θOH = θO = K1 PO2 θv (3:69)
k4 k4
250 3 Engineering reactions

Taking into account the balance equation, an explicit expression for the fraction of
vacant sites could be obtained as:
1
θv = pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi (3:70)
k3
1 + K1 PO2 1 + k4 + K2 PH2

The reaction rate is equal to the sum of the rates along two routes:

pffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffi
r3 r5 k3 K1 PO2 K2 PH2
r = rI + rII = + = pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi 2
2 2 k
2 1 + K1 PO2 1 + k3 + K2 PH2 (3:71)
4
pffiffiffiffiffiffiffiffiffiffiffiffi
k5 PH2 K1 PO2
+ pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
k
2 1 + K1 PO2 1 + k3 + K2 PH2
4

In (3.62) there were steps with a stoichiometric number equal to zero, which means
that these steps do not take part in a particular route. In fact a step can have a stoi-
chiometric number equal to zero along all the routes. For example when solvent Y
adsorbs on the surface a mechanism can take the form:

1. z + A ≡ zA 1
2. zA , zB 1
3. zB ≡ z + B 1 (3:72)
4. z + Y ≡ zY 0
A=B

The stoichiometric number of step 4 in eq. (3.72) is zero, which is logical as the sol-
vent does not take part in the chemical transformations. At the same time, the solvent
concentration can influence the reaction rate by blocking some catalytic sites, which
will be otherwise accessible to reactants, resulting in the following rate expression:

k + 2 KA CA − k − 2 KB CB
r= (3:73)
1 + KA CA + KB CB + Ky Cy

One of the papers I submitted to the Journal of Catalysis in the early 1990s contained an analy-
sis of a mechanism similar to eq. (3.72). The referee responded that the “stoichiometric number
in chemistry can never be zero.” Apparently the reviewer, who was supposed to be a specialist
in catalytic kinetics, was not at all familiar with the theory of complex reactions, developed by
Horiuti and Temkin in the 1960s!

A reaction mechanism can be rather complex, comprising several routes and types
of sites. For example, a mechanistic model describing the experimental data of
 3.4 Kinetics of complex reactions 251

L-arabinose oxidation over Au/Al2O3 catalyst with oxygen in the aqueous phase [1]
can be considered as:

HO HO OH OH
O O2 O H2O
HO HO O
Au/AI2O3 Au/AI2O3 HO O
OH OH OH
OH OH
(3.74)

This was based on the concept of oxidative dehydrogenation:

Nð1Þ Nð2Þ Nð3Þ Nð4Þ

1. A + * ≡ A* 0 0 0 0
2. A + *′ ≡ A*′ 2 0 0 0
3. O2 + 2* ≡ 2O* 1 0 0 0
4. O* + *′ ≡ O*′ + * 2 1 1 1
5. A*′ + O*′ ! B*′ + H2 O + *′ 2 0 0 0
6. B*′ + OH *′ ! D*′ + *′
−
0 1 0 0
6a. D*′ + H + ! ðfastÞC + *′ 0 1 0 0
− −
7. O* + H2 O + 2e ! 2OH + * 0 0 −1 1
8. OH − + * ! O* + H + + 2e − 0 1 0 2
9. O*′ + H2 O + 2e ! 2OH + *′
− −
0 1 1 1
10. OH − + * ≡ OH − * 0 0 0 0
11. OH + *′ ≡ OH *′
− −
0 1 0 0
12. B*′ ≡ B + *′ 2 −1 0 0
Nð1Þ : 2A + O2 = 2B + 2H2 O, Nð2Þ : B + H2 O = C, Nð3Þ : 0 = 0, Nð4Þ : 2H2 O = 2OH − + 2H + .
(3:75)

where * is the active sites on the edges of the catalyst, * ′ is the active sites on the
faces, A is L-arabinose, B is arabinolactone, C is arabinonic acid and D is the intermedi-
ate. The elementary steps above can be described by four reaction routes. It was as-
sumed that L-arabinose can adsorb on both the faces and the edges of gold clusters
(steps 1 and 2, respectively). Oxygen is adsorbed on the edges with dissociation (steps
3), thereafter migrating to the faces (steps 4). The sugar oxidation involves dehydroge-
nation of the adsorbed L-arabinose to arabinolactone and oxygen reduction in step 5.
The product, arabinonic acid, is formed in step 6 and 6a via the intermediate species
252 3 Engineering reactions

D, while step 7 and 8 account for electron transfer, which involves oxygen adsorbed on
the edges. Finally, as the oxidation rate was found to be dependent on pH, adsorption
of OH– on the faces was also included in the model (step 11). Electron transfer involving
adsorption of OH– on the edges was also included in the model (step 9). In addition,
step 10 accounts for direct adsorption of OH– on the edges without electron transfer.
The oxidative dehydrogenation of L-arabinose takes place in two steps (5 and 6) via the
intermediate arabinolactone B. The intermediate product, arabinolactone, has been
found only in small amounts in the reaction mixture (step 12). It is assumed to react
much faster than L-arabinose and, as a reactive intermediate, it will have very low sur-
face coverage.
One of the routes (i.e., N(3)) is a so-called empty route, as it does not result in
any chemical transformation.
According to the theory of complex reactions developed by Horiuti and Temkin,
the rate of the s step is obtained by summation over all P basic routes, hence:
X
P
rjsj = rs − r − s = vðpÞ
s r
ðpÞ
(3:76)
p=1

where rs and r–s are the rates of the forward and reverse elementary reactions com-
prising the stage, r(p) is the rate along the basic route N(p), vðpÞ
s is the stoichiometric
(p)
number of the sth step along the route N . For example the rate of step 5 in (3.75) is:

rj5j = r5 − r − 5 = 2*rð1Þ + 0*rð2Þ + 0*rð3Þ + 0*rð4Þ (3:77)

which looks trivial at first. However, such analysis is very helpful for multiroute
mechanisms.
If for some steps rs ≈ r − s  rjsj the steps are considered as “fast” (for example,
steps 1, 2, 3, 4 in eq. 3.75) or quasi-equilibrium steps (as illustrated in Fig. 3.4).
Strictly equilibrium steps are also possible, when the stoichiometric numbers of a
step for all basic routes are zero, then rs = r–s. Reversible adsorption of a solvent
[eq. (3.72)] may serve as an example.
For non-linear steps there could be several solutions for (3.76). Sometimes the
steady- state solution cannot be reached and the reaction system starts to oscillate.
For a one-route mechanism for a steady-state reaction the overall reaction rate
follows immediately from (3.75):
r+s − r−s r+1 − r−1 r+2 − r−2 r+3 − r−3 r+4 − r−4
r= = = = = = ... (3:78)
vs v1 v2 v3 v4

General expressions for the reaction rate can be constructed knowing the expres-
sions for the rate of steps if the following equation is used:
Y
s Y
s
r+i − r − i = ðr + 1 − r − 1 Þr + 2 . . . r + s + r − 1 ðr + 2 − r − 2 Þr + 3 . . . r + s + . . .
i=1 i=1 (3:79)
+ r − 1 r − 2 . . . ðr + s − r − s Þ
 3.4 Kinetics of complex reactions 253

The numbers of steps in (3.79) are chosen in an arbitrary way. For a four-step reac-
tion one arrives at:
Y
4 Y
4
r+i − r − i = ðr + 1 − r − 1 Þr + 2 . . . r + 4 + r − 1 ðr + 2 − r − 2 Þr + 3 . . . r + 4 + . . .
i=1 i=1 (3:80)
+ r − 1 r − 2 . . . ðr + 4 − r − 4 Þ

which could be transformed into:

r + 1 r + 2 r + 3 r + 4 − r − 1 r − 2 r − 3 r − 4 = rj1j r + 2 r + 3 r + 4 + r − 1 rj2j r + 3 r + 4 + . . . r − 1 r − 2 r − 3 rj4j (3:81)

As steady-state the global rate is expressed by the rate of a step and its stoichiomet-
ric number is then:

r + 1 r + 2 r + 3 r + 4 − r − 1 r − 2 r − 3 r − 4 = ðv1 rÞr + 2 r + 3 r + 4 + r − 1 ðv2 rÞr + 3 r + 4

(3:82)
+ . . . + r − 1 r − 2 r − 3 ðv3 rÞ

giving the rate expression as:

r + 1r + 2r + 3r + 4 − r − 1r − 2r − 3r − 4
r= (3:83)
v1 r + 2 r + 3 r + 4 + r − 1 v2 r + 3 r + 4 + r − 1 r − 2 v3 r + 4 + r − 1 r − 2 r − 3 v3

which, in more general form, is as follows:

Q
s
r+i
i=1
r = r+ − r− =
v1 r + 2 . . . r + s + r − 1 v2 r + 3 . . . r + s + r − 1 r − 2 . . . vs
Q s
r−i
i=1 (3:84)
−
v1 r + 2 . . . r + s + r − 1 v2 r + 3 . . . r + s + r − 1 r − 2 . . . vs
Q
s Qs
r+i − r−i
i=1 i=1
=
v1 r + 2 . . . r + s + r − 1 v2 r + 3 . . . r + s + r − 1 r − 2 . . . vs

The ratio of rates in the forward and reverse directions is thus expressed as:
r − r − 1r − 2r − 3 . . . r − s
= (3:85)
r+ r1 r2 r3 . . . rs

Note that such analysis is valid for linear mechanisms.

For a two-step catalytic sequence (eq. 3.48) and when numbering the step 1 as
the first one and the step 2 as the second, it follows from (3.84) that:

k + 1 ½* CA1 k + 2 ½*I CA2 − k − 1 ½*I CB1 k − 2 ½* CB2

r= (3:86)
k + 2 ½*ICA2 + k − 1 ½*I CB1
254 3 Engineering reactions

or after rearranging:

k + 2 CA2 + k − 1 CB1
½* = r (3:87)
k + 1 CA1 k + 2 CA2 + k − 1 CB1 k − 2 CB2

If numbering step 1 as the second one and step 2 as the first, another expression for
the rate and coverage of the intermediate I follows from eq. (3.84) as:

k + 2 ½*ICA2 k + 1 ½*CA1 − k − 2 ½*CB2 k − 1 ½*ICB1

r= ; (3.88)
k + 1 ½*CA1 + k − 2 ½*CB2
k + 1 CA1 + k − 2 CB2 (3.89)
½*I = r
k + 1 CA1 k + 2 CA2 − k − 1 CB1 k − 2 CB2

From the balance equation:

k + 1 CA1 + k + 2 CA2 + k − 1 CB1 + k − 2 CB2

½*tot = ½*I + ½* = r (3:90)
k + 1 CA1 k + 2 CA2 − k − 1 CB1 k − 2 CB2

the reaction rate can be easily obtained as:

k + 1 CA1 k + 2 CA2 − k − 1 CB1 k − 2 CB2 ω + 1ω + 2 − ω − 1ω − 2

r= = (3:91)
k + 1 CA1 + k + 2 CA2 + k − 1 CB1 + k − 2 CB2 ω + 1 + ω + 2 + ω − 1 + ω − 2

For a more complex case of three steps:

1. * + A1 $ *I + B1
2. *I $ *I1
(3:92)
3. *I1 + A2 $ * + B2
A1 + A2 $ B1 + B2

the reaction rate can be derived either using the steady-state approximation leading to:
ω + 1ω + 2ω + 3 − ω − 1ω − 2ω − 3
r=
ω + 2ω + 3 + ω − 3ω + 2 + ω − 3ω − 2 + ω + 3ω + 1 + ω − 1ω + 3 + ω − 1ω − 3 + ω + 1ω + 2 + ω − 2ω + 1 + ω − 2ω − 1
(3:93)

or directly from (3.84) when changing the counting order in the reaction mecha-
nism (3.92) for steps 1–3. The terms in the denominator for (3.84) are obtained
when counting starts from the first, second or the third step, respectively:

s = 1, r + 2 r + 3 ;r − 3 r + 2 ;r − 3 r − 2
s = 2, r + 3 r + 1 ;r − 1 r + 3 ;r − 1 r − 3
s = 3, r + 1 r + 2 ;r − 2 r + 1 ;r − 2 r − 1
 3.4 Kinetics of complex reactions 255

which in combination with other terms gives eq. (3.93).

A Mars-van-Krevelen catalytic oxidation mechanism involving oxygen from the
lattice can be used to illustrate the three-step mechanism:

1. * + O2 ! *O2
2. *O2 + R ! RO + *O
(3:94)
3. *O + R ! * + RO
2R + O2 $ 2RO

For the mechanism with three irreversible steps the equation is simply:
rR ω + 1ω + 2ω + 3
r O2 = =
2 ω + 2ω + 3 + ω + 3ω + 1 + ω + 1ω + 2
(3:95)
k + 1 k + 2 k + 3 PO2 PR
=
k + 2 k + 3 PR + k + 1 ðk + 2 + k + 3 ÞPO2

In a similar fashion using pressure instead of concentrations for gas-phase reac-

tions the rate equation for a four-step linear mechanism (simplified mechanism for
methane steam reforming):

1. CH4 + z , H2 + CH2 z
2. H2 O + CH2 z , zCHOH + H2
3. zCHOH , H2 + zCO (3:96)

4. zCO , z + CO
CH4 + H2 O = CO + 3H2

takes the form:

k + 1 PCH4 k + 2 PH2 O k + 3 k + 4 − k − 1 PH2 k − 2 PH2 k − 3 PH2 k − 4 PCO

r= (3:97)
D
with:

D = k + 2 PH2 O k + 3 k + 4 + k − 1 PH2 k + 3 k + 4 + k − 1 PH2 k − 2 PH2 k + 4 + k − 1 PH2 k − 2 PH2 k − 3 PH2 +

k + 1 PCH4 k + 2 PH2 O k + 3 + k − 4 PCO k + 2 PH2 O k + 3 + k − 4 PCO k − 1 PH2 k + 3 +

k − 4 PCO k − 1 PH2 k − 2 PH2 + k + 4 k + 1 PCH4 k + 2 PH2 O + k − 3 PH2 k + 1 PCH4 k + 2 PH2 O + ð3:98Þ

k − 3 PH2 k − 4 PCO k + 2 PH2 O + k − 3 PH2 k − 4 PCO k − 1 PH2 + k + 3 k + 4 k + 1 PCH4 +

k − 2 PH2 k + 4 k + 1 PCH4 + k − 2 PH2 k − 3 PH2 k + 1 PCH4 + k − 2 PH2 k − 3 PH2 k − 4 PCO

The hydrogenation of benzene is an example of a single-route reaction mechanism

with five linear steps:
256 3 Engineering reactions

1. C6 H6 + * , C6 H6 *
2. C6 H6 * + H2 , C6 H8 *
3. C6 H8 * + H2 , C6 H10 *
(3:99)
4. C6 H10 * + H2 , C6 H12 *
5. C6 H12 * , C6 H12 + *

C6 H6 + 3H2 = C6 H12

The general form of the kinetic equation is:

ω + 1ω + 2ω + 3ω + 4ω + 5 − ω − 1ω − 2ω − 3ω − 4ω − 5
r= (3:100)
D

where the denominator D in its most general form consists of 25 terms. If reaction
occurs far from overall equilibrium and all the steps are irreversible then after sev-
eral simplifications:
ω + ω + 2ω + 3ω + 4ω + 5
r=
ω1 ω2 ω3 ω4 + ω1 ω2 ω3 ω5 + ω1 ω2 ω4 ω5 + ω1 ω3 ω4 ω5 + ω2 ω3 ω4 ω5
PC6 H6 PH2
= (3:101)
1
k+5 PC6 H6 PH2 + 1
k+4 + 1
k+3 + 1
k+2 PC6 H6 + 1
k+1 PH2

with the reaction order in hydrogen and benzene varying between zero and unity.
In heterogeneous catalysis the reaction mechanism is often rather complex and
cannot be represented by a one-route multistep reaction sequence, as there are sev-
eral routes leading to a variety of products.
If the reaction is described by P routes the rate expression is:
!
XP ðpÞ ðpÞ ðpÞ
r − 1 r − 2 r − 3 . . . r − s − 1 vs r − 1r − 2r − 3 . . . r − s
ðpÞ v1 r − 1 v2
r + + ... + =1−
p=1
r +1 r r
+1 +2 r r
+1 +2 +3 r . . . r +s r + 1r + 2r + 3 . . . r + s
(3:102)

3.4.2 Relationship between thermodynamics and kinetics

Linear free energy relationships (LFER) introduced by Brønsted are widely used in
homogeneous and heterogeneous catalytic reactions, relating reaction constants k
with equilibrium constants K in a series of analogous elementary reactions:

k = gK α 0 < α < 1 (3:103)

 3.4 Kinetics of complex reactions 257

where g and α (the Polanyi parameter) are constants. Such relationships suppose a cer-
tain relation between reaction thermodynamics and kinetics, for example, between the
Gibbs free energy of activated complexes and the Gibbs energy of a reaction:

ΔG# ⁓αΔG (3:104)

Evans and Polanyi introduced this relationship to organic reactions, Semenov ex-
tended the application to chain reactions and Temkin applied LFER to heteroge-
neous catalysis. Equation (3.104) is often expressed through a relationship between
the energy of activation and the reaction enthalpy:

E = Eo + αΔH (3:105)

where Eo is a constant for a reaction family.

3.4.3 Non-ideal surfaces

In the treatment of catalytic kinetics the surface is often considered to be uniform.

However, adsorbed molecules change the structure of the surface layer, influencing
each other, and the surface sites are not equal in the ability to bind chemisorbed
molecules. This is reflected, for example, in the dependence of the heats of adsorp-
tion on coverage. The case of different adsorption sites is defined as biographical
non-uniformity while the mutual influence of the adsorbed species is considered as
induced non-uniformity, In the former case there is a certain distribution of proper-
ties, which can be either chaotic – when adsorption energy on a given site is inde-
pendent on the neighbor site – or discrete.
The physical nature of the biographical (intrinsic or a priori) non-uniformity
can be attributed to the difference in the properties of the different crystal faces and
the occurrence of dislocations, defects and other disturbances.
The special Temkin model of non-uniformity assumes that each site is charac-
terized by a definite adsorption energy of a particular substance and that there is a
certain distribution of the total number of sites with this energy. The values of ad-
sorption energy on different sites lie between a certain minimum and a certain max-
  
imal value ΔGa min ≤ ΔGa ≤ ΔGa max , and the number of sites dl with standard
  
Gibbs energy of adsorption within ΔGa and ΔGa + d ΔGa is given by an expo-
nential dependence:
ΔGa 
dl = Ce RΘ d ΔGa (3:106)

where C and Θ are constants. The values of Θ could be positive, negative or infinity,
reflecting cases of Freundlich, Zeldovich-Roginski and Temkin isotherms, respec-
tively. In case when Θ → ∞ instead of (3.106) the following distribution is valid:
258 3 Engineering reactions


dI = Cd ΔGa (3:107)

The sites are numbered in the sequence of increasing ξ (decreasing adsorption

strength). The relative number of sites is s = l/L, where L is the total number of sites
(0<s<1). The distribution of sites function φ(ξ)dξ, which is the number of sites with
ΔGa =RT within ξ + dξ, can be defined as:

ξ < ξ o :φðξÞ = 0; ξ o ≤ ζ ≤ ξ 1 :ϕðξÞ = A expðγξÞ;ξ > ξ o :φðξÞ = 0 (3:108)

where ξo is lowest value of ξ and ξ1 is the highest value of ξ, A is a constant, γ = T/Θ.

For even distribution, γ = 0, for example, φ(ξ) = A. The value of A is determined by
the normalized condition:
ð
∞ ξð1

φðξÞdξ = L or φðξÞdξ = L (3:109)

−∞ ξ

The expressions for the constant A are

γL L
γ≠0, A = ;γ = 0, A = (3:110)
eγξ 1 − eγξ 0 ξ1 − ξ0

For each surface site (there are overall L sites) the reaction rate for the two-step se-
quence using pressures, rather than concentrations is expressed in the same way as
for ideal surfaces:

1 k + 1 PA1 k + 2 PA2
ρ+ = (3:111)
L k + 1 PA1 + k − 1 PB1 + k + 2 PA2 + k − 2 PB2

The rate in the forward direction is the sum of all the rates on each catalytic site:
ξð1

r+ = ρ* φðξÞdξ (3:112)
ξ0

Changing the variable from ξ to λ (ξ = λ + ξ0) the reaction rate is:

ðf
γ k + 1 0 PA1 k + 2 0 PA2 eðn − mÞλ dλ
r+ = γf
(3:113)
e − 1 ðk + 1 PA1 + k − 2 0 PB2 Þe − mλ + ðk − 1 0 PB1 + k + 2 0 PA2 Þenλ
0
0

where n = β + γ; m = α – γ. After some rearrangements discussed in detail by Murzin

and Salmi [2] an expression for the reaction rate is obtained, which is valid in the
region of medium coverage:
 3.4 Kinetics of complex reactions 259

γ π k + 1 0 PA1 k + 2 0 PA2 − k − 1 0 PB1 k − 2 0 PB2

r= (3:114)
eγf − 1 sin mπ ðk + 1 0 PA + k − 2 0 PB Þm ðk − 1 0 PB + k + 2 0 PA Þ1 − m
1 2 1 2

For evenly non-uniform surfaces γ = 0, m = α and the expression for the rate is:

1 π k + 1 0 PA1 k + 2 0 PA2 − k − 1 0 PB1 k − 2 0 PB2

r= (3:115)
f sin απ ðk + 1 0 PA + k − 2 0 PB Þα ðk0 − 1 PB + k + 2 0 PA Þ1 − α
1 2 1 2

When both steps in the two-step sequence are irreversible a power law model is
obtained:

r = kPAn 1 PA1 −2 n (3:116)

where 0 ≤ n ≤ 1. A comparison between eq. (3.116) and the corresponding one for
uniform surfaces:

k + 1 PA1 k + 2 PA2
r= (3:117)
k + 1 PA1 + k + 2 PA2

shows the difference between the reaction kinetics on uniform and non-uniform
surfaces.
These two models give qualitatively different behavior at boundary values of
partial pressures. For example, according to an ideal surface model, at high partial
pressures the reaction rate obeys zero order kinetics, while at low pressures the re-
action order is equal to unity, in stark contrast to the mechanism for a non-ideal
surface, which does not show zero order dependence in a wider range of partial
pressures.
Temkin’s kinetic models for intrinsic non-uniform surfaces with essentially two
non-equilibrium steps were extended by Snagovskii and Avetisov to several types
of reaction mechanisms. Only linear steps were considered that occur either with or
without changes in the number of adsorbed species leading respectively to:

A1 + z $ zI1 + A2 or B2 + zI1 $ zI2 + B2 (3:118)

It was assumed that the rate constants of elementary steps do not depend on the pa-
rameter ξ. The rate constants of the steps in the forward direction (for example, ad-
sorption type) that correspond to the left side of eq. (3.118) contain the term e − αξ ,
while the desorption type constants have the term eβξ . The values of the transfer coef-
ficients α and β are considered to be the same and distribution functions are similar
for all surface species. The kinetic equation for a uniform patch is expressed by:

1 Ceðβ − αÞξ
ρ= (3:119)
L De − αξ + Heβξ
260 3 Engineering reactions

where L is the number of surface sites, and C, D and H are some functions of the
reactants partial pressures. The reaction rate on non-ideal surfaces is given by:

πγ C
r= (3:120)
ðeγf − 1Þ sin nπ Dm H n

with m = α – γ and n = β + γ. The values of C, D and H for the two-step sequences

are correspondingly:

C = ω + 2 D − ω − 2 H, D = ω + 1 + ω − 2 , H = ω − 1 + ω + 2 (3:121)

For a mechanism with three steps and two surface intermediates the parameters in
eq. (3.120) are:
 
ω−2
C = ω + 3 ðω + 1 + ω − 3 Þ − ω − 3 H, D = 1 + ðω − 3 + ω + 1 Þ,
ω+2
(3:122)
ω − 1ω − 2
H = ω+3 +
ω+2

The models for induced non-uniformity (presence of lateral interactions) usually

consider the repulsive interactions between nearest neighbors and attractive inter-
actions between next-nearest neighbors. Information on lateral interactions (which
are not prominent at low coverage) can be extracted from high-resolution scanning
tunneling microscopy data and analysis of desorption kinetics.
The simplest representation of adsorbate-adsorbate-adsorbent interactions is
the Fowler-Guggenheim isotherm, which differs from the Langmuir one in the expo-
nential term, accounting for the interactions:

θ vθ
aP = e (3:123)
1−θ

Using a similar concept the reaction rate for the region of medium coverage was
derived [3] as:

T ln U ′ k + 1 PA1 k + 2 PA2 − k − 1 PB1 k − 2 PB2

r=   (3:124)
η C k + 1 PA + k − 2 PB α k + 2 PA + k − 1 PB
2 1−α
1 2 2 1

 
where η is a constant and U ′ = k + 1 PA1 + k − 2 PB2 = k + 2 PA2 + k − 1 PB1 . Equation (3.124)
is similar to the expression for biographically non-uniform surfaces. However, for
induced non-uniform surfaces in the region of medium coverage, the concentra-
tion-dependence is slightly more complicated than the derivation for biographical
non-uniform surfaces, as parameter U’ is dependent on partial pressures of reac-
tants and products.
 3.4 Kinetics of complex reactions 261

3.4.4 Kinetic aspects of selectivity

One of the most important requirements for catalytic reactions is proper selectivity,
which in a broad sense should be understood as chemo-, regio- and enantioselectiv-
ity. Selectivity is the ability of a catalyst to selectively favor one among various com-
petitive chemical reactions. Intrinsic selectivity is associated with the chemical
composition and structure of surface (support), while shape selectivity is related
with pore transport limitations (see Figs 2.115, 2.116 and 3.9).

Fig. 3.9: Reactant selectivity in catalysis by zeolites.

Chemoselectivity and regioselectivity describe the ability of a catalyst to discrimi-

nate among the different and the same functional group, or several orientations,
respectively. Diastereoselectivity defines control of the spatial arrangement of the
functional groups in the product, while enantioselectivity is related to the catalyst
ability to discriminate between mirror-image isomers or enantiomers.
Notions of differential and integral selectivity are often used. The former is the
ratio between the accumulation rate of the desired product and the total rate of con-
versions of one of the original substances in various directions, including the for-
mation of the desired product at a given temperature and with a given composition
of the reaction mixture: S = rdes=r.
Differential selectivity relates the rates, integral selectivity is the ratio of the de-
sired product concentration to the total concentration of all of the products, de-
scribing the course of the reaction up to a given point. In a gradientless system
differential and integral selectivity coincide.
As selectivity depends on conversion, it is extremely dangerous to compare selec-
tivity for different catalysts at just one end-point or at a certain period of time. For
parallel reactions of the same kinetic order it still could be done, as selectivity for a
system 1. A ) B; 2. A ) C is independent of the concentration of A and therefore of
conversion.

r+1 k + 1 KA CA k+1
SB = = = (3:125)
r + 1 + r + 2 k + 1 KA CA + k + 2 KA CA k + 1 + k + 2
262 3 Engineering reactions

At the same time, for consecutive reactions 1. A → C; 2. C → D the expression for

selectivity that assumes equilibrium binding of the reactants A and C is conversion
dependent (Fig. 3.10):

100
Selectivity (%)

50

0
0 25 50 75 100
Conversion (%)

Fig. 3.10: Selectivity vs conversion dependence for consecutive reactions.

r+1 − r+2 r+2 k + 2 K2 CC

SC = =1− =1− (3:126)
r+1 r+1 k + 1 K1 CA

The steepness of the decline depends on k+2K2/k+1K1 ratio. When adsorption proper-
ties of A and C are similar, selectivity could be enhanced by applying a certain sub-
stance, which displaces C on the surface or impedes its re-adsorption. Such strategy
is implied in selective hydrogenation of acetylene to ethylene when CO is added to
the feed in minute (ppm) amounts.
Let us expand the kinetic analysis to a parallel-consecutive hydrogenation reac-
tion (Fig. 3.11).
This mechanism can be visualized using the catalytic cycle approach with the
vertices (nodes) and edges corresponding to the intermediates and steps, respec-
tively. In the case of linear elementary steps the edges join the vertices, which cor-
respond to the intermediates (adsorbed species ZA, ZB, ZC, ZD and vacant sites Z).
The rates along three routes, represented by three cycles in the graph (Fig. 3.11B)
can be easily derived assuming that hydrogenation occurs in the liquid-phase by
Eley-Rideal mechanism and that all reactants and products in principle could adsorb
on the catalyst surface.
If deriving kinetic equations from the mechanism above, one arrives at follow-
ing equations:
 3.4 Kinetics of complex reactions 263

B Z C ZD

C
+H2 (2) +H2 (3) A +H2
ZB ZC
A D

+H2 (1) +H2 ZA +H2

(A) B (B)

Fig. 3.11: Parallel-consecutive hydrogenation: (A) mechanism, (B) kinetic graph.

k + 1 KA cA cH2 k + 2 KA cA cH2
r1 = , r2 =
1 + KA cA + KB cB + KC cC + KD cD 1 + KA cA + KB cB + KC cC + KD cD
(3:127)
k + 3 KC cC cH2
r3 =
1 + KA cA + KB cB + KC cC + KD cD

where k+1 is rate constant of reaction 1 in the forward direction, KA is the adsorption
constant and cA is the concentration of component (A) etc., and cH2 is hydrogen
concentration in the liquid-phase. Equation (3.127) should be combined further to
the mass balances of the components:

1 dcA 1 dcB 1 dcC 1 dcD

= − r1 − r2 ; = r1 ; = r2 − r3 ; = r3 (3:128)
ρ dt ρ dt ρ dt ρ dt

where ρ is the catalyst bulk density. Selectivity towards the compound C is:

dcC r2 − r3 k + 2 KA cA − k + 3 KC cC
SC = =− =−
dcA r1 + r2 k + 1 KA cA + k + 2 KA cA
(3:129)
k+2 k + 3 KC cC cC
=− + = −L+M
k + 1 + k + 2 ðk + 1 + k + 2 ÞKA cA cA

An analytical solution of (3.129) can be readily obtained:

L CA M
CC = − CA (3:130)
1 − M ðCAo ÞM − 1

where CAo is the initial concentration of the reactant at which CCo =0,
L = k + 2 =ðk + 1 + k + 2 Þ and M = k + 3 KC =ððk + 1 + k + 2 ÞKA Þ.
Equation (3.130) can be rearranged [2] to

L 1h i
SC = ð1 − δÞM − ð1 − δÞ (3:131)
1−Mδ

where δ is conversion.
264 3 Engineering reactions

It follows from eq. (3.131) that selectivity is independent of hydrogen pressure

being dependent on conversion.
Equation (3.131) implies and Fig. 3.10 illustrates that there is a substantial dif-
ference in selectivity of the intermediate on conversion. Thus, behavior of different
catalysts should be compared in terms of selectivity at similar conversion levels.
Selectivity values are usually more accurate than the values of reaction rates, even
with comparatively large errors in these reaction rates. Selectivity analysis is thus
crucial for elucidation of reaction mechanisms in complex reactions, making it pos-
sible to discriminate among rival reaction mechanisms in cases when statistical
data fitting provides very similar results.
Analytic solutions are readily available for first-order reaction networks be-
cause the overall system of differential equations remains linear. For a consecutive
reaction network consisting of a second-order reaction followed by the first-order
reaction, an analytical solution is very complicated. In fact, the exact solutions for
a mixed second-order process followed by a first-order reaction appeared only very
recently [4, 5]. These expressions have a very high degree of mathematical complex-
ity containing hypergeometric functions. They are not tractable for analysis of, for
example, selectivity dependence on conversion and reaction parameters. An alter-
native to analytic expressions is numerical solutions, which can be rather easily ob-
tained using commercially available software as well as freeware.
Such analysis performed in [6] demonstrated that for the reaction network com-
prising the second-order reaction followed by the first-order reaction just by dilut-
ing the reaction mixture 10fold selectivity to the intermediate at approximately 20%
conversion can dramatically drop from 95% to 5%. Such variations in selectivity
when different catalysts of different research groups are compared can incorrectly
be attributed to intrinsic nature of catalysts rather than different initial substrate
concentrations (solvent amounts), which often is not discussed.
The reaction mechanism with linear steps can easily be shown graphically (see
Fig. 3.11B). For non-linear mechanisms the situation is somewhat different. For a
simple Langmuir-Hinshelwood mechanism (eq. 3.34) visualization can be done as
presented in Fig. 3.12 with the non-linear step 3 depicted in the graph as a line con-
necting node (*A) with (*), touching (*B) and additionally as a broken line connect-
ing (*B) and (*).
The mechanism in eq. (3.34) contains only one route, however from Fig. 3.12 a
wrong conclusion could be made about two catalytic cycles, which emphasizes that
visualization of non-linear mechanisms is not straightforward. There is, however,
an incentive to treat heterogeneous complex catalytic kinetics in the same spirit as
homogeneous organometallic systems using the notion of catalytic cycles. Thus in
the literature there are mechanisms with non-linear steps, which are incorrectly dis-
cussed in terms of catalytic cycles, giving a misleading impression of reaction
mechanisms and leading to incorrect rate expressions.
 3.4 Kinetics of complex reactions 265

+A

*A 1
+B

2
*B *

C Fig. 3.12: Graph for a mechanism with a non-linear step.

A convenient way to overcome the difficulties in the visualization of reaction

networks is to use the catalytic cycle approach, in a similar way to linear sequences
with the number of nodes (and respective edges) corresponding to the number of
steps. For example, CO oxidation with stepwise (molecular to dissociative) oxygen
adsorption and two non-linear steps 2 and 4 is considered below:

1. O2 + z ) zO2 1
2. zO2 + z ) 2zO 1
3. CO + z , COz 2 (3:132)
4. COz + Oz ) CO2 + 2z 2
2CO + O2 = 2CO2

The graph for such mechanism can be visualized either with a conventional approach
(Fig. 3.13A) or with a concept of a reaction route cycle [2] (Fig. 3.13B). In Fig. 3.13B the
nodes contain all the surface species taking part in the cycle.

2CO2 4Z +2COC

+O2 +CO 4 3
1
3
ZO2 2 Z ZCO 2ZO + 2ZCO 2ZCO + 2Z
2
CO2 2 1 +O2
4 4
ZO ZO2 + 2ZCO + Z
(A) (B)

Fig. 3.13: Langmuir-Hinshelwood CO oxidation mechanism with dissociative oxygen adsorption:

(A) kinetic graph and (B) catalytic cycle.
266 3 Engineering reactions

In the reaction route cycle (similar to Fig. 3.13B) the number of species in the
node NN of a cycle should at least correspond to free sites required for adsorption/
reaction steps, taking into account stoichiometric numbers of these steps. This num-
ber can be calculated by considering all the free sites on the left-hand side of equa-
tions for elementary reactions multiplied by the respective stoichiometric numbers.
Additionally, sites (corrected with the stoichiometric numbers of steps) that are re-
leased and appear on the right hand side of steps, preceding those adsorption/reac-
tion steps, should be subtracted. This analysis should be performed for all routes.

3.4.5 Structure sensitivity

In the particle size regime between 1 and 20 nm, three types of reactivity change are
distinguished. This is illustrated in Fig. 3.14. Some reactions are independent of parti-
cle size. The rates of other reactions, normalized per exposed metal surface atom, de-
crease, increase or even pass through a maximum when particle size increases. This
dependence of catalytic rate on size or dispersion of catalytically active particles has
been known since Boudart [7] introduced the notion of structure-sensitive and insen-
sitive reactions in the 1960s.
TOF

Diameter Fig. 3.14: Structure sensitivity plots.

The thermodynamic approach was used quantitatively to explain the cluster size
effect [8]. It relied on the concept of chemical potential μ(r) (or the surface energy
excess of the metal atom in a metal particle) dependence on the cluster radius r
size. The following expression for the Gibbs energy of large clusters was obtained:

ΔGads ðrÞ − ηRT

ΔGads, ∞ = (3:133)
r

where η is a parameter, which can be either negative or positive. Taking into ac-
count relationships between equilibrium constant and Gibbs energy, as well as the
relationship between thermodynamics and kinetics, a following expression can be
written for adsorption:
 3.4 Kinetics of complex reactions 267

αη
kðrÞ = gKðrÞα = k∞ e r (3:134)

For the two-step sequence it was demonstrated [8] that turnover frequency v(r) is:
η
ðk + 1 PA1 k + 2 PA2 − k − 1 PB1 k − 2 PB2 Þeðα2 + α1 − 1Þ r
vðrÞ = η η η η (3:135)
k1 + PA1 eα1 r + k + 2 PA2 eðα2 − 1Þ r + k − 1 PB1 eðα1 − 1Þ r + k − 2 PB2 eα2 r

where PA1 , etc., are partial pressures (for gas-phase reactions) or concentrations (for
liquid-phase reactions), ki are kinetic constants and α are the Polanyi parameters of
steps. If these parameters a are equal to each other and the overall reaction is
irreversible:
η η
ω + 2 eðα − 1Þ r p1 eðα − 1Þ r
vðrÞ = ω+2 +ω−1 η = η (3:136)
1+ ω+1 e− r 1 + p2 e − r

with ωi -frequencies of steps (i.e., ω + 1 = k + 1 PA1 , etc.), p2 = ðω + 2 + ω − 1 Þ=ω + 1 and

p1 = ω + 2 .
Comparison of eq. (3.136) with the experimental data demonstrated a very good
correspondence of theory and experiments and even the ability to describe maxima
in TOF as a function of cluster size.
The same equation can be derived [9], taking into account differences in the
Gibbs adsorption energy between edges, corners and terraces:

ΔGads = ΔGads, terraces fterraces + ΔGads, edfes fedges + ΔGads, corners fcorners (3:137)

where fterraces, fedges denote fractions of these surface sites, whose sum is obviously
equal to unity. The geometrical considerations for different particle size shapes are
presented in Fig. 3.15.
Theoretical analysis of the models presented above demonstrated that, in line
with experimental data, significant changes in kinetic regularities (for example, re-
action orders and selectivity) could be anticipated with varying cluster size.
An alternative way of implementing the cluster size dependence in reaction ki-
netics is to view reaction constants as size independent. Then each type of site (ter-
races, corners and edges) contributes to the overall rate according to:

vðdÞ = rcorners fcorners + redges fedges + rterraces fterraces (3:138)

In (3.138) it is assumed that adsorption/desorption and reaction rates on terraces,

edges and corners are always the same, independent of the cluster size, which
might be unrealistic especially for reactions involving large organic molecules
when an ensemble of sites is required for multi-centered adsorption, and reactivity
can depend on the cluster size even for terraces.
268 3 Engineering reactions

100

80
Surface atoms (%)

Faces
60

40
Edges

20
Corners

0
0.2 0.4 0.6 0.8 1.0
Dispersion

Fig. 3.15: Fraction of exposed surface atoms as a function of metal dispersion. (From [10] copyright
Elsevier).

3.4.6 Mechanism-free kinetics – kinetic polynomial

For many catalytic reactions with non-linear steps, derivation of kinetic equations
could be challenging. In order to avoid such difficulties Lazman and Yablonsky ap-
plied constructive algebraic geometry to non-linear kinetics that expresses the reac-
tion rate of a complex reaction as an implicit function of concentrations and
temperature. This concept of kinetic polynomial [11] has found important applica-
tions, including parameter estimation, analysis of kinetic model identifiability and
finding all steady states of kinetic models. The Lazman-Yablonsky four-term rate
equation for polynomial kinetics is:
−1
k + ðf + ðcÞ − Keq f − ðcÞÞ
r= Nðk, cÞ (3:139)
Σðk, cÞ

The terms in eq. (3.139) are as follows: (1) the kinetic apparent coefficient k+; (2) the
potential term, or driving force related to the thermodynamics of the net reaction;
(3) the term of resistance, i.e., the denominator, which reflects the complexity of
reaction, both its multistep character and its non-linearity; finally, (4) the non-
linear term [N(k,c)], a polynomial in concentrations and kinetic parameters, which
is caused exclusively by non-linear steps. In the case of a linear mechanism, this
term vanishes. In classical kinetics of heterogeneous catalysis (LHHW equations)
such term is absent.
Such phenomenological or mechanism-free expressions, although useful for
some applications and for preliminary analysis of kinetic data, are in general not
reliable as they do not predict reaction rates, concentration or temperature depen-
dence outside the experimental conditions. Models based on the knowledge of
 3.4 Kinetics of complex reactions 269

elementary processes provide reliable extrapolation outside of the experimentally

studied domain. In addition, the system is better understood.

3.4.7 What is behind a rate constant?

Transition state theory of surface reactions was developed independently by Laidler

and Glasstone [12], Eyring [13] and Temkin [14] shortly after the appearance of the
more general approach by Eyring and Polanyi. According to this theory, the reac-
tion rate for a reaction nAA +nB B +Iads+Jads→ C comprising reactants A and B from
the gas phase and adsorbed species i and j is given by [14]:
 
kB T q≠ ΔE
r = χL exp − ½AnA ½BnB θi θj (3:140)
h qA qB RT

where kB is the Boltzmann constant, h is the Plank constant, q≠ is the modified par-
tition function of the transition state, and qA and qB are partition functions of the
reactants, χ is the transmission coefficient, reflecting nonadiabaticity in the quan-
tum-mechanical sense, L is the total number of sites per unit area, and nA and nB
are reaction orders in respective compounds. The concentrations of the gas-phase
molecules are given in number of molecules per volume of gas. Quite often instead
of gas-phase concentrations, respective pressures are used; therefore, the equation
for the rate should be modified [14]
 
kB T q≠ ΔE
r = χL exp − ½PA nA ½PB nB θi θj (3:141)
hðkB TÞnA + nA qA qB RT

Partition functions are important because some useful properties follow from
them. To calculate the values of the partition functions of the molecule, each of
the motions (electronic, translational, rotational and vibrational) should be con-
sidered, keeping in mind that a molecule of n atoms has 3n degrees of freedom.
For a molecule there are three degrees of translational freedom and 3n – 6 and
3n – 5 for the vibrational freedom of non-linear and linear molecules, respectively.
Thus an atom (n = 1) has three translation degrees of freedom; a molecule of two
atoms has three translational, one vibrational, and two rotational degrees of free-
dom. For a linear molecule of three atoms, in addition to three translational de-
grees of freedom, there are four vibrational and two rotational degrees of
freedom, while for a non-linear molecule there will be three vibrational and three
rotational degrees of freedom. There is one specific feature in the partition func-
tions for the transition state as the one vibrational degree of freedom is replaced
by the movement along the reaction coordinate. Thus the total number of degrees
of freedom is 3n – 1.
270 3 Engineering reactions

The translational partition function can be calculated knowing the particle

mass, temperature and is conventionally calculated per unit volume: q3D
trans or per
dimension qtransl
1 !3 1
ð2πmkB TÞ2 ð2πmkB TÞ2
q3D
trans =V ; qtransl = I (3:142)
h h

where m is the reduced mass. For example, for a molecule AB this is mAmB/(mA
+ mB). The vibrational energy levels are usually calculated using the harmonic os-
cillator approximation. For a diatomic molecule, the vibrational partition function
per degree of freedom is:
1
qvib =   (3:143)
− hv
1 − e kB T

IR spectra are applied for calculations of vibrational partition functions, clearly rep-
resenting a challenge in the calculation of vibration frequency of short-lived (10–12 s)
transition states. However, vibrational partition functions are usually equal to unity,
unless the vibrations are low frequency. For calculations of rotational partition func-
tions the moment of inertia I (for example, the molecule structure) must be known.
In the case of a diatomic molecule, the rotation partition function is:

8kB π2 IT
qAB = (3:144)
h2
The electronic partition function qel is usually equal to unity.
For an indirect molecular adsorption on a surface, the adsorbed molecule is
free to move across the surface, rotate and vibrate. For such precursor-mediated ad-
sorption the pre-exponential factor in eq. (3.140) takes the form:

ðkB TÞq#2 q# #
rot qvib ðkB TÞq#2
kads = trans
× = trans
× Sid (3:145)
hqtrans 3 qrot qvib hqtrans 3
ð#Þ ð#Þ
where the former term is the collision factor and Sid = ðqrot qvib Þ=ðqrot qvib Þ is the
sticking coefficient for indirect adsorption. If there are no changes in the values of
rotational and vibrational partition functions of the adsorbed molecule, the sticking
coefficient is equal to unity. The transition state is associated, however, with the
decrease of entropy, therefore the sticking coefficient is less than unity and is usu-
ally a function of surface coverage, temperature and the surface structure of the
adsorbent.
For direct adsorption on solid surfaces with limited mobility of adsorbed spe-
cies the pre-exponential factor is:
 3.4 Kinetics of complex reactions 271

#ads
kB T qvib kB T q#2
kads = 3
= trans
× Sd (3:146)
h qtrans qrot qvib hq3trans

with the same collision factor as in (3.145) but a much smaller (10–6–10–8) sticking
q#ads
coefficient Sd = vib
caused by considerable losses in entropy when the
ðq#2
trans qrot qvib Þ
molecule is strongly fixed on the surface after direct adsorption.
For desorption of a molecule with similar freedom for adsorbed species and the
transition state:

kb T qA#
k= ≈ 1013 s − 1
h qA*
(3:147)
qA # kb T
as ≈ 1 and ≈ 1013 s − 1 .
qA* h

When transition state is mobile, the prefactor for desorption is in the range 1014–16 s−1
in line with experimental data [15]. The values of prefactors for surface reactions of
the type A*→B* are approximately 1013 s−1, while the prefactor for dissociation of an
adsorbate AB*→A*+B* is related to the value of desorption prefactor of the same mol-
ecule, that is, kdiss=~10–3 kdes [15].
Transition state theory can be applied to complex reactions, theoretically pre-
dicting the values of pre-exponential factors for multistep reaction mechanisms,
where because of over-parametrization the direct numerical fitting of rate constants
results in values with a large error interval.
Kinetic treatment based on the theory of complex reactions – which went be-
yond the simplified Langmuir-Hinshelwood-Hougen-Watson approach with one
rate-determining step (Tab. 3.1) – made the calculation of parameters necessary
(pre-exponential factors, activation energies of elementary reactions, etc.). Such
theoretical approaches started to appear in 1970s and were based on thermodynam-
ics and transition state theory, as well as surface science experimental tools (ultra-
high vacuum studies, spectroscopy) in combination with enhanced computer
power. Later on they were nicely summarized, refined and popularized by
J. Dumesic and co-workers [16] as microkinetic modeling. In essence, micro-kinetic
modeling assembles molecular-level information obtained from quantum chemical
calculations, atomistic simulations and experiments to quantify the kinetic behav-
ior at given reaction conditions on a particular catalyst surface. Some examples will
be provided in Chapter 4.
272 3 Engineering reactions

3.4.8 Apparent activation energy of complex reactions

Apparent activation energy is an often used concept for complex reactions, describ-
ing the overall dependence of a reaction on temperature. The expression for the ap-
parent (observed) activation energy is given as follows:

∂ ln r ∂ ln r
Eact, apparent = − R = RT 2 (1:3)
∂1=T ∂T

As an example, the two-step mechanism (3.48) with two irreversible steps gives an
expression for the rate

ω + 1ω + 2 k + 1 CA1 k + 2 CA2
r= = (3:148)
ω + 1 + ω + 2 k + 1 CA1 + k + 2 CA2

Considering eq. (1.3) the apparent activation energy can be calculated as

k + 1 CA k + 2 CA
∂ ln k 1 2
∂
+ 1 CA + k + 2 CA
Eact, apparent = RT 2 1 2
= RT 2 ðln k + 1 + ln CA1 + ln k + 2 + ln CA2
∂T ∂T
− lnðk + 1 CA1 + k + 2 CA2 ÞÞ ð3:149Þ

Neglecting dependence of the prefactor on temperature in ln k + 1 in comparison

with the exponential term of the Arrhenius equation one gets

∂ ∂ E+1
ðln k0 + 1 expð − E + 1 =RTÞÞ = ðln k0 + 1 − E + 1 =RTÞ = (3:150)
∂T ∂T RT 2

and

∂ E+2
ðln k0 + 2 expð − E + 2 =RTÞÞ = (3:151)
∂T RT 2

The last term in eq. (3.149) can be transformed

∂ ∂ lnðk + 1 CA1 + k + 2 CA2 Þ ∂ðk + 1 CA1 + k + 2 CA2 Þ

ðlnðk + 1 CA1 + k + 2 CA2 ÞÞ = =
∂T ∂ðk + 1 CA1 + k + 2 CA2 Þ ∂T

∂ðk + 1 CA1 + k + 2 CA2 Þ CA1 k0 + 1 e − E + 1 =RT RT+ 21 + CA2 k0 + 2 e − E + 2 =RT

E E+2
1 RT 2
= = =
k + 1 CA1 + k + 2 CA2 ∂T k + 1 CA1 + k + 2 CA2
E+1 E+2
k + 1 CA 1 + k + 2 CA 2
= RT 2 RT 2 ð3:152Þ
k + 1 CA1 + k + 2 CA2

The final expression for the apparent activation energy is

 3.4 Kinetics of complex reactions 273

k + 1 CA1 k + 2 CA 2
Ea, app = E + 1 + Ea + 2 − E − E+2 (3:153)
k + 1 CA1 + k + 2 CA2 + 1 k + 1 CA1 + k + 2 CA2

A special case of the two-step sequence is the Eley–Rideal mechanism, where the
first step in the mechanism [3.48] can be considered as quasi-equilibrium, resulting
in eq. (3.32). The activation energy of such reaction is illustrated in Fig. 3.16.

Eact1
Potential enregy

ΔH1
Eact2
ΔHrxn

Reaction coordinate

Fig. 3.16: Activation energy profile for a two-step reaction with a low value of activation energy for
the binding step.

Equation (3.32) can be written as

ko+ 2 KAo CA eð − ΔH1 − E + 2 Þ=RT

r= (3:154)
1 + KAo CA e − ΔH1 =RT

When binding of the substrate to the catalyst is not strong enough and the cata-
lyst mainly exists in the form of unbound species, KAo CA e − ΔH1 =RT < < 1; the appar-
ent activation energy is simply Eact = E + 2 + ΔH1 and can be even negative when the
enthalpy of the exothermal intermediate complex formation (being negative) is
larger than the activation energy of the second step. Experimentally, such nega-
tive values of activation energy were reported for a homogeneous Diels–Alder re-
action of dimethylanthracene and tetracyanoethylene in the series of inert
solvents [17, 18] and for the organocatalytic Mannich reaction of an imine and
fluorinated ketoester catalyzed by thiourea [19]. In the latter case, theoretical cal-
culations of the enthalpy showed that binding of the catalyst to substrate occurs
without an activation barrier, and that the overall enthalpy change of the
274 3 Engineering reactions

transition state is negative relative to starting substrates. In heterogeneous cata-

lytic cracking of n-alkanes over ZSM-5, the observed activation energy decreases
from 140 to −50 kJ/mol when the carbon chain length increases from three to 20
[20]. This dependence is a consequence of the competition between an intrinsic
kinetics dependence with temperature, and a decrease in the adsorption strength
and the concentration of active intermediates with temperature.
In fact, eq. (3.154) can have a maximum as a function of temperature, which
can be easily demonstrated by considering the reciprocal function and taking a de-
rivative of it. When the derivative is equal to zero
 ′
1 + KAo CA e − ΔH1 =RT
=0 (3:155)
ko+ 2 KAo CA eð − ΔH1 − E + 2 Þ=RT

an expression for the temperature at optimum is

− ΔH1 =R
Tmax = − ΔH1 − E + 2
(3:156)
ln K1o CA E + 2

which shows clearly that positive values of Tmax are achieved when the enthalpy of
the exothermal intermediate complex formation is larger than the activation energy
of the second step.

3.4.9 Dynamic catalysis and deactivation

In the chemical industry the majority of processes are designed to be conducted in

stationary conditions. Hence, it is not surprising that steady-state kinetics is mainly
studied as the surface concentration of reactants is time independent. The investi-
gation of reaction kinetics in gradientless reactors further simplifies the experimen-
tal routine.
Although steady-state kinetic experiments give valuable information about the
overall rates of catalytic reactions, they cannot unequivocally reveal the underlying
reaction mechanisms because different reaction mechanisms might give similar
steady-state rate equations.
From an industrial viewpoint the main drawback of kinetic models, based
only on steady-state data, is that start-up and transient regimes cannot be reliably
modelled. Kinetic models for non-stationary conditions should be also applied for
the processes in fluidized beds, reactions in riser (reactor) – regenerator units
with catalyst circulation – as well as for various environmental applications
of heterogeneous catalysis, when the composition of the treated gas changes
continuously.
For the construction of a kinetic model for non-stationary conditions, knowl-
edge about the evolution of the concentrations of adsorbed species on the catalyst
 3.4 Kinetics of complex reactions 275

surface is needed. Under non-stationary conditions, the changes of concentration

fields in time, reactor space and catalyst surface (for heterogeneous catalysis), are
interrelated by complex dependencies. Therefore, knowledge about the gas and sur-
face composition is required for experimental investigations under non-stationary
conditions.
In addition, in many catalytic reactions changes in the rate constants of some
elementary steps are associated with the side reactions, which are not present in
catalytic cycles. When such deactivation is prominent it requires some counterbal-
ancing at the catalyst and reactor levels, as will be illustrated in Chapter 4. The
time scale of deactivation depends on the process type and can vary from a few sec-
onds, as in fluid catalytic cracking (FCC), to several years, for example, in steam
reforming of natural gas or oxidation of o-xylene, which are also discussed in
Chapter 4.
The main causes of deactivation in heterogeneous catalysis are poisoning, foul-
ing, thermal degradation (sintering, evaporation) initiated by the often high tem-
perature, mechanical damage and corrosion/leaching by the reaction mixture
(Fig. 3.17).

S S

Sintering Catalyst
Selective poisoning particle Fine

S
Attrition
S

Nonselective poisoning
Fouling

Active site
Support
Leaching Component in reaction medium

Fig. 3.17: Deactivation mechanisms. (From [21]).

Poisoning is usually caused by strong adsorption of impurities in the feed. Poisoning

depends upon adsorption strength of such poisons relative to the other species com-
peting for catalytic sites. Adsorbed poisons may not only block active sites, but change
the electronic or geometric structure of the surfaces as well. Fouling is associated with
covering of the catalyst surface by deposits, which are quite often hydrogen-deficient
carbonaceous materials (i.e., coke), making the active sites inaccessible.
For liquid-phase reactions catalysis leaching can be prominent, with the leached
metals responsible for high catalytic activity in some reactions. Such leaching is par-
ticularly problematic in oxidation catalysis because of the strong complexation and
276 3 Engineering reactions

solvolytic properties of oxidants (H2O2, RO2H) and/or the products (H2O, ROH,
RCO2H, etc.).
Strong stresses of packed catalyst beds during start-ups, shut-downs and catalyst
regeneration could lead to mechanical deactivation. Finally, thermal degradation (be-
cause of sintering, chemical transformations, evaporation, etc.) could cause deactiva-
tion. Of particular prominence is sintering, or, the loss of catalyst active surface
caused by crystallite growth of either the support material or the active phase.
Various kinetic models to account for deactivation have been presented in the
literature, of different empirical and semi-empirical natures. Such models often sep-
arate activity function a (the current reaction rate divided by the initial intrinsic
rate, a = rt/r0) and deactivation:

da
− = f ðci , TÞφðaÞ (3:157)
dt

In eq. (3.157) f(ci, T) reflects the activity on non-deactivated catalyst and φ(a) is de-
activation function. This approach of separability is applied when the reaction and
deactivation rates are of different magnitudes; the reactions proceed relatively fast,
while deactivation requires much longer time (hours, days or months) and in addi-
tion deactivation does not affect selectivity.
Different equations have been presented in the literature for the deactivation
function, for example, a power law one:

φðaÞ = am (3:158)

As activity changes in fixed-bed reactors are observed with time-on-stream, deacti-

vation is often expressed in terms of time, which is not, however, a true variable.
More correctly, the deactivation function has to be expressed in terms of the deacti-
vating agent: the coke precursor or the poison, which means that the amount of
coke (or poison) on the catalyst site should be known. The determination of a rate
equation for the formation of the coke precursor is thus an integral part of the ki-
netic study of the process.
For the formal treatment of deactivation for separable kinetics, the fraction of
non-coked or non-poisoned catalyst (f) can be defined in terms of the concentration
of the poison or coke on the catalyst surface (cP) and the catalyst capacity for coke
or poison (cP,0):
cP, 0 − cP
f= (3:159)
cP, 0

Accumulation of coke (or poison) on the catalyst is given by:

dcP
= RP ðcP , p, T, WHSVÞ (3:160)
dt
 3.4 Kinetics of complex reactions 277

where WHSV stands for weight hour space velocity for fixed-bed operation in con-
tinuous mode. Separating activity and deactivation, as well as assuming a power
law for the dependence of RP on coke concentration (rate of coke formation is pro-
portional to some power (n) of unused capacity for adsorbing coke):

dcP
= kðcP, 0 − cP Þn ϕðp, T, WHSVÞ (3:161)
dt

after integrating for constant values of cP,0 and f the linear exponential and hyper-
bolic relationships between fraction of coke and time could be obtained:
" # 1
n−1
1 
f= n≠1; f = exp − k′cnP,−01 t , n = 1 (3:162)
1 + ðn − 1Þk′cnP,−01 t

where k′ = kϕ(p, T, WHSV) is only an apparent constant, as it depends on tempera-

ture, WHSV and the reactant pressure p.
However, when deactivation is considered as a constituent part of the reaction
scheme, it gives more possibilities for clarifying deactivation mechanisms. For a
scheme:

A + * $ A* $ B* $ B + *
"# (3:163)
C*

Where C represent adsorbed coke, deactivation rC and self-regeneration r–C proceed

simultaneously with these steps being essentially slower than the reaction steps.
Applying the quasi steady-state only to the main reaction dθA =dt ≈ 0 and supposing
for simplification that adsorbed A is the most abundant surface intermediate, the
changes in coke coverage as a function of time-on-stream are:

dð1 − θA Þ
= kc θA − k − C ð1 − θA Þ (3:164)
dt

where kC and k–C are deactivation and self-regeneration constants. When integrat-
ing eq. (3.164) with the boundary conditions t = 0, θOA = 1 (i.e., the surface is initially
totally covered by A), an analytical expression for the reaction rate is obtained:

r = r0 θA = a3 + a1 e − a2 t (3:165)

where r0 is the reaction rate at deactivation free conditions and:

kC r0 k − C r0
a1 = ; a2 = kC + k − C ; a3 = (3:166)
kC + k − C kC + k − C
278 3 Engineering reactions

At the infinite time the reaction rate corresponds to a3 because of self-regeneration.

The steepness of activity loss is characterized by a2, while the sum of a1 and a3 cor-
responds to the rate at deactivation free conditions.
For sintering of the metal crystallites on a catalyst support, models for activity
decline could also include certain residual activity level similarly to eq. (3.166).
Generally, it can be stated that derivation of a rate equation with deactivation
can be based on the theory of complex reactions, considering deactivation as an
independent route leading to coke (C) on the catalyst surface (Tab. 3.2).

Tab. 3.2: Reaction routes and kinetic equations for a mechanism including deactivation.

Nð1Þ Nð2Þ 1 − θC* rA = − r2 − jvA jr4

θV =
1. A + * ≡ A* 1 1 1 + KA cA + KB cB rB = r2 − jvB jr4
2. A* ! B* 1 0 1 − θC* rC* = r4
r2 = ðk + 2 KA c − k − 2 KB cB Þ
3. B* ≡ B + * 1 0 1 + KA cA + KB cB jvA j = 1, jvB j = 0
1 − θC*
4. A* ! C* 0 1 r4 = k + 4 KA cA
1 + KA cA + KB cB
Nð1Þ A ! B
Nð2Þ A ! C*

More examples are provided in the monograph [2], which covers several linear and
non-linear models, including mono- and bimolecular deactivation mechanisms for
reactant and product molecules as well as a reversible deactivation reaction.
Contrary to experiments in fixed-bed continuous reactors, deactivation of cata-
lytic reactions performed in batch modes can be hidden, as the observed behavior
could be similar to first-order kinetics. It is typical to then perform experiments
under identical conditions but recycling the catalyst.
For the modelling of deactivation, a semi-empirical equation can be used:

CR = CR0 − a1 te − kd τ (3:167)

where a1 is a complex parameter (could include reactant concentration), CRo is the

initial concentration of reactant, kd is the deactivation constant, t is the time, and τ
is the total time starting from the beginning of the series.
In multilayer coking, as observed when organic compounds react on catalyst
surfaces, several carbonaceous layers are build-up (Fig. 3.18): R + θ0 ! θ1 ,
R + θ1 ! θ2 ...R + θn − 1 ! θn where R denotes the reactant.
The rate of coke generation is composed from the rate of monolayer and multi-
layer formation:
 3.4 Kinetics of complex reactions 279

θn

θ1

n
0 l
θ0 = ∑θl θp = ∑θn
n=1 Fig. 3.18: The model for multilayer coking.

1 dCC XN
= km θ0 cR + kp cR θn − 1 = km cR ð1 − θP Þ + kp cR ðθP − θN Þ (3:168)
ξ dt n=2

where ξ is the quantity of coke, derived from one mole of reactant and θp is the cov-
erage of coke.
As activity changes in the main reaction are only caused by blocking of the ac-
tive sites by coke formation in the first layer, the coverage of coke θp is:

Cm dθP
= km cR ð1 − θP Þ (3:169)
ξ dt

where Cm is the capacity of a monolayer, or the maximum amount of coke (gram of

coke/gram of catalyst), which could be adsorbed on the catalyst for monolayer cok-
ing. Equations (3.168) and (3.169) should be solved simultaneously, which is a very
demanding task. Simplifications of the model for the linear mechanism of the main
reaction are a = 1 – θp and θN = 0. After integration an equation relating activity to
the coke concentration can be obtained:

CC
= ð1 − φÞð1 − aÞ − φ ln a (3:170)
Cm

where φ = kp/km. Equation (3.170) includes linear and exponential dependences as

special cases. If multilayer coking is not profound, then φ = kp/km ≈ 0 and CC/Cm =
(1 – a). When the deactivation constants for monolayer and multilayer coking are
the same and φ = kp/km = 1, then CC/Cm = –lna and a = 1 – CC/Cm. Similar treatment
could be extended to cases of non-linear deactivation and non-linear mechanisms
for the main reaction.
When a poison molecule (P) deposits on a vacant surface site, P + * Ð P *a cer-
tain part of the catalytically active sites is blocked. For the first-order kinetics in a
batch reactor it would mean that instead of m (catalyst mass) an effective catalyst
280 3 Engineering reactions

mass (m* = m – f), should be used, where f is the amount of catalyst irreversibly
blocked by the poisons. If f is independent on reactant concentration, one arrives at:

C = Co e − kðm − f Þt (3:171)

where C = Co at t = 0.

3.4.10 Mathematical treatment of experimental data

In addition to the selection and construction of experimental equipment, plan-

ning of experiments and implementing them, the procedures for obtaining ki-
netic parameters also involve checking the experimental data consistency of the
developing kinetic models and evaluation of the kinetic data. Modern statistical
methods of kinetic evaluation require parameter estimation through numerical
data fitting with subsequent evaluation of the physical and statistical consis-
tency of the parameters.
A heuristic approach to the generation of kinetic data includes experiments
with varying parameters that are considered to be important, i.e., temperature, con-
centrations, pressures, pH, catalyst amount, volume, etc. After estimation of the
rate constants by regression analysis and evaluation of the adequacy of the model,
additional experiments are performed if the results are not satisfying. This is fol-
lowed by another round of parameter estimation and possible simulations outside
the studied parameter domain, checking if it is possible to predict the kinetic be-
havior outside of the studied initially domain.
In experimental design, possible experimental errors and the necessity of re-
peated experiments should be accounted for, and a decision on the minimum set of
experiments should be made. In addition, performing experiments under non-
isothermal conditions should be considered because the temperature profile can be
followed on-line. For non-isothermal reactors, the experimental temperature field
can be interpolated by fitting an empirical model for each temperature profile, T = f
(reactor length), using, for example, a polynomial model.
The structure of kinetic models used in parameter estimation includes expres-
sionsXfor the reaction rates rj = fj(k,K,c), generation rates of the components
ri = vij rj and the mass balances of reactors, presented in the corresponding sec-
tions jof this book for some reactor types.
Differential models for batch and fixed-bed reactors can generally be described
by ordinary differential equations (ODEs), such as:

d^y
= f ðy, pÞ (3:172)
dΘ
 3.4 Kinetics of complex reactions 281

where Θ represents time, length, volume, etc., depending on the particular case.
Partial differential equations, particularly the hyperbolic and parabolic ones, can
be transformed into the form of ordinary differential equations by discretization.
Non-linear regression analysis is applied in the estimation of parameter values
(typically rate and equilibrium constants, more rare mass and heat transfer coeffi-
cients), as most models in catalytic kinetics are non-linear with respect to the
parameters.
The model equation is written according to ^y = f ðx, pÞ, where y is the dependent
variable, function (f) contains independent variables (x) and parameters (p). Non-
linear regression is, strictly speaking, valid for algebraic models, but it can be applied
to differential models, because the solution (y) is obtained numerically from the
model equation. Different criteria for minimization could be applied, including most
often the sum of squares between the experimental values and model predictions:
XXXh i
QðpÞ = yi, s, t − f ðxi, s, t′ pÞ2 wi, s, t (3:173)
i=1 s=1 t=1

where Q(p) is the objective function in minimization, i denotes each component in

the reaction mixture, s denotes the different data sets and t refers to the data points,
yi,s,t is the experimental value and f(xi,s,t) is the corresponding model prediction, wi,
s,j,t is the weight matrix that is used if experimental scattering varies between differ-
ent data. In addition to this objective function, the sum of squares of the relative
errors:
" #
X X X yi, s, t − f ðxi, s, t′ pÞ2
QðpÞ = wi, s, t (3:174)
i=1 s=1 t=1
yi, s, t

or even the maximal deviation between the experimental and calculated

parameters:
XXXh i
QðpÞ = ðyi, s, t − f ðxi, s, t′ pÞ wi, s, t (3:175)
i=1 s=1 t=1

could be used. Various optimization methods exist for the minimization of the ob-
jective function to be included in modern parameter estimation software. The
Levenberg-Marquardt method is a rapid and efficient one provided there are good
initial guess of the parameters. Otherwise, the minimization can start with a slower,
but a more robust simplex algorithm, switching in the vicinity of the minimum to
the Levenberg-Marquardt algorithm. The latter method belongs to the group of the
gradient methods, which use the derivative vector of the objective function with re-
spect to the parameter directions to determine the direction where this gradient
changes most, i.e., the steepest descent direction. The steepest descent is not neces-
sarily the optimum one, thus some methods therefore use the second derivative ma-
trix of the objective function with respect to the parameters.
282 3 Engineering reactions

The main reason for deviations between experimental and predicted data origi-
nates from systematic deviations, which are caused by inadequate stoichiometry,
inappropriate kinetic models and poor calibration of the analytical equipment.
Analysis of residuals as a function of variables can pinpoint cases with inappropri-
ate weighting, systematic deviations caused by an incorrect model and poor experi-
mental design (Fig. 3.19).

(A) (B)

(C) (D)

Fig. 3.19: Residuals as a function of variables: (A) adequate model with random distribution, (B)
inappropriate weighting, (C) systematic deviations and incorrect model, (D) poor experimental
design. From [22].

The parameter estimation results are evaluated by a standard statistical analysis.

Degree of explanation can be estimated with the coefficient of determination, the
R2-value:

ky − ^yk2
R2 = 1 − (3:176)
ky − yk2

where y is the observed variable, ^y is the prediction of the variable and y is

the average of all the data points. The estimated sum of squares is divided by
the sum of squares of the simplest possible model, the average of the values.
Equation (3.176) implies that R2 values approaching 100% are desired while
values exceeding 95% represent a good fit. The R2-value does not, unfortu-
nately, distinguish between random errors caused by scattered data, and sys-
tematic deviations caused by an inadequate model also require the sensitivity
analysis.
Because the standard statistical analysis of parameters based on linearization
does not always give a realistic picture of the accuracy of the parameters, it is rec-
ommended to check the objective function (Q) as a function of each parameter
value (Fig. 3.20A,B)
Such sensitivity analysis provides additional information regarding accuracy
and identifiability of parameters. Another option is to investigate the value of the
objective function for a pair of parameters, resulting in contour plots (Fig. 3.20C).
 3.4 Kinetics of complex reactions 283

Sensitivity Sensitivity
Q 90 Q 950

85 900

80 850

75 800

70 750
0 1 2 3 4 0.1 0.2 0.3 0.4 0.5
(A) k1 B) k2

100,000

90,000 4.93

80,000 2.4
EA

70,000 0.984

0.872

60,000
2.39

50,000
0.03 0.04 0.05 0.06 0.07
(C) k

Fig. 3.20: Sensitivity plots: (A, B) objective function dependence on a parameter, (C) contour plots.

For a strong mutual correlation between parameters, contour plots become

elongated, i.e., the numerical value of one parameter can be easily compensated by
another parameter. In such cases the model cannot be used for extrapolation and
the physicochemical significance of the parameters is questionable. One of the op-
tions to suppress the correlation between pre-exponential factor and activation en-
ergy is to use the modified Arrhenius equation:
    
− Eact 1 1 − Eact
k = kave exp − ; where kave = k0 exp (3:177)
R T Tave RTave

Here Tave is the reference temperature, chosen within the experimentally covered
temperature domain, which is typically the average temperature.
284 3 Engineering reactions

In addition to strong correlations between parameters, sometimes too small or

too large values of some parameters are obtained, indicating only a magnitude of
the values without a reliable confidence range. Rearrangement of the model should
thus be done prior to further parameter estimation. For complex reactions with
many routes in addition to using microkinetic approaches, it could be recom-
mended to perform additional experiments such as kinetic measurements with reac-
tions intermediates as well as adsorption measurements.
Since parameters in a mechanistic model have a clear physico-chemical mean-
ing, there are certain constraints including positive values of kinetic parameters,
magnitude of rate and equilibrium constants, and activation energies. Adsorption
entropies are negative because binding of the substrate to the catalyst results in the
loss of entropy, therefore 0 < − ΔS0ads < S0gas . The adsorption enthalpy should be also
negative to allow negative values of the Gibbs energy of adsorption. Moreover, as
discussed in Section 3.4.7, it is possible to theoretically calculate at least some of
the values of kinetic parameters based on the transition state theory even if there
are difficulties in taking into account properly surface heterogeneity and lateral
interactions.
Sometimes after data fitting several rival models have similar values of the ob-
jective function. Several approaches have been devised to allow a selection between
these models, discussed in detail by Kapteijn et al. in [23].
Several software packages are available for modeling kinetics of chemical and
biochemical systems, including catalytic programs. Athena Visual Studio can be
used for modeling, optimization, estimation, optimal experimental design, and
model discrimination of chemically reactive and non-reactive process systems.
ModEst (from model estimation) is a software that has been designed for parame-
ter estimation of mechanistic mathematical models as well as for experimental de-
sign, and is able to deal with explicit algebraic, implicit algebraic (systems of
non-linear equations) and ordinary differential equations. This software package
includes the possibility for advanced statistical analysis, including sensitivity
data, contour plots and evaluation of estimated parameters accuracy by the
Markov chain Monte Carlo (MCMC) method, which is based on the Bayesian ap-
proach. The MCMC method, provides a tool for the evaluation of the reliability of
the model parameters by treating all the uncertainties in the data and the model-
ling as statistical distributions.

3.5 Mass transfer

3.5.1 Diffusion effects in heterogeneous catalysis

In any catalytic system it is not only chemical reactions that should be considered,
but also mass and heat transfer effects. These effects are present inside the porous
 3.5 Mass transfer 285

ci

Fig. 3.21: Concentration gradients and temperature profiles for

an exothermal fluid-solid reaction with interphase and
intraparticle diffusion.

catalyst particles and in the surrounding fluid films, resulting in concentration gra-
dients across the phase boundaries and within the particle (Fig. 3.21).
Because of heat and mass transfer the observed rate in a catalytic reaction
(macrokinetics) is different from an intrinsic rate of a catalytic transformation (mi-
crokinetics), thus modeling of a two-phase (fluid-solid) catalytic reactor includes
simultaneous reaction and diffusion in the pores of the catalyst particle. In three-
phase systems (gas-liquid-solid) diffusion effects in the liquid films at the gas-
liquid interphase should also be considered (Fig. 3.22).

ca
Gas Gas–liquid
diffusion
pA
Liquid Catalyst

Absorption
equilibrium

cA

Fig. 3.22: Mass transfer processes in three-phase systems. (From [2].).

The intraparticle and interphase mass transfer coefficients display lower tempera-
ture dependence than the intrinsic rate, as shown in Fig. 3.23 and discussed quanti-
tatively later in the book.
The design of experiments for a particular reaction system starts with defining
an experimental domain, which is usually constrained by temperature, pressure,
concentrations, solubilities, slurry density, mixing effects, kinetic rates, mass trans-
fer rates, reaction enthalpies, heat transfer rates and the thermal stability of the
286 3 Engineering reactions

ln ka
Film
diffusion Pore
diffusion
Slope = 0

Kinetic
–EA region
Slope =
2R

–EA
Slope =
R
Transition Transition
region region

1/T

Fig. 3.23: Temperature dependence of catalytic reactions. (From [2].).

species in the reaction medium. It is preferable to separate the reaction kinetics

from the transfer phenomena and to carry out kinetic experiments under mass
transfer free conditions. Establishing the intrinsic catalytic kinetics and not the ap-
parent mass transfer limited pseudo-kinetics in practice for three-phase systems,
implies that the experiments are carried out at lower temperatures and at lower cat-
alyst loadings with small catalyst particle sizes and with efficient mixing to ensure
negligible influence of various mass transfer phenomena. Such measurements are
preferably carried out for at least three temperature levels to determine reliably ap-
parent pre-exponential factor and activation energy.
For determination of the mass transfer parameters from experimental data the
detailed reactor model that contains kinetic and mass transfer could be used. The
mass transfer parameters are estimated when the kinetic parameters are already
available and are implemented as fixed parameters.

3.5.2 Reactor dependent external diffusion (interphase mass transfer, film

diffusion)

First we will consider an isothermal case when there is only transfer of mass from
the bulk to the external surface of the catalyst and internal diffusion does not play
a role (Fig. 3.24).
 3.5 Mass transfer 287

Bulk phase Nonporous catalyst

cA (reactant A) (reaction of A)

cAb A B
1 1

2 2

3 3
4 4

Fig. 3.24: External mass transfer. 1, no limitations by external mass transport (film diffusion); 2 and
3, external mass transfer and reaction; 4, maximum limitations by external mass transfer.

The molar flux (mol m–2 s–1) J or number of moles i diffusing in the unit of time
per unit of surface in z direction according to the Fick’s law is proportional to the
concentration gradient in this direction:

∂ci
Ji = − D (3:178)
∂z

where D is the diffusion coefficient (m2 s–1), which for binary diffusion in gases in-
creases with increase an in temperature and a decrease in pressure as discussed
below.
The heat flow q (J m–2 s–1) or heat transmitted per unit time per unit of surface
area in z direction is determined analogously to mass flow by the Fourier’s law:

∂T
q= −k (3:179)
∂z

and is proportional to the temperature gradient. In eq. (3.179) k is thermal conduc-

tivity of the fluid.
In the thin (boundary) layer around the solid particle the mass and heat trans-
fer is governed by molecular diffusion and the overall mass transfer is proportional
to the molecular diffusion coefficient. The molar flow rate N (mol s–1) can be ex-
pressed by the following equation:

N = kf Ap ðcb − cS Þ (3:180)

where Ap is the external surface area of the catalyst particle and kf(β) is the mass
transfer coefficient in the film layer surrounding the catalyst particle, cb concentra-
tion at the bulk, cs concentration at the surface. At steady-state, this flux is equal to
the reaction rate in the particle:
288 3 Engineering reactions

N = Vp rv ðcS Þ (3:181)

where Vp is volume of catalyst particle, and rv is the rate per particle volume, which
is related to the rate per unit mass (rw) and particle density (pp) in the following
way: rv = pprw. For a first-order irreversible reaction, eqs. (3.180) and (3.181) give:

Vp 1
kf ðcb − cs Þ = kv cs = kv cs (3:182)
Ap a′
where a’ = Ap/Vp is volumetric external surface area. The concentration at the exter-
nal surface and the expression for the reaction rates can be easily deduced:

kf a′ 1 1
rvobs = kv cs = kv cb = kv c b = kv cb = ηext kv cb (3:183)
k v + kf a′ kv
+1 1 + Da
kf a′

This equation contains the external effectiveness factor ηext which is defined as the
ratio of effective (observed) rate to the intrinsic chemical rate under bulk fluid con-
ditions ηext = 1=ð1 + DaÞ.
In eq. (3.183) Da is the Damköhler number kv kf a′. Large values of Da corre-
spond to strong mass transfer limitations; therefore the observed kinetics in the do-
main of mass transfer is of first-order. For strong external mass transfer limitations
increasing catalyst activity does not influence the rate. Catalyst poisoning and deac-
tivation might have an influence on the observed rate when catalyst activity is de-
creased to such extent that kinetics is becoming the limiting step.
It is apparent that the effectiveness factor depends on the mass transfer coeffi-
cient, which in turn depends on the reactor and hydrodynamic conditions, physical
properties of the liquid, as well as the size of the catalyst grain.
Estimation of the catalyst effectiveness factor in eq. (3.183) requires calculation
of the mass transfer coefficient, which can be done using dimensionless numbers,
such as the Sherwood (Sh), the Schmidt (Sc) and the Reynolds (Re) numbers:

kf d Vd Vdρ v
Sh = , Re = = , Sc = (3:184)
D v μ D

where d is the characteristic length, kf is the mass transfer coefficient, representing

the ratio of total mass transfer to diffusive mass transfer, V is velocity, ρ density, μ
dynamic viscosity and v kinetic viscosity.
The coefficient of mass transfer could be estimated using the mass transfer fac-
tor jD, which is defined in the following way:
 2  2
Sh kf μ 3 kf ρ μ 3
jDðMÞ = = = (3:185)
ReðScÞ3
1
V ρD G ρD
 3.5 Mass transfer 289

where G is the mass flow (G = Vρ). Experimentally established correlations between

the mass transfer factor and the Reynolds number jD = f(Re) are presented in Tab. 3.3
and are graphically illustrated in Fig. 3.25. The mass transfer coefficient kf can be cal-
culated by taking jD from experimental correlations and inserting it into eq. (3.185).
The following procedure could be applied. If G, d and μ are known the Reynolds
number can be computed, followed by calculation of the mass transfer factor, for ex-
ample Fig. 3.25. Finally, by taking the known mass transfer factor and V, μ, ρ and D,
the mass transfer coefficient kf can be calculated.

Tab. 3.3: Mass and heat transfer correlations in packed beds.

Mass transfer Heat transfer Mass transfer factor Range

Gases

0.357 0.641 1 0.428 0.641 1 0.416 < ε < 0.788  < Re < 
Sh = Re Sc3 Nu = Re Pr3
εp εp jD = 0.357Re − 0.359

Liquids

0.25 0.69 1 0.30 0.69 1 0.35 < ε < 0.75  < Re < 
Sh = Re Sc3 Nu = Re Pr3
εp εp jD = 0.25Re − 0.31 . < Re < 
1.09 1 1 1.31 31 1
Sh = Re3 Sc3 Nu = Re Pr3 jD = 1.09Re − 0.67
εp εp

ε, the interparticle void fraction of the bed of particles.

JH
JD
j 0.1

0.01
100 1,000
Reynolds number

Fig. 3.25: Correlation between mass and heat transfer factors and Reynolds numbers.
290 3 Engineering reactions

It follows from eq. (3.185) that kf ∝ jM V. Considering dependence of the mass

transfer factor on the Reynolds number, it can be written that jM ∝ Re − 0.5 ∝ ð1=Vdp Þ0.5 ,
giving the following dependence of the mass transfer coefficient on the velocity and
the diameter of catalyst particles:

kf ∝ Vð1=Vdp Þ0.5 ∝ ðV=dp Þ0.5 (3:186)

Equation (3.186) is demonstrated graphically in Fig. 3.26, showing linear propor-

tionality between the mass transfer coefficient and the square root of (V/dp) at low
values of the latter. Thus, with increasing velocity and diminishing the catalyst par-
ticle size, the impact of mass transfer on the intrinsic catalytic rate could be elimi-
nated. Another way to calculate mass transfer coefficients is to use correlations
between the Sherwood number and the Reynolds and the Schmidt numbers if the
values of Re and Sc are known. The Sherwood number can be calculated using cor-
relations in Tab. 3.3 and therefore the mass transfer coefficient can be determined.

robs

(V/dp )0.5

Fig. 3.26: Dependence of the observed rate on the square root of (V/dp).

From eq. (3.185) it can be shown that kf ∝ ðDÞ2=3 . Temperature dependence of the dif-
fusion coefficient is expressed for diffusion in the gas-phase by a Chapman-Enskog
equation, which gives D ∝ T 3=2 , finally resulting in kf ∝ ðDÞ4=6 ∝ ðT 3=2 Þ2=3 ∝ T and a
very minor temperature dependence of the observed reaction rate with the apparent
activation energy less than 5–10 kJ mol–1.
The diffusion coefficient according to the Chapman-Enskog equation (discussed
in Section 3.5.3) is inversely proportional to pressure, thus mass transfer is becom-
ing more prominent with an increase in pressure. The dependence of the mass
transfer coefficient on pressure is complicated however by the influence of pressure
on other parameters.
Analogously to mass transfer, heat transfer coefficients can be computed,
which are correlated in the same way as mass transfer. In fact, the Sherwood
 3.5 Mass transfer 291

number (the ratio of convective mass transfer rate to diffusive mass transfer rate) is
an analogue of the Nusselt number, which corresponds to the ratio of the total heat
transfer to conductive heat transfer:

hf d
Nu = (3:187)
λ

where hf is the heat transfer coefficient, λ is the thermal conductivity of the fluid.
Tab. 3.3 contains correlations between the Nusselt number and the Reynolds and
Prandtl numbers, where the latter is defined through the specific heat capacity Cp,
thermal conductivity and viscosity Pr = Cpμ/λ. For heat transfer calculations the
heat transfer factor could be also used:
 2
Nu hf C p μ 3
jH = = (3:188)
ReðPrÞ3
1
Cp G λ

For calculation of heat transfer coefficients, the Chilton-Colburn analogy can be ap-
plied. From Fig. 3.25 it can be seen that jD/jH ≈ 0.7 and by knowing the Reynolds num-
ber first the heat transfer factor and finally the heat transfer coefficient are obtained:
 2
1 λ 3
hf = j H Cp 3 G (3:189)
μ

Several correlations have been proposed in the literature for slurry reactors [24] that
relate the dimensionless Sherwood number to the Reynolds and Schmidt numbers.
Data collected for various slurry reactors show that the Sherwood number can be
described by the following equation:
1 1
Sh = 1.0Re2 Sc3 (3:190)

In the case of low Reynolds numbers, dependencies in Table 3.3 should be modified
to contain a correction term, accounting for mass transfer in the absence of stirring:
1
Sh = 2 + 0.4Red 0.5 Sc3 (3:191)

The dependence of the Reynolds number on the local velocity can be established in
terms of the Kolmogorov theory of turbulence. According to this theory, the only
parameter needed to describe the probability distribution of the relative velocity
field Vλ in a homogeneous isotropic turbulent fluid is the average energy of dissipa-
tion ε, i.e.,:
1
Vλ ∝ ðελÞ3 (3:192)
292 3 Engineering reactions

which gives the following expression for the Reynolds number:

 1
εd 3
Red ∝ (3:193)
v3

finally providing a way to estimate the liquid/solid mass transfer coefficients kLS:
!1
εD4 ρ 6
kLS = (3:194)
ηdp 2

where ε denotes the specific mixing power. DoAB is the mutual diffusion coefficient of
solute A in solvent B, ρ is solvent density, η solvent viscosity, and dp is the diameter
of the catalyst particles. A similar equation can be used for gas-liquid mass transfer
with the only difference is that instead of catalyst particle size diameter the size of
bubbles should be introduced in eq. (3.194). In practice, for gas-liquid systems the
gas- liquid mass transfer coefficient is difficult to decouple from the diameter of gas
bubbles, because the latter is determined by the multiphase hydrodynamics. Specific
correlations between kLga (where a is the interfacial area of gas per unit of
volume, m2m–3) and energy of dissipation are used, whose validity is usually limited
thus disallowing extrapolations beyond the ranges of these correlations.
Theoretical calculations of the maximum specific mixing power are based on
the assumption that all of the energy of the impeller is dissipated in the liquid. As
the power of the motor Pstirrer (Watt) is known, the energy dissipation is defined
through it and the liquid mass:

Pstirrer
ε= (3:195)
ρL V L

where ρL and VL are the liquid density and volume.

Typically, not all the power of the motor is dissipated in the liquid therefore it
is advisble to determine the specific mixing power by experimentally determining
the rate of mass transfer. This is done, for example, by dissolution of some solid
materials (e.g. benzoic acid), whose rate is limited by mass transfer. It should be
also noted that the energy is not evenly dissipated throughout the vessel, as it is
much higher close to the impeller and drastically decreasing away from the impel-
ler and leads to even dead zones without any mixing. Visualization of mixing could
be done, for example, by tracer addition or using electrical resistance tomography.
For fixed-bed reactors an experimental way to determine whether external mass
transfer is present is to determine the dependence of conversion α on the flow rate
(Fig. 3.27). If the conversion at a fixed residence time (catalyst mass m divided by the
mass flow FA0 ) in the reactor does not change with changing flow rate (and respec-
tively, the catalyst amount), for example, αA, 1 = αA, 2 at m1 =FA,
0
1 = m2 =FA, 2 (Fig. 3.27A)
0

then external mass transfer limitations can be neglected. Experimentally is it usually

 3.5 Mass transfer 293

αA,1 αA,2 αA,1 jαA,1

m1 m2 m1 m1

(A) F 0A,1 F 0A,2 (B) F 0A,1 nF 0A,1

Fig. 3.27: Elucidation of mass transfer by (A) changing the flow rate and catalyst amount or (B)
changing the flow rate.

easier to keep the catalyst amount the same (Fig. 3.27B). When conversion is linearly
proportional to the residence time in the reactor, it is usually assumed that there are
no external mass transfer limitations.
A common test to verify if mass transport controls a catalytic reaction in three-
phase slurry reactors is to vary the mass of catalyst (gas-liquid mass transfer) and
the rate of agitation (liquid-solid mass transfer). When the reactor productivity rate
is independent on the catalyst mass the observed rate is thus governed by the gas-
liquid mass transfer (Fig. 3.28A).
Reactor productivity (a.u.)

Reactor productivity (a.u.)

5 5

4 4

3 3

2 2

1 1

0 0 1,000 2,000 3,000 4,000

(A) Catalyst mass (B) Stirring speed (rpm)

Fig. 3.28: Experimental verification of (A) gas-liquid and (B) liquid-solid mass transfer.

Agitation of the reaction mixture is usually done by stirring, although reactors in

the form of shakers are also used for laboratory purposes.
The mass transfer coefficient depends on the energy dissipation, which is a
function of mixing power W, often related to the stirring speed according to:

P = Kρliquid N 3 D5 ðstirrerÞ (3:196)

where N is the stirring speed (rev s–1), K is a constant, ρliquid is the liquid density
and Dstirrer is the stirrer diameter. The dependence of the mass transfer coefficient
on the stirring speed can be then established as kf ∝ n1=2 . If the rate is then
294 3 Engineering reactions

independent on the agitation efficiency at a sufficiently high stirring speed

(Fig. 3.28B), then it is usually assumed that mass transport effects are minimized.
However, it should be beared in mind that energy dissipation can have a com-
plex behavior depending on the selection of solvent, stirring rate, presence of baffles,
liquid volume and design of the reactor internals, including impellers. Experiments
with an Autoclave Engineers EZE Seal 0.5 l reactor [25] demonstrated that the dissipa-
tion power at high stirring rates does not follow the square root dependence of mass
transfer coefficient on the stirring speed. The energy of dissipation can also depend
on the liquid load in an autoclave or a shaking reactor [26]. At the higher liquid load,
increasing the shaking frequency does not necessarily lead to improved energy dissi-
pation, therefore the rate of catalytic reactions could be independent on the agitation
because the mass transfer coefficient is the same. As an illustration, Fig. 3.29 shows
the dependence of the power number Po = P/(ρliquidN3D5(stireer)) as a function of the
Reynolds number Re. For laminar flow Po ∝ Re–1 while for turbulent flow it is con-
stant. Such constant values of Po depend however on the impeller configuration, in
particular the number of blades and the blade width, W.

500

1 2 3 4 5 6

100

50
W/D = 1/5 W/D = 1/5 W/D = 1/8 W/D = 1/8 W/D = 1/8 W/D = 1/8
Np

10

5 7 1
2
Glassed steel
3
3-blade retreat 4 5
5
6

7
0.5
1 10 102 103 104 105
Re

Fig. 3.29: Power curves for some typical impellers. (From [27]).

This shows the apparent danger of establishing the kinetic region based exclusively
on the results of catalytic experiments at different stirring speeds, as the indepen-
dence of catalytic activity on the agitation could be explained simply by the fact
that higher stirring speed or shaking frequency does not necessarily influence the
 3.5 Mass transfer 295

specific mixing power. Therefore, experimental verification of the absence of mass

transfer should be combined with calculations.
Note that for lab reactors the stirring speed could be 1,000–2,000 rpm, while in
plant conditions it is usually difficult to obtain sufficient agitator speeds.

When I was a PhD student in the 1980s I visited a caprolactam production plant and had a dis-
cussion with a plant manager. Experiments in the laboratory reactor for one of the process
steps – hydrogenation of benzoic acid – were performed using a kinetic regime. Comparison
with the plant data clearly demonstrated that liquid-solid mass transfer was present in indus-
trial settings and it was proposed that they also increase the stirring speed in the industrial
reactor. The response of the plant manager was negative because 120 rpm was the maximum
technically feasible stirring speed for that plant reactor.

The absence of mass could be verified by applying several criteria, as discussed in

more detail in Section 3.5.8. The criterion of external mass transport can be derived
considering the mass transfer rate equal to the observed rate:
 
4
kLS ðcb − cs Þ 4πR2 = robs πR3 (3:197)
3

and subsequently:

R
cb − cs = robs kLS (3:198)
3

It can be assumed that cb – cs < 0.05cb, so the criterion for the nth order reaction
takes the form:

robs R 0.15
< (3:199)
kLS cb n

3.5.3 Calculation of diffusion coefficients

Reliable calculation/estimation of diffusion coefficients should be done for correct

evaluation of the influence of mass transfer. Different methods have been presented
in the literature [28] to calculate the binary diffusion coefficients in the gas-phase.
The most common is the Chapman-Enskog equation, which is based on kinetic gas
theory:
0.5
ðTÞ1.5 1
MrA + 1
MrB
2 −1 −3
DAB ðcm s Þ = 1.8829 × 10 (3:200)
pðσAB Þ2 ΩAB
296 3 Engineering reactions

Here MrA and MrB are relative molecular masses (dimensionless), p is total pres-
sure (kPa), σAB (nm) is the characteristic length (the Lennard-Jones parameter) for
a pair of molecules, ΩAB is a collision integral and is a function of kBT/εAB, where
εAB (J) is another Lennard-Jones parameter and kB (1.38 × 10–23 J K–1) is the
Boltzmann constant. In the literature values of σAB are often given in Å (10–10 m),
then total pressure should be in 105 Pa, if the units of the diffusion coefficient
are cm2 s–1. The Lennard-Jones parameter, σAB, is calculated from σAB = 0.5 (σA +
σB), while the collision integral depends on the constant εAB, which is calculated
pffiffiffiffiffiffiffiffiffi
for a binary system according to εAB = εA εB . The Lennard-Jones force constants
for some molecules are given in Tab. 3.4 and the values for the collision integral
are presented in Tab. 3.5.

Tab. 3.4: The Lennard-Jones force constants for some molecules.

Substrate g Molar σ, Å ε=kB, k Substrate Molar σ, Å ε=kB, k

mol– mass g mol– mass

H . . . CO . . .

He . . . CO . . .
Ne . . . NO . . .
Ar . . . NO . . .
Kr . . . SO . . .
Xe . . . Cl . . .
air . . . Br . . .
N . . . l . . .
O . . . CH . . .
CH . . . CHCl . . .
CH . . . CHCl . . .
CH . . . CCl . . .
CH . . . CN . . .
i-CH . . . COS . . .
n-CH . . . CS . . .
n-CH . . . CH . . .
n-CH . . . CHCl . . .
Cyclohexane . . .

Typical values of gas-phase diffusion coefficients are ~ DA ~ 10–5m2s–1. Although

values of diffusion coefficients calculated using the Chapman-Enskog equation gen-
erally agree quite well with experimental data, some modifications of this equation
are applied as well as discussed in more specialized textbooks [2, 29] where values
for binary diffusion coefficients are also presented.
For the calculation of binary diffusion coefficients in the liquid-phase DoAB semi-
empirical equations such as Wilke-Chang equation are often used, giving accurate
results for diffusion coefficients of gases in liquids:
 3.5 Mass transfer 297

Tab. 3.5: Values for the collision integral.

kB T =εAB ΩAB kB T =εAB ΩAB

. . . .

. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. .

1
7.4 × 10 − 8 ðΦMB Þ2 T
DoAB = μb 0.6 ½cm2 s − 1  (3:201)
V
cP bðAÞ

In eq. (3.201) the dimensionless association factor f is equal to 2.6 for water, 1.9 for
methanol, 1.5 for ethanol and 1 for benzene, heptane, and other non-associated
solvents, MB is the molecular mass of solvent, μB is solvent viscosity in cP at temper-
ature T[K], Vb(A) is the liquid molar volume at the solute’s normal boiling point
in cm3/mol. Molar volumes of some molecules are given in Tab. 3.6.
The molar volume can be also estimated from the atomic increments given in
Tab. 3.7.

Tab. 3.6: Molar volumes of some molecules.

Methane . Methanol . Acetonitrile .

Propane . Dimethylether . Methyl chloride .
Heptane  Acetone . Phosgene .
Cyclohexane  Acetic acid . Ammonia 
Ethylene . Methyl formate . Water .
Benzene . Ethyl acetate  SO .
298 3 Engineering reactions

Tab. 3.7: Atomic increments.

Atom Increment, cm mol− Atom Increment, cm mol−

C . Br .
H . Cl .
O (except as noted below) . F .
– in methyl esters and ethers . I 
– in ethyl esters and ethers . S .
– in higher esters and ethers . Ring
– in acids . – -membered –
– joined to S,P,N . – -membered –.
N – -membered –.
– double bonded . – -membered –
– in primary amines . – naphthalene –
– in secondary amines . – anthracene –.

3.5.4 Size dependent internal (pore) diffusion

In a porous catalyst particle, the reacting molecules diffuse first through the fluid
film surrounding the particle surface and then diffuse into the pores of the catalyst
to the active sites. In a similar way the reaction products are diffusing out of the
catalyst grains. As an outcome of pore diffusion for most common reaction kinetics
the reaction rates inside the pores have lower values than expected with the con-
centration levels of the main bulk.
The mathematical treatment of pore diffusion is based on the general conserva-
tion laws. The amount of mass or heat transported per unit of time into and out of a
differential volume element is balanced by accumulation in this volume and gener-
ation or consumption within the same volume

IN + GENERATION = OUT + ACCUMULATION

The complex diffusion process could be represented by Fick’s first law, where for
the species i the flux Ni (mol m2 s–1) is proportional to the concentration gradients
of the components (dci /dr) and its effective diffusion coefficient (Dei): Ni = –De(dc/
dx). In the calculation of the diffusion in a porous media the random pore model
and the effective diffusion coefficient is applied. The latter is defined as De = D(ε/τ)
which is smaller than the diffusion coefficient because the diffusional cross section
is smaller than the geometric cross section (thus porosity ε is introduced) and the
catalyst has irregular pore structure (expressed via tortuosity τ) as illustrated in
Fig. 3.30. Typically ε /τ = 0.05 ÷ 0.2
 3.5 Mass transfer 299

Fig. 3.30: Illustration of porosity and tortuosity.

Pore diffusion may occur by molecular diffusion, Knudsen diffusion and surface
diffusion; the surface diffusion usually not seriously influencing catalytic kinetics.
In the gas-phase and within large pores, molecular diffusion prevails when the
molecules collide with each other (Fig. 3.31A).

(A) (B)

Fig. 3.31: (A) Molecular and (B) Knudsen diffusion.

In small pores or when the gas density is low, it is not only molecular but also
Knudsen diffusion (Fig. 3.31B) that could play an important role. The mean free
path becomes comparable to the size of the pore and molecules are colliding with
the walls rather than with each other. Such diffusion is thus independent of pres-
sure and is not observed in the liquid-solid catalytic reactions when the fluid den-
sity is much higher compared to the gas-solid catalysis.
The Knudsen diffusion coefficient is proportional to the pore radius re and the
mean molecular velocity, giving proportionality of the Knudsen diffusion coefficient
to the square root of temperature:
rffiffiffiffiffiffiffiffiffiffi
2 8 RT
DK = re (3:202)
3 πM

where R is the gas constant, T is the absolute temperature and M is the molecular
mass.
As the pore radius in the case of cylindrical pores can be calculated from the
total pore volume Vp(cm3 g–1) and the total surface area S (cm2 g–1), (i.e., re = 2Vp /S)
the Knudsen diffusion coefficient for a porous solid becomes:
rffiffiffiffiffiffiffiffiffiffi
8εp 2 RT
Dk = (3:203)
3Sρp π M

where εp is porosity and ρp is the particle density.

300 3 Engineering reactions

A special case of diffusion is the so-called configurational diffusion in zeolite

pores, where the pore size is comparable with the size of reacting molecules. Steric
and diffusional constraints in zeolites are the origin of their shape-selective proper-
ties. Comparison of the values of diffusion coefficients for different cases (configu-
rational, Knudsen and molecular diffusion) is given in Fig. 3.32. Such comparison
should be taken with care because transition between different types of diffusion
depends not only on the size of pores but also on the process conditions.

Fig. 3.32: Values of diffusion coefficients depending on the pore diameter [30].

If both molecular DAB and Knudsen DKA diffusion are present (for example, the cata-
lyst contains broad distribution of pore radii), the Bosanquet approximation is
used:
1 1 1
= + (3:204)
D DAB DKA

where the first term stems from molecular diffusion and the second from Knudsen
diffusion.
Reaction and diffusion will be first considered for an isothermal case with only
internal mass transfer control (Fig. 3.33).
The main aim of the mathematical treatment is to obtain an expression for the
effectiveness factor η = robs =rkinetics , which relates the observed reaction rate to the
intrinsic chemical rate. If the value of the effectiveness factor is within 0.95 ÷ 1 then
internal diffusion could be neglected. As concentration changes with the position
 3.5 Mass transfer 301

cs
c Fig. 3.33: Reactant concentration profiles (A) external
c and internal mass transfer control, (B) internal mass
(A) (B) transfer control only.

inside the catalyst grain the reaction rate for non-zero order reactions is thus also a
function of the position. Therefore, the concentration profiles should be obtained
first.
Considering a pseudo-homogeneous model and the first law of Fick for a differ-
ential element of a spherical particle (Fig. 3.34) the following mass balance can be
written:

Fig. 3.34: Illustration of a differential element of a spherical particle.

dc dc
Ax + dx De  − Ax De  = rAdx dx (3:205)
dx x + dx dx x

where A is the geometrical surface area. As dx is very small, x + dx = x and eq. (3.205)
is rearranged to:

dc dc
4πx2 De  − 4πx2 De  = 4πx2 rdx (3:206)
dx x + dx dx x

A simple manipulation for the nth order kinetics gives:

d2 c 2dc kcn
+ = (3:207)
dx2 xdx De
Introducing dimensionless concentration y = c/c0 and the position inside the pore
x* = x/L(R), eq. (3.207) is rewritten in a simple form:

d2 y 2 dy
+ = ϕ2 y (3:208)
dx*2 x* dx*
where ϕ is the Thiele modulus:
302 3 Engineering reactions

sffiffiffiffiffiffiffiffiffiffiffiffi
kcn − 1
ϕ=R (3:209)
De

The analytic solution of differential eq. (3.209) is well-known. The boundary condi-
tions reflect symmetry at the pellet center, giving dy/dx * = y′(0) = 0. Moreover at
the external pellet surface the surface concentration is equal to the bulk substrate
concentration y(1) = 1. The change of the dimensional concentration across the pel-
let radius is thus:

ðeϕx* − e − ϕx* Þ sinhðϕx*Þ

y= = (3:210)
eϕ − e − ϕ sinhðϕÞ

Despite the simplicity of this case is it of great practical importance, because in many
cases only small heat consumption or release is observed and the catalyst particle
could be treated as an isothermal one. Normalized concentration profiles depending
on the position inside the spherical particle are presented in Fig. 3.35 and it is appar-
ently clear that for small values of Thiele modulus the reaction rate has a less steep
decline of concentrations along the pore, while for large values of Thiele modulus the
reactant concentration diminishes dramatically, indicating that in fact catalytic reac-
tions occur only in the vicinity of the catalyst outer layer. Practical implementation of
this conclusion for the design of catalyst particles used in methane steam reforming
will be discussed in the corresponding section in Chapter 4.
During the stationary operation the amount of mass converted is equal to the
flux across the external surface:

1.0
Φ = 0.5
Φ=1
0.8

Φ=2
0.6
c/cs
Φ=3
0.4

Φ=5
0.2
Φ = 10

0.0
0.0 0.2 0.4 0.6 0.8 1.0
x*

Fig. 3.35: The concentration profile dependence inside a spherical particle at different values of the
Thiele modulus for isothermal, first-order irreversible reaction.
 3.5 Mass transfer 303

4 3 dc
πR rave = 4πR2 De  (3:211)
3 dx x = R

and the average rate is:

3 dc
rave = De  (3:212)
R dx x = R
and subsequently the catalyst effectiveness factor takes the form:

dy 
3De dx* 
x* = 1 3 dy 
η= =  (3:213)
R2 kcbn − 1 ϕ2 dx* x* = 1

By differentiating eq. (3.210) and substituting the result into eq. (3.213) the effective-
ness factor η of a first-order irreversible reaction in a spherical pellet is obtained as
a unique function of the Thiele modulus:
 
3 1 1
η= − (3:214)
ϕ tanhðϕÞ ϕ

Equation (3.214) clearly indicates that the Thiele modulus and the effectiveness factor
depend strongly on the size of catalyst grains. At small values of the Thiele modulus
(i.e., small particle size) the effectiveness factor is approaching unity. The flat depen-
dence of effectiveness factor on the Thiele modulus occurs at small catalyst particles,
low catalyst activity (small k), and large pore size and high porosity (large De). When
φ >>3, the following dependence is valid η ∝ 1=ϕ and the effectiveness factor is in-
versely proportional to the Thiele modulus and thus to the particle size. The asymp-
totic value for the effectiveness factor is η = 3=ϕ as limðφ ! ∞Þ tanh φ = 1. For large
values of Thiele modulus the overall rate is controlled by pore diffusion and for very
active catalysts, or for catalysts with small pores, low porosity and/or large diameter
the reactant concentration approaches zero in the center of a particle.
Obviously, in laboratory-scale reactors the size of catalyst particles can be
rather small to diminish the impact of internal diffusion, while in fixed-bed pilot or
industrial reactors the size of catalyst grains is unavoidably much higher because
of increased pressure drop, resulting in a significant influence of internal diffusion.
For slurry reactors, even at the pilot stage the size of catalyst powder could be still
in the range of 50–100 μm, which in most cases (i.e., when catalytic reactions are
not very fast) is sufficient to eliminate internal diffusion. However, external diffu-
sion limitations can still play a role.
The analysis above was done for a spherical particle and the first-order kinetics.
Analytical solutions are available for some isothermal steady-state systems and
ideal geometric forms (slabs and spheres) and simple kinetics.
The treatment could be extended for arbitrary shapes typical for heterogeneous
catalysis (Fig. 3.36) by introducing a generalized Thiele modulus, which is related
to the ratio of the pellet volume to the external pellet surface:
304 3 Engineering reactions

Fig. 3.36: Typical shapes of industrial heterogeneous catalysts.

sffiffiffiffiffiffiffiffiffiffiffiffi
Vp kcn − 1
ϕp = (3:215)
Ap De

where Vp/Ap = L, the characteristic length of diffusion. The effectiveness factor for
different geometries for an isothermal, first-order irreversible reaction is presented
in Fig. 3.37, demonstrating that the particle geometry is of less importance for the
effectiveness factor.
Fig. 3.37 (presented as a double logarithmic plot) illustrates that for the val-
ues of Thiele modulus above four to five, dependences of the effectiveness factor on
the Thiele modulus almost coincide independent to the pellet geometry. Therefore,
for evaluation of the influence of internal mass transfer, even the simplest case [a
porous catalyst slab (flat plate), for which the catalyst effectiveness factor is given
by η = tanh(ϕ)/ϕ], could be used.

Slab
Effectivenes factor

0.5 Cylinder

Sphere

0.1
0.1 1 10
Thiele modulus

Fig. 3.37: Effectiveness factor η as a function of the generalized Thiele modulus φp for different
pellet geometries.
 3.5 Mass transfer 305

For evaluation of the importance of internal diffusion in the literature it is

mainly the expressions for first-order kinetics that are used. In general, for arbitrary
kinetics a numerical solution of the balance equation that takes into account the
boundary conditions is necessary.
Plots for a family of curves (irreversible Langmuir kinetics) are presented in
Fig. 3.38, ranging from the zero order kinetics (denominator KcAs is much larger
than unity and approaches infinity) to first-order kinetics ðKcAs ≈ 0Þ.

1
∞ (zero order)
100
10
0.1
0 (1st order)

0.1
0.1 1 10
kKCAS
Φ=L
Dei(1 + KCAS)

Fig. 3.38: Effectiveness factor as a function of Thiele modulus for different kinetic expressions [32].

As is clear from Fig. 3.38, the catalyst effectiveness factors can be obtained for
Langmuir (Eley-Rideal) kinetics by interpolating between 0th and first-order kinet-
ics. For some more complicated Langmuir-Hinshelwood rate equations negative
order could be observed, because the numerator can be of a lower order than de-
nominator. For such reaction kinetics, the effectiveness factor can even exceed
unity, because the reaction rate increases along the pore length as the reactant con-
centration inside the particle becomes lower than in the bulk phase. Detailed analy-
sis of the effectiveness factor for complex rate expressions is given in the review by
Dittmeyer and Emig [31].
It is interesting to note that kinetic regularities and the values of activation
energy are different in the regime, which is influenced by internal mass transfer.
For the nth order reaction, when the catalyst effectiveness factor is inversely pro-
portional to the Thiele modulus, an expression for the observed reaction rate is:
pffiffiffiffiffiffi
1 n 1 De 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
robs = ηr = kC = kC pffiffiffiffiffiffiffiffiffiffiffiffi =
n
kDe Cn + 1 (3:216)
φp L kC n − 1 L
306 3 Engineering reactions

Analysis of eq. (3.216) demonstrates that the observed reaction rate is inversely propor-
tional to the characteristic length of diffusion and has the reaction order (n + 1)/2.
Moreover observed activation energy has half the value of the corresponding to the
true kinetics activation energy (Fig. 3.23) if neglecting the minor influence of temper-
ature on the effective diffusion coefficient. Therefore, mechanistic conclusions
based solely on kinetic observations obtained in the internal diffusion regime
could be false, as the observed reaction orders for zero and second order reac-
tions are 0.5 and 1.5, respectively. For the first-order kinetics the values of re-
action orders are the same in kinetic and internal mass transfer regimes.
When internal diffusion is combined with the film diffusion the catalyst effec-
tiveness factor can be extended to:

 
3 1 1 ′ 1
η = ηint next = − (3:217)
ϕ tanhðϕÞ ϕ 1 + ϕ 1
− 1
BiM tanhðϕÞ ϕ

which gives the overall effectiveness factor for first-order reactions in a spherical cata-
lyst particle. In eq. (3.217) the Biot number for mass transport BiM is defined as the
ratio between the diffusion resistance in the fluid film and in the catalyst particle:

kf R
BiM = (3:218)
Dei

Usually BiM >>1 for porous catalyst particles and internal diffusion resistance prevails.
The practical difficulties in using the eq. (3.217) are associated with the necessity to
apply the Thiele modulus, which contains the rate constant, which might be un-
known. The practical way to solve this problem is to use the Weisz modulus ψ, which
only contains observable parameters and is defined as the ratio of effective (observed)
reaction rate to the maximum effective rate of diffusion at the external surface:

robs R2
ψ= = ϕ2 η (3:219)
cs De

The effectiveness factor as a function of the Weisz modulus for different Biot num-
bers is demonstrated in Fig. 3.39. The effectiveness factor is diminished when the
Biot number is decreased, or in other words when external diffusion becomes more
prominent. In practice the values of the Biot numbers are typically between 100
and 200 thus the influence of internal (pore) diffusion is more important than the
impact of interphase (external) mass transport.
 3.5 Mass transfer 307

Bim = 0.1 Bim = 106

Effectiveness factor

Bim = 1
0.1

Bim = 10

Bim = 100
0.01
0.01 0.1 1 10 100
Weisz modulus

Fig. 3.39: Effectiveness factor η as a function of the Weisz modulus ψ for an isothermal, first-order
irreversible reaction in a sphere.

3.5.5 Non-isothermal conditions

For some heterogeneous catalytic reactions (oxidations, hydrogenations, dehydro-

genations) substantial consumption or release of heat results in non-isothermal
temperature profiles inside the catalyst particle and in the film surrounding the par-
ticle (Fig. 3.40).
In general, the mass and the heat balances should be solved using the corre-
sponding boundary conditions in order to get the concentration and temperatures
profiles. The non-linear character of differential equations leads to a need for nu-
merical solutions.

Gas film

Ts
T Ts T

c cs cs c
Bulk gas

Exothermic Endothermic

Fig. 3.40: Temperature and concentration gradients in the gas film and catalyst particle. (From [23].).
308 3 Engineering reactions

When interphase heat and mass transfer could be neglected the effectiveness
factor depends on the Thiele modulus, the Prater and the Arrhenius numbers de-
fined below. The Arrhenius numbers appear during the considerations of the heat
and mass balances. The heat fluxes q are calculated according to the Fourier’s law,
relating the heat flux through a surface with the temperature difference along the
path of heat flow:

dT
q = − λe (3:220)
dx

Another form of representing the Fourier law has been already presented (eq. 3.179)
with just another notation also frequently applied in the literature. In eq. (3.220) the
heat conductivity in the pores λe [k in eq. (3.179)] can be computed from:

λe = ð1 − εÞλsolid + ελg (3:221)

where λsolid, λg are heat conductivity of the solid and gas in the pores, respectively,
and ε is the porosity. As the heat conductivity of solid materials is several orders of
magnitude higher than the conductivity of gases, the second term in eq. (3.221) can
be neglected. The values of heat conductivities for industrial catalysts are ~10–1–10–3
Watt m–1 K–1.
When the interphase heat and mass transfer is neglected for a spherical parti-
cle, in addition to the mass balance equation in the particle (3.205), a heat balance
equation for particles also needs to be considered, which defines the difference be-
tween the heat flows in and out of a differential shell as equal to the heat produc-
tion within the shell because of a catalytic reaction of nth order:
 
dT  dT
− Ax + dx λe  − − Ax λe  = kcn ð − ΔHÞ4πx2 Δx (3:222)
dx x + dx dx x

The system of molar mass balances and the energy balance can be now written as:

d2 c 2 dc kðTÞcn d2 T 2 dT kðTÞcn
+ = , 2 + = ð − ΔHÞ (3:223)
dx2 x dx De dx x dx λe
with the boundary conditions: x = R, c = cb, T = Tb (surface temperature is equal to
the temperature in the bulk and the surface concentration is equal to the bulk con-
centration when external heat and mass transfer is absent) and x = 0, dT/dx = dc/
dx = 0. This system is strongly coupled through concentrations in the reaction rate
expressions and through the exponential temperature dependencies of the rate con-
stants that require numerical solutions.
If introducing dimensionless concentration y = c/cb; positions inside the pore
x* = x/L(R), and temperature; Θ = T/Tb eqns. (3.223) are rewritten as:
 3.5 Mass transfer 309

  
d2 y 2 dy 1
+ = φ2 n
y exp γ 1 − , (3:224)
dx*2 x* dx* Θ

and
  
d2 Θ 2 dΘ 1
+ = β φ2 n
y exp γ 1 − (3:225)
dx*2 x* dx* Pr
Θ

with the boundary conditions x = 1, y = 1, Θ = 1 and x = 0, dΘ/dx* = dy/dx* = 0 and

the dimensionless Thiele, Prater (βPr) and Arrhenius (γ) numbers:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kðTÞcb n − 1 ð − ΔHÞDe cb Ea
ϕ=R , βPr = ,γ= (3:226)
De λe Tb RTb

The Arrhenius number is a measure for temperature sensitivity, while the Prater
number is the normalized maximum temperature difference between the pellet cen-
ter and bulk βPr = ΔTmax/Tb. For exothermal reactions values of βPr are positive,
rarely exceeding values of 0.1, which means that at reaction temperatures of
500–600 K the maximum temperature difference between the fluid bulk and the
particle center is 50–60 K. Fig. 3.41 contains some numerically obtained typical
curves for the effectiveness factor as a function of the Thiele modulus. As can be

10.0

Exothermic
β = 0.3 reaction
E 0.4
γ= = 20
RT 2.0

0.1

1.0

Endothermic 0
reaction 0.2
β = 0.4

0.1
1.0 10.0 100.0
Thiele modulus

Fig. 3.41: The catalyst effectiveness factors as a function of Thiele modulus at different values of
the Prater numbers.
310 3 Engineering reactions

seen from this figure, for exothermic processes the effectiveness factor can even ex-
ceed unity, because of a temperature rise inside the particle and increased values of
the rate constants, which is not overcompensated by the lower concentrations in-
side the pellet because of the diffusion. This effect is particularly visible at small
values of the Thiele modulus. Another interesting feature under non-isothermal
conditions and strongly exothermic reactions is steady-state multiplicity, even for
the simplest first-order reaction. It implies that there are several possible solutions
for the effectiveness factor, with the middle one usually unstable. For the typical
range of Arrhenius numbers (10–30) multiple steady states are possible only at
large Prater numbers (highly exothermal reactions, low thermal conductivity). As
mentioned above, for industrial catalysis the values of the Prater number are below
0.1 and multiplicities of the effectiveness factor are not important, although the val-
ues of the effectiveness factor in the real systems could be above unity.
The effective heat conduction with the catalyst particle is typically not that crit-
ical and the dominating part of the overall heat resistance is in the external bound-
ary layer, but not inside the catalyst pellet. The mathematical treatment in this case
is complicated and discussed in specialized monographs.
The effectiveness factor for this combined case is presented as a function of the
observable Weisz modulus in Fig. 3.42 with a modified Prater number:

100

10
β =1
Effectiveness factor

0.75
1
0.5

0.1 0.25
γ =20
Bim=100 0
β =–1
0.01
0.1 1 10 100
Weisz modulus

Fig. 3.42: Effectiveness factor η as a function of the Weisz modulus Ψ for combined intraparticle
mass transfer and interphase heat and mass transfer and the first-order irreversible reaction in a
sphere.

2
ð − ΔHÞcb Pr3
β* Pr = 2
(3:227)
ρcp Tb Sc3
 3.5 Mass transfer 311

where ρ is density, cp is heat capacity, Sc and Pr are, respectively, the Schmidt and
the Prandtl numbers. In a similar way to the previous considerations for the large
values of the Weisz modulus and exothermal reactions (β*Pr > 0) the effectiveness
factor can exceed unity, as the decline in concentrations is compensated by the
temperature increase. For higher Arrhenius numbers the increase of the effective-
ness factor above unity will be even more pronounced as the reaction rate will pos-
sess stronger temperature dependence.

3.5.6 Multiple reactions and diffusional limitations

It is not only catalytic activity but also selectivity that can be influenced by mass
transfer phenomenon. Analytical treatment is available for simple reactions and as
the complexity increases, numerical treatment becomes necessary. In the relevant
literature it is typically independent reactions (less commonly observed in practice,
although interesting from the theoretical viewpoint), consecutive and parallel reac-
tions that are treated. The consecutive and parallel reactions will be considered here.
For a consecutive reaction sequence A ) B ) C in a batch reactor when the
rate equations are of the first-order, the differential selectivity is obtained from:

dcB kA cA − kB cB kB cB
s= − = =1− (3:228)
dcA kA cA kA cA

An analytical solution of (3.228) can be readily obtained that provides expressions

for the concentration of the intermediate CB and selectivity SB:
!  M − 1 !
1 cM M cA
cB = A
−
− cA , sB = 1 − −1 (3:229)
1 − M ðc Þ0 M 1 1 − M c0A
A

where c0A
is the initial concentration of the reactant at which c0B =0 and M = kB/kA.
The mass balances for pore diffusion with a slab geometry and a characteristic
dimension L for the dimensionless concentrations y are:

d2 yA d2 y B
2
= ϕ2 yA , = − ϕ2 yA + γ2 ϕ2 yB (3:230)
dx dx2
Where:

kA 2 kB De, A
ϕ2 = L2 ,γ = (3:231)
De, A kA De, B

The solution of eq. (3.231) results in the concentration profiles, first making it possi-
ble to calculate the rates:
312 3 Engineering reactions

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi bulk
rAbulk = kA De, A cA tanh ϕ (3:232)

pffiffiffiffiffiffiffiffiffiffiffiffiffi bulk pffiffiffiffiffiffiffiffiffiffiffiffiffiffi tanh ϕ − γ tanhðϕγÞ

rBbulk = kB De, B cB tanhðϕγÞ − kA De, A cbulk (3:233)
A
1 − γ2

and then the selectivity towards the intermediate product

tanhðϕγÞ
rBbulk 1 − γ tanh ϕ γ tanhðϕγÞ De, B cbulk
s= = − B
(3:234)
rAbulk 1 − γ2 tanh ϕ De, A cbulk
A

Differential selectivity in consecutive reactions depends on thepvalues of the Thiele

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
modulus ϕ and parameter γ. The higher is the value of γ = kB De, A =kA De, B , the
more pronounced is the influence of diffusion resulting in lower selectivity towards
the intermediate B (Fig. 3.43). This is an important conclusion – that internal diffu-
sion limitations, prominent in industrial conditions because of the large size of cat-
alyst pellets, lead to diminished selectivity towards the intermediate product in
comparison with the intrinsic kinetic conditions.

1.0
1
2
0.8
s 10

0.6
γ = 100

0.4
0.1 0.2 0.5 1 2 5 10
ϕ

Fig. 3.43: Dependence of selectivity for consecutive reactions on Thiele modulus at different values
of γ. (Redrawn from [33].).

For parallel reactions A ) B1 and A ) B2 the differential equation describing the

concentration profile of component A in the isothermal case is:

d2 yA 1
= ϕ21 ðf1 ðyA Þ + f2 ðyA ÞÞ (3:235)
dx 2
γ′

where:
 3.5 Mass transfer 313

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u ðcbulk Þ
uk1 r1 Abulk
t c k1 r1 ðcbulk
A Þ
ϕ1 = L A
, γ′ = (3:236)
De, A k2 r2 ðcbulk
A Þ

If the reaction orders are different, for instance the reactions are of the nth and
the mth order, respectively f1 ðc′A Þ = ðc′A Þn ; f2 ðc′A Þ = ðc′A Þm , then the concentration
profile of component A is:
!
d2 yA 2 n ðyA Þm
= ϕ1 ðyA Þ + (3:237)
dx2 γ′

with:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n−1
k k1 ðcA bulk Þ
γ′ = ðcA bulk Þn − m ;ϕ1 = L
1
(3:238)
k2 De, A

The ratio of the selectivity towards the product B1 in the internal diffusion regime to
the selectivity under kinetics when γ′ ≪ 1 is:
sdiffusion m+1
= (3:239)
skinetics 2n − m + 1

implying, that when the reactions are of the same order, the differential selectivity
is independent on the presence of internal diffusion. If the desired reaction is of
lower order (for example, n < m), then it is preferential to conduct the reaction in
the diffusion region, as the highest penalty in the case of pore diffusion limitations
is on the highest order reaction. Under non-isothermal conditions changes in appar-
ent selectivity caused by temperature variations inside the pellet are expected when
the activation energies are different irrespective of the reaction orders.
It is possible that one of the reactions occurring in the system is an undesired
one, for example poisoning. For the slab geometry the diffusion model is:

d2 y
= ϕ2 rðyÞa (3:240)
dx2
where a is the relative activity (or fraction of unpoisoned sites). For uniform poison-
ing and the first-order kinetics, eq. (3.240) takes the form:

d2 c′
= ðϕ2 aÞc′ (3:241)
dx2
Apparent activity is defined as the ratio of the effectiveness factors for the poisoned:
pffiffiffi
tanhðϕ aÞ
η poisoned
= pffiffiffi (3:242)
ϕ a
314 3 Engineering reactions

and unpoisoned catalyst under conditions of pore diffusion limitations:

pffiffi
tanhðϕ aÞ
pffiffi pffiffiffi pffiffiffi
ηpoisoned ϕ a a tanhðϕ aÞ
χ = unpoisoned = = (3:243)
η tanhðϕÞ tanhðϕÞ
ϕ

pffiffiffi pffiffiffi
For negligible diffusion limitations χ = a because tanh ϕ a ≈ ϕ a; tanhðϕÞ ≈ ϕ.
Conversely, when intraparticle mass transport is significant, the Thiele modulus is
pffiffiffi
large and χ = a (Fig. 3.44), which means that poisoning is less prominent in the
case of internal diffusional limitations.

1.0

LLarge
arge TThiele
hiele
Fraction of initial activity

0.8
m
modulus
odulus

0.6

Small TThiele
Small hiele
0.4
m
modulus
odulus

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
Activity

Fig. 3.44: Behavior of apparent activity for small and large Thiele moduli.

3.5.7 Diffusion in micropores

Mass transfer in micropores (configurational diffusion) has a complex behavior and

is influenced by the interactions of the diffusing molecules with the walls of the
pores as well as adsorbate-adsorbate interactions.

As a result, the diffusivity in this case (which could be anisotropic) is much lower
than for molecular or Knudsen diffusion (Fig.3.32). There could be a great variation
of several orders of magnitude in the diffusivity of molecules with not very different
 3.5 Mass transfer 315

kinetic diameters. For mixtures caused by the similar size of the reacting species
and the pores, diffusion of fast reacting species could be delayed.
In case of catalysis on zeolites, diffusion is sometimes described based on the
adsorbed phase concentration, while the reaction rate is based on the bulk gas-
phase concentration or pressure. Subsequently, an adsorption constant K is added
to the Thiele modulus to maintain its dimensionless character
sffiffiffiffiffiffiffiffiffiffiffi
k
φ=L (3:244)
KDeff

The practical consequences of configurational diffusion are shape-selective proper-

ties of materials containing micropores, such as zeolites, as discussed in Chapter 1.
In fact, reactant and product selectivity is a result of these mass transfer effects.
Modeling of shape selectivity effects is a complicated task and requires ad-
vanced numerical treatment. Molecular dynamics, stochastic and continuous mod-
els are applied. For molecular dynamics models a solution is needed for Newton’s
laws of motion for the molecules in the zeolites, which calls for the use of quantum
mechanics. Stochastic or probabilistic models (such as Monte Carlo methods) use
rather simple model reactions. Finally, continuous models are based on the conser-
vation equations for mass and energy, which are certainly valid for macropores,
while their applicability could be questioned for micropores when only few mole-
cules are present within the pore cross section.

3.5.8 Criteria for the absence of diffusional limitations

In order to clarify the influence of mass transfer it is possible to calculate the mass
transfer coefficients and catalyst effectiveness factors as explained above. Such
analysis requires calculations of the diffusion coefficients and elucidation of several
other physical and materials properties. Interested readers could consult the relevant
literature [25, 34, 35] where details of calculations are provided.
For internal diffusion, analytical solutions can only be used directly for simple
kinetics such as first or zero orders, thus approximations should be applied because
many reactions are not of zero or first-order. Alternatively, it is possible to use the
generalized diagrams by Aris [36] (Fig. 3.45). Such diagrams relate various kinetics
effectiveness factors and the generalized Thiele modulus:

ϕ
ϕ* = (3:245)
R1
ð2 r′dyÞ0.5
0

where y is the relative concentration and r’ is the reaction rate for relative concen-
r
trations r′ = .
rðcb Þ
316 3 Engineering reactions

10

p=
–5
Effectiveness factor

–4
–3
–2
–1
–0.6
p = –0.78 –0.33
1
–0.1

p = 0.1
0.5
1
2
5

0.1
0.1 1 10
Normalized Thiele modulus

Fig. 3.45: Isothermal effectiveness factor for reactions of pth order.

For example, for a second order reaction:

sffiffiffiffiffiffi rffiffiffirffiffiffiffiffiffiffiffiffiffiffiffi
ϕ 3 3 robs
ϕ* = R1 = = R (3:246)
2ϕ 2 cs De η
ð2 y2 dyÞ0.5
0

The intrinsic rate is evaluated from kinetic experiments for small catalyst particles
(when the effectiveness factor is unity). The calculated value of the effective diffu-
sion coefficient and the initial bulk phase concentration of the reactant (neglecting
external diffusion and supposing that surface and bulk concentrations are equal)
are inserted in eq. (3.246) and give a dependence of the generalized Thiele modulus
on the effectiveness factor and the radius of the catalyst pellet. Then for a large cat-
alyst particle size the generalized Thiele modulus is calculated by first setting the
effectiveness factor equal to unity. The next step in the iteration loop is to check the
value of effectiveness factor from the diagrams (Fig. 3.45) for the respective reaction
order (2 for example).
This value of the effectiveness factor is further applied to recalculate the gener-
alized Thiele modulus, followed by another estimation of the corresponding effec-
tiveness factor from the plots. This iteration procedure is applied until convergence
is reached.
Dittmeyer and Emig [31] proposed and reviewed some theoretical criteria for the
absence of interphase (external) and intraparticle (internal) transport for simple
 3.5 Mass transfer 317

reactions with arbitrary kinetics. Such theoretical criteria contain explicit expres-
sions of the reaction rates and are more difficult to use than experimental criteria.
Because of the complexity of using these criteria and a possibility of simultaneous
concentration and temperature gradients in the catalyst grain, in order to make the
theoretical approach more sound it could even be recommended that the level of
complexity be increased and a numerical solution of the balance equation for the
particle be generated.
Apparent complexity of the procedures described above led to the development
of alternative approaches based on various experimental criteria, which became
rather popular and are often applied to determine if heat and mass transfer are
influencing a process under investigation. Such criteria are usually derived in a
way that deviations of the observed rates from the ideal rates are not more than 5%.
robs
η= = 1 ± 0.05 (3:247)
rðTb , cb Þ

The advantage of applying different criteria is that observed values of reaction rates
are used, while the kinetics parameters (requiring calculation of, for example,
Thiele modulus) could be still unknown, at least at the initial stage of a process in-
vestigation. The approximate nature of such criteria calls for a conservative ap-
proach in their use. A decision on the possible influence of mass transfer could be
made when the calculated value is significantly different from the limiting value,
otherwise an uncertainty remains. The criterion for external heat transfer was pro-
posed by Mears [37]:

jΔHr jrobs R Rg Tb 0.15

χ= < 0.15 = (3:248)
hTb EA γ

where h is the heat transfer coefficient, Tb is the bulk temperature, ΔHr is the reaction
enthalpy, EA is the activation energy and Rg is the universal gas constant. The criteria
of external mass transport [see also eq. (3.197) for the derivation with kLS = kf] for a
simple irreversible reaction of nth order (n ≠ 0) is:

robs R 0.15
w= < (3:249)
kf cb n

and for a combined intra/interphase heat and mass transfer are also proposed by
Mears [38]:

robs R2 1 + 0.33γχ
< (3:250)
De cb ðn − γβPr Þð1 + 0.33nwÞ

where γ and βPr are the Arrhenius and the Prater numbers, χ and w are defined,
respectively, by eqns. (3.248) and (3.249). It is apparent that the application of the
latter criteria is not straightforward, as it requires knowledge of the Arrhenius
318 3 Engineering reactions

number and thus the activation energy. However, activation energy is usually deter-
mined from the studies when the influence of mass and heat transfer is yet to be
determined.
Internal heat transport limitations were considered by Anderson [39] leading to
the following conditions when such limitations do not occur:

ΔjHr jrobs R2 Rg Tb 0.75

< 0.75 = (3:251)
λTs EA γ

where λ is the thermal conductivity.

The most widely applied criterion is probably the one for internal mass transfer
limitations in an isothermal catalyst particle, that is, for pore diffusion. Following
Weisz and Prater [40] no pore diffusion limitation occurs, if the Weisz modulus de-
fined by eq. (3.219) is:
8
> < 6, n = 0
2<
robs R
ψ= < 0.6, n = 1 (3:252)
cs De >
:
< 0.3, n = 2

Recently I reviewed a scientific paper where the calculation of the criterion for pore diffusion re-
sulted in a value of 0.29, while for this particular reaction with kinetic order equal to 2 it should
be below 0.3. Based on the approximate calculations of the criterion (which also required calcula-
tions of an effective diffusion coefficient and thus an estimation of porosity and tortuosity) the
authors made a premature and possibly incorrect conclusion that their data are not influenced by
mass transfer.

Strictly speaking, such criterion is applied for the spherical particles. For other geom-
etries, the specific radius could be replaced by the pellet volume Vp, divided by the
pellet surface area Ap. According to Villermaux [41] internal mass transfer is negligi-
ble when:

robs Vp2
< 0.1 (3:253)
cs De Ap 2

According to Weisz and Hicks [42], for a non-isothermal catalyst particle no internal
heat and mass transport limitations occur if:
 
γβPr
ψ exp <1 (3:254)
1 + βPr

As the accuracy is questionable in the calculations of the various parameters in the

criteria presented above, such as heat, mass transfer and diffusion coefficients, poros-
ity, tortuosity, and particularly for bimodal pore size distribution, experimental tests
are also conducted in addition to calculations of criteria for heat and mass transport.
 3.5 Mass transfer 319

As explained above an experimental way to determine whether external mass

transfer limitations are present in flow reactors is to measure conversion dependence
on the flow rate. At low Reynolds numbers the heat and mass transfer coefficients
could be, however, insensitive to changes in the flow velocity. For the slurry systems
experiments with different catalyst masses are used to determine the influence of
film diffusion on the gas/liquid interface. At high catalyst mass, the rate becomes in-
dependent of the catalyst mass and is only limited by gas-liquid mass transfer, which
is independent of the catalyst amount. The stirring speed is typically varied to check
the influence of external (fluid/solid) mass transfer. As already emphasized the mass
transfer coefficient depends on the energy of dissipation, rather than the agitation
rate, thus special care should be taken to ensure that while varying the agitation rate
the energy of dissipation (which depends also on the liquid load) is also changing.
An experimental approach to verify the impact of internal diffusion is to per-
form experiments with catalysts of different particle sizes. It should be kept in
mind, however, that the flow field could be also changed while altering the pellet
size. In addition, if there are no changes in activity for different pellet sizes for a
catalyst with bimodal pore distribution, the influence of internal diffusion in the
small pores may still be important, while the test may just indicate that there are no
diffusional limitations in the larger set of pores. Another experimental criterion
could be applied that was presented by Koros and Nowak [43] and elaborated on by
Madon and Boudart [44]. This criterion is based on making rate measurements with
catalysts in which the concentration of the catalytically active material has been
changed. As the reaction rate in the kinetic regime is directly proportional to the
concentration of active material, an increase in activity will be proportional to
the amount of the active phase. The limitations of this method are associated with
the practical difficulties of preparing catalysts samples with different loading with-
out simultaneously altering some other important parameters, such as metal disper-
sion, diffusion characteristics or distribution of the active sites along the pore.

Exercise 3.1: The oxidation of methanol to formaldehyde is carried out over a Fe-Mo oxide catalyst.
The following operating conditions are used: Tbulk = bulk gas-phase temperature = 536 K; P = total
pressure = 0.12 MPa; CA = feed composition = 6.0 mol% methanol in air; porosity = 0.4, tortuosity
= 4; λe = effective thermal conductivity of catalyst = 0.17 (J m–1 s–1 K–1); heats of formations for
methanol, formaldehyde and water are correspondingly: 190, 112 and –239 kJ mol–1. The reaction
rate is first-order in methanol and zero order in oxygen. Activation energy is 80 kJ mol–1.

1. Calculate the maximum temperature difference between the surface and center of catalyst
pellet.
2. Determine whether temperature gradients in the catalyst pellets will have a significant influ-
ence on the reaction rate.
3. Is it possible for these reaction conditions that, depending on the catalyst particle size, the
effectiveness factor will exceed 1?

Neglect temperature effects on the heat of the reaction

320 3 Engineering reactions

Exercise 3.2 (From [32])

A first-order catalyzed decomposition of N2O in helium over spherical particles has been studied
in a laboratory-scale reactor. Use the data below to answer the following questions:

1. What is the impact of internal diffusion?

2. Do temperature differences exist over the gas film or within the particle?

Data: Catalyst dp = 2.4 mm, porosity = 0.2, tortuosity = 5, pore radius = 0.8 nm,
λe = effective thermal conductivity of catalyst = 0.45 (J m−1 s−1K−1), Gas film kf = 0.083 m
s−1, h = 46 (J m−1 s−1 K−1), Reaction heat = –163 kJ mol−1, cb = 20 mol m−3 (at 1 bar, 609 K),
rateobserved = 27 mol s−1 m−3.

3.6 Catalytic reactors

The choice of reactors is an important part of the engineering process and is dictated
by the purpose of the investigation (the study of mechanism and kinetics; generation
of data for process simulation and design; the investigation of catalyst performance
in a certain range of parameters), process economics, deactivation, etc.

3.6.1 Laboratory reactors

Laboratory reactors are needed for the collection of representative kinetic data for ac-
tivity and selectivity in the absence of mass transfer limitations and preferably deacti-
vation. A high precision of the data is needed because large deviations in the values
of the experimentally measured rates will have serious implications for quantitative
considerations. Reproducibility of rate measurements over a broad range of parame-
ters is also of importance. Another necessary feature is the possibility of obtaining
the maximum amount of kinetic information in the minimum time. Analysis of prod-
ucts as well as reactor lay-out should preferably be as easy as possible.
The essential feature for catalytic reactions is the readiness in reduction/activa-
tion of heterogeneous catalysts and a possibility to utilize them in the needed geo-
metrical form. Despite the strict definition of catalysis, which states that the
catalyst does not change during the catalytic reactions, some activity deterioration
takes place and therefore measurements of catalytic kinetics should always monitor
the catalyst activity.
For example, tubular reactors with fixed (packed) beds of catalyst and plug
flow (piston flow) (PFR) are often operated in a continuous mode (Fig. 3.46). In the
set-up presented in Fig. 3.46 the check valves are needed to stop back diffusion and
the tubing volume is minimized to avoid the “dead” volume.
 3.6 Catalytic reactors 321

Pressure Mass flow Check Back-pressure

reduction controller valve Reactor regulator

Fig. 3.46: Tubular reactor set-up.

This is a simple and inexpensive reactor type, which could be applied both as a
differential or integral reactor. In the first case the main requirements is the ideal
plug flow, low degree of conversion (typically below 10%) and high space veloci-
ties. Differential operation mode allows for direct measurement of the reaction rates
with simple equipment, while obvious disadvantages are the difficulties in achiev-
ing steady-state rates because of transients and deactivation, and the long time to
get just one experimental point, thus imposing serious limitations in studying mul-
tiroute processes. Conversely, the catalyst can be easily charged or discharged, and
there is even an option to use the reactor tube with the catalyst only once and dis-
charge the whole reactor. In addition, the catalyst deactivation could easily be seen
with time-on-stream. The tube could be uniformly heated in a tubular furnace al-
lowing isothermal conditions.
Dimensions of the reactor and the amount of catalyst depend on the scale (lab-
oratory/bench) and the tasks. A typical tubular minireactor could contain 0.1–1 g of
catalyst, while a bench scale reactor, operating in an integral mode might contain
50–200 g. Integral fixed-bed reactors of this size are suitable for measuring overall
conversion and catalyst lifetime under realistic conditions, which is important in
the scaling-up from a laboratory to a large-scale industrial reactor. Differential
fixed-bed reactors, not suitable for simulation purposes, are rather efficient for mea-
suring steady-state activity, selectivity and could be used for catalyst screening, es-
pecially when they are arranged in a parallel mode.
Quantitative treatment of plug flow reactors is cumbersome therefore several as-
sumptions are usually made. The fluid compositions is considered to be uniform
along the reactor cross section (i.e., there is no radial dispersion), which is valid only
at certain conditions dcat/dreactor < 1/30, while for larger catalyst particles some gas
can by-pass close to the reactor walls. When the reactants move along the reactor the
composition of the mixture changes and axial diffusion could become prominent. To
avoid possible impact of it in kinetic experiments the conversion level, as mentioned
above, should be preferentially below 10–15%. It should also be mentioned that, in
general, non-isothermicity of a tubular reactor should be considered (Fig. 3.47).
322 3 Engineering reactions

T0 Tc

T0
Fig. 3.47: Possible temperature profile in a single tube
x reactor.

Temkin and Kul´kova [45] developed a concept of single pellet string reactor
(Fig. 3.48), where real-size catalyst particles can be tested by placing catalyst
spheres and inert cylinders alternately in a tube of a diameter slightly larger than
the catalyst spheres.

Catalyst particle Dilution particle

Fixation wire
a) b)
Reactor wall

Fig. 3.48: Reactor concept of Temkin and Kul´kova [45]. (From [46]).

Absence of any contact between catalyst particles and a small space between
the inert spheres and the reactor walls ensure efficient heat transport. The orig-
inal design by Temkin and Kul´kova [45] was further optimized by Clariant [46]
to minimize irregularities in the flow patterns and allow accelerated testing
with the several so-called Temkin reactors in parallel. The current design com-
prises seven parallel tubular Temkin reactors, each filled with a series of reac-
tor modules with eight spherical cavities that are connected by cross-shaped
channels (Fig. 3.49).
Stirred tank reactors (STR) are very often applied for multiphase heterogeneous
catalytic reactions and can be operated batchwise (batch reactor, BR, Fig. 3.50A),
semi-batchwise (semi-batch reactor, SBR, Fig. 3.50B) or continuously (continuous
stirred tank reactor, CSTR, Fig. 3.50C).
Batch mode of operation brings several advantages as it allows for monitor-
ing the progress of the reaction over time and thus acquiring the whole kinetic
curve in one experiment. High precision and a wide range of parameters af-
forded by this operation mode made batch reactors very popular for kinetic stud-
ies especially in the field of fine and pharmaceutical chemicals. The catalyst
particles could be immersed in the slurry using a conventional autoclave
(Fig. 3.51A), which could be arranged in series (6, 16 or 32 reactors) allowing
 3.6 Catalytic reactors 323

Fig. 3.49: Accelerated testing with the Temkin–Kul´kova reactor [47].

Fig. 3.50: Stirred tank reactor (STR) operating (A) batchwise, (B) semi-batchwise and (C) continuously.

A A

A-A top
sectional view
(A) (B) (C)

Fig. 3.51: (A) autoclave, (B) Berty and (C) Carberry reactors.
324 3 Engineering reactions

high throughput screening. Alternatively, the catalyst can be introduced into

baskets (Fig. 3.51A,B).
For example, in a Berty reactor (Fig. 3.51B) the flowing gases are recirculated
through the catalyst bed using an impeller. The Carberry reactor (Fig. 3.51C) con-
tains a spinning basket with the solid catalyst, which is retained because of woven
wire mesh basket, allowing efficient gas-liquid circulation with low pressure drop.
While the advantage of such arrangements is the possibility to use heterogeneous
catalysts of different geometrical shapes, large catalyst particles required for this
operation mode often lead to conducting the reaction in the region of external diffu-
sion. Such reactors or other similar types, when the catalyst is attached as a fiber or
monolith to the impeller (Fig. 3.51), enable catalyst testing in the absence of the
thermal and concentration gradients, representing gradientless reactors.
These gradientless reactors belong to the group of stirred flow reactors, charac-
terized ideally by very small concentration and temperature gradients within the
catalyst region.

(A) (B)

Fig. 3.52: Reactors when the catalyst is attached to the impeller: (A) monolithic catalyst, (B) holder
for fiber catalysts.

Alternatively, recirculation reactors can be applied (Fig. 3.53), when there is exter-
nal mixing, ensuring constant pressure in the system and no changes in the gas
composition with time and in space as each single pass of reactants through the
bed results in very small conversion because of the very high circulation rate.
In the flow circulation reactor developed by Temkin, Lukyanova and
Kiperman (Fig. 3.53) the mixture of gases is fed through pipe (1) to the reactor (4)
and then to the outlet (2) for further analysis, for example by GC. The pump (3)
provides the necessary circulation of the mixture. The system is free from concen-
tration gradients along the catalyst bed, concentration gradients caused by axial
 3.6 Catalytic reactors 325

4 3

1 Fig. 3.53: Flow circulation reactor [48].

dispersion, and temperature gradients. The treatment of data is simplified as the

rate is extracted directly in the differential form. Efficient circulation pumps are
required and the inventory of reactants should be sufficient. This reactor can also
operate in a batch mode, allowing measurements of the whole kinetic curve in
one experiment with the disadvantage that the by-products are accumulated in
the loop. The relaxation time (time to reach steady-state) could be significant.
These disadvantages are, however, compensated by the easy mathematical treat-
ment of experimental data.
The concept of circulation flow reactor was further developed on an industrial
scale in the Buss reactor (Fig. 3.54), which can operate in batch, semi-continuous
and continuous mode. Large quantities of the reaction gas are introduced via a
mixer creating a well-dispersed mixture, which is rapidly circulated by a special

Venturi ejector

Hydrogen

Reactor Cooler

Recirculation
pump
Fig. 3.54: Buss loop reactor. (From [49]).
326 3 Engineering reactions

pump at high gas/liquid ratios throughout the volume of the loop, allowing the
maximum possible mass transfer rates. Independent optimization of the heat trans-
fer is arranged by a heat exchanger in the external loop. Separation of the catalyst
from the product is possible in a continuous operation, retaining the solid catalyst
within the loop.
During recent years, research activities have focused on carrying out reactions
on a small scale (Fig. 3.55).

Catalytic zone

Mixing zone
(A) (B) (C)

Fig. 3.55: Microchanneled reactor. (A) Two-phase reactor. (B) Flow pattern in a three-phase reactor.
(C) Wash-coating stability.

Because of higher heat and mass transfer, high ratio of surface area to volume and
short residence time, microreactors (or microstructured or microchannel reactors),
with the dimensions of channels in microreactors typically between ≈10 μm ± 2 mm
(Fig. 3.55A), provide efficient control over the exothermic reactions, reactions when
low inventory is needed for toxic reactants or reactions carried out under runaway
conditions. As a consequence, more aggressive conditions can be applied, leading
to higher yields than in conventional reactors. Another advantage of the microreac-
tors is a possibility of on-site production enabling the use of chemicals at the place
of production and avoiding the transportation and storage of dangerous materials.
Improvement of safety in microreactors is based on a small reaction volume and
thus diminished inventory of dangerous chemicals. Among the main challenges
related to the application of microchanneled reactors, flow distribution can be
extremely uneven not only for three-phase systems (Fig. 3.55B), but also for two-
phase (gas-solid and liquid-liquid) systems. In addition, the introduction of cata-
lytic particles in microreactors, ensuring their robustness, is not a trivial task.
 3.6 Catalytic reactors 327

Fig. 3.55C illustrates the instability of alumina wash-coating used as a support for a
ruthenium catalyst in the hydrogenation of sugars.
While an obvious use of microreactors is related to their application in generating
kinetic data, somewhat larger scale production is also feasible. Options in the scal-
ing-up of micro reactors could be external or internal numbering up. In the external
numbering up many devices are connected in a parallel fashion, which implies a par-
allel connection of functional elements that provides compact architecture at reason-
ably high throughput. External numbering up preserves all the transport properties
and hydrodynamics of a single device, requiring, however, sophisticated control sys-
tems, a large quantity of housing materials and thus high fabrication costs.
Parallelization of reactors is also related to maldistribution of the reactants, as al-
ready mentioned.
According to Bartolomew and Farrauto [50], laboratory and bench scale reac-
tors are selected to satisfy the intended application; avoid deactivation, heat and
mass transfer limitations; minimize temperature and concentration gradients while
maximizing the accuracy of measurements. In addition, minimization of the con-
struction and operation costs is also taken into account. Among the laboratory reac-
tors presented above for two-phase reactions differential fixed-bed reactors are
usually selected to study intrinsic kinetics, with the preference given to gradientless
reactors either with an external loop (flow circulation reactors) or to the stirred gas
reactor (Berty or Carberry type).
For three-phase reactions kinetic experiments are typically done in autoclaves,
although some researchers prefer CSTR. As batch operation mode is rarely used in
industry for large-scale production, the preferred reactor types are either slurry reac-
tors (bubble columns, CSTR and cascades of stirred tanks) or packed-bed reactors
(up-flow packed-bed or down flow trickle bed reactors). Packed bed reactors are illus-
trated for up-flow (Fig. 3.56B) and down flow (Fig. 3.56A) modes. Trickle bed reactors
can have multiple thermocouples (Fig. 3.57A) to measure temperature profiles and
can be arranged in parallel fashion for throughput operation (Fig. 3.57B).

Liquid

Gas

Liquid

(A) (B) Gas

Fig. 3.56: Concurrent three-phase reactors: (A) down flow (trickle bed), (B) up-flow.
328 3 Engineering reactions

(A) (B)

Fig. 3.57: Continuous reactors: (A) temperature control with multiple thermocouples (design by Dr.
A. Tokarev), (B) parallel trickle bed reactors (design Dr. K. Eränen).

3.6.2 Industrial reactors

The discussion below will be focus separately on two (gas-solid or liquid-solid) and
three-phase (gas-liquid-solid) reactors.

3.6.3 Two-phase reactors

Fixed-bed reactors with a packed catalyst bed are most commonly used in indus-
trial practice (Fig. 3.58) for the production of petrochemicals and various basic
chemicals. The catalyst is loaded in a vessel, through which the fluid is directed.
The catalyst particles are in the form of pellets (extrudates, tablets) of a size that
does not lead to too high-pressure drop, while increasing the diffusional length at
the same time. Fig. 3.58 shows adiabatic fixed-bed reactors, when there is no tem-
perature control in the bed and there is a monotonic temperature increase along
the bed. For single-route reactions (such as, for example, water gas shift) the con-
version can be calculated simply from the adiabatic temperature increase or de-
crease. When heat should be removed to prevent damaging the catalyst particles or
because equilibrium composition is unfavorable at a high temperature, multibed re-
actors (Fig. 3.58B and 3.58C) can be used with heat removal achieved either using
heat exchanges or cooling by cold reactants, respectively. Such procedures allow
for adapting the temperature profile to the requirements of an optimal reaction
pathway (Fig. 3.59).
Injection of hot or cold gas between the stages (or beds) as visualized in
Fig. 3.58B is a simple option. The disadvantages for constant tube diameter are
cross-sectional loading, which increases from stage to stage, and energetically
 3.6 Catalytic reactors 329

Feed Feed Feed

Interstage Interstage
gas feed heat exchange
Inert

Catalyst

(A) (B) (C) (D)

Fig. 3.58: (A) Single-bed adiabatic packed-bed reactor. (B) Adiabatic reactor with quenching. (C)
Multibed adiabatic fixed-bed reactor with interstage heat exchange (from [51].)
(D) Adiabatic reactor with radial flow.
Conversion

1
2

Temperature

Fig. 3.59: Conversion vs temperature profiles in a multistage reactor compared with an optimum
one reactor. 1, equilibrium; 2-optimum.

unfavorable mixing of hot and cold streams. As the composition is changing it can
have an either positive or negative effect on the reaction kinetics and thermody-
namics. Adiabatic multistage reactors with interstage heat transfer are typically
used in reactions giving a single product, but limited by equilibrium (sulfur triox-
ide, ammonia, methanol syntheses) where intermediate cooling is used to displace
the gas temperature in the direction of higher equilibrium conversion.
For strongly exothermic (partial oxidations and hydrogenations) or endother-
mic reactions (dehydrogenations) multitubular reactors with cooling or heating in
between the tubes are applied. This allows for better temperature control (Fig. 3.60),
although in this case the temperature profile is characterized by a maximum
330 3 Engineering reactions

Fig. 3.60: A multitubular reactor. (From [51].).

(hot spot), which is related to enhanced heat release at the reactor inlet and a de-
cline in heat release along the tube length caused by a diminished concentration
of reagents. An important issue with a multitubular reactor is the uniform flow
distribution among the tubes, which requires special measures when the catalyst
is loaded in the reactor. The loading procedures will be separately addressed in a
special section below in this Chapter.
The flow conditions in packed-bed reactors are very close to the plug flow (i.e.,
minor back-mixing), which implies a higher conversion rate for the most common
reaction kinetics and better selectivity towards the intermediate products for conse-
cutive and parallel-consecutive reactions.
The construction of the packed-bed reactors is simple, not requiring any
moving parts, and therefore this is generally a low-cost, low-maintenance reac-
tor. Among other advantages it is possible to have large variations in operating
conditions (including high pressure) and contact times, little catalyst attrition
and subsequent catalyst losses, high catalyst to reactant ratio and thus long resi-
dence time. In addition, principles of mathematical modeling of packed-bed re-
actors are well established. The most serious disadvantages are related to poor
heat transfer in large fixed-bed reactors and thus the inability to control temper-
ature properly. High-pressure drop is associated with limited productivity, while
non-uniform flow patterns and broad-residence time-distribution may lead to
poor selectivity. As mentioned above, the temperature control problems could be
overcome by introducing internal and external heat exchanges, interstage cool-
ing or heating and the application of diluents. The revamp of adiabatic axial
flow operation to radial flow and the use of special geometrically shaped cata-
lysts can improve pressure drop, simultaneously helping in removal pore diffu-
sional problems.
 3.6 Catalytic reactors 331

In some cases, even making a hole in a catalyst pellet helps to improve the pressure drop. I
was once involved in a case where the plant operator loaded the same amount of a tablet-
ized catalyst as was used in a similar plant by the catalyst supplier, without taking into
account different requirements for the maximum pressure allowed in these reactors (1.8 bar
in this case and above 3 bar in the other plant). As the reactor at this plant was not de-
signed to operate above 1.8 bar and the pressure drop was above 0.8 bar with the newly
installed catalyst batch, the required productivity was not reached. The plant operator had
to remove one-third of the catalyst load almost immediately and had to replace the catalyst
charge after a while with the same size of tablets but with an additional hole, which solved
the problem of the pressure drop.

Some examples of fixed-bed reactors will be discussed in Chapter 4. Regeneration

and replacement of the catalyst are difficult and require shut-downs.
Uniform flow distribution could be achieved if the pressure drop is significantly
high. It should be still ensured that the catalyst is packed uniformly in order to
avoid bypassing. Such uniform flow distribution is more difficult to achieve if the
pressure drop is low, i.e., in monolith reactors. The pressure drop in fixed-bed reac-
tors depends very strongly on the void fraction, as follows from the Ergun equation,
which is often applied for pressure drop calculations. For spherical particles the
pressure drop Δp (Pa) per reactor length is

ð1 − εÞ2 ð1 − εÞ 2
Δp=L = 150η vG + 1.75ρ 3 v (3:255)
ε3 d2p ε dp G

where vG is the interparticle velocity (m/s), ɛ is the void fraction, m3 of the

free space/m3 packing, ɳ is dynamic viscosity, N s m‒2, and ρ is density, kg/
m3, L is the length, m, dp the particle diameter, m. The first term in the Ergun
equation corresponds to the laminar flow, while the second reflects the turbu-
lent flow.

Recently, the author has received an e-mail, which had a following question: “We have bought
your book “Engineering Catalysis” and teaching catalysis reaction engineering from it. I would
like to know what is the maximum pressure drop that can be allowed while designing a fixed
bed reactor for large scale industrial processes? Please excuse me for asking this very trivial
question. I asked this because these issues are hardly mentioned in most of the undergraduate
textbooks on reaction engineering.” In my view, there is no general answer to this question as
it depends on a particular case. Obviously, the pressure drop influences the reaction rates and
the overall process economics, thus should be diminished still ensuring uniform flow distribu-
tion. In some cases, the maximum allowed pressure drop could be say 0.5 bar, while for other
cases it could be several bars. The reactor design including the pressure drop calculations
should be viewed as a part of the overall (upstream and downstream) technology.

An important parameter during the reactor selection is the adiabatic temperature rise,
which is defined as ΔTad = ΔHr Co/cp where ΔHr is the heat of the reaction (kJ/mol), Co
332 3 Engineering reactions

is the initial reactant concentration (mol m −3), and cp is the heat capacity (kJ/m3 K).
When ΔTad < 300 K a multibed reactor can be used with heat exchanges or quenching
in between the beds. If the adiabatic heat rise is ΔTad = 300 ÷ 700 K a multitubular
reactor is preferable. Sometimes two of these options are applied for the same reac-
tion. Fig. 3.61 shows adiabatic and isothermal options for methanol synthesis reactor,
offered by Uhde.

Adiabatic Multitubular
methanol methanol
reactor reactor
Steam

BFW
Product

Product
Feed
BFW
(A) (B) Feed

Fig. 3.61: (A) Adiabatic and (B) isothermal reactors for methanol synthesis. From [52].

The low-cost radial axial adiabatic multibed reactor with quenching is used in
plants when extra steam is required in the synthesis unit. In the isothermal reac-
tor the reaction heat is removed by partial evaporation of the boiler feed water
(BFW) generating 1 metric ton of medium pressure steam per 1.4 metric ton of
methanol. This isothermal reactor allows easier temperature control by regulating
steam pressure, conditions close to isothermal, high heat recovery and low by-
product formation.
An interesting option related to heat control is a reaction control with direct,
regenerative heat exchange in the reactors with periodic flow reversal developed by
Matros (Fig. 3.62). The catalyst packing acts simultaneously as the regenerative
heat exchanger. The catalyst is initially pre-heated to the reaction temperature and
then the cold reaction is put in contact with the hot catalyst packing. Thus, the re-
action gas is heated by the catalyst. The front moves into the packing, however, be-
fore it reaches the outlet of the bed, the flow is reversed by special valves and then
moves back again, heating the cooled part of the reactor. This operation ensures
that a periodic steady- state is established. The hot zone is in the center of the pack-
ing, while both ends act as regenerative heat exchanger. Although such periodic
flow operation was successfully tested in several medium-scale production plants
for the synthesis of basic inorganic and organic chemicals, the main application
 3.6 Catalytic reactors 333

Fig. 3.62: Autothermal reaction control with direct

(regenerative) heat exchange.

was found in catalytic off-gas purification, i.e., catalytic oxidation of volatile or-
ganic compounds (VOC) present in minute amounts in exhaust gases.
When the adiabatic temperature rise is above 700 K, fluidized bed reactors
(Fig. 3.63A) are usually used. When heat exchanges are located directly in the fluid-
ized bed of catalyst, it is possible to maintain the temperature at, for example,
300–400 °C, the order of magnitude is higher even in the adiabatic temperature

Product Catalyst in Catalyst in

Product Feed

Products

Heat
exchange

Feed Catalyst out Feed Catalyst out Product Reactants

(A) (B) (C)

Fig. 3.63: Reactors with moving catalysts. (A) Fluidized bed reactor; (B) a moving-bed reactor (left,
counter-current and right, concurrent flow of the feed and the catalyst); (C) entrained flow (riser)
reactor.
334 3 Engineering reactions

rise. Because of intensive heat transfer the temperature in almost all parts of the
fluidized bed reactor is the same, affording isothermal conditions.
When the catalyst undergoes strong deactivation the application of packed-bed
reactors is cumbersome, as continuous addition, removal and regeneration of the
catalyst is required. In the moving-bed reactor the catalyst moves downwards and
the gas flows upwards (Fig. 3.63B, left). The removed catalyst is transported in a
regenerator unit. For up-flow operation the catalyst, which is removed, is further
deactivated because it is in contact with the feed most rich in the reactant, while
less deactivated catalyst is in the reactor.
For the concurrent operation mode (Fig. 3.63B, right) applied for heavier feed-
stock, as for example in catalytic reforming, uniform catalyst activity along the
reactor is maintained. Typically, in such reactors the catalyst flows by gravity and
the reaction mixture can be withdrawn from the reactor and cooled or heated.
Heating is required in catalytic reforming, where one of the dominant reactions is
dehydrogenation.
When deactivation is even more prominent and the catalyst life is just few sec-
onds, an entrained flow or riser reactor is used when the catalyst powder is en-
trained by the catalyst in feed, thereafter being separated in a cyclone, regenerated
(not shown in Fig. 3.63C) and sent back to the riser. A detailed discussion on the
use of such reactors is given in a section devoted to fluidized catalytic cracking in
Chapter 4.
The obvious advantages of the fluidized bed reactors are associated with the
possibility to regenerate the catalyst continuously, high thermal efficiency and thus
efficient temperature control. In some cases, for example, vapor phase oxidations
of organic compounds, fluidized bed reactors are effective flame barriers that allow
operation at high concentration of organic compound, possibly even within the ex-
plosion limits. Moreover, small catalyst particles (50–100 microns) are needed for
fluidization, which leads to less prominent pore diffusional resistance and makes it
possible to run a process free from mass transfer limitations. The construction is,
however, much more complicated than for simple packed-bed reactors, requiring
extensive investments and higher maintenance costs. The flow conditions are com-
plex and could vary from an ideal plug flow to complete back-mixing. Various flow
regimes are illustrated in Fig. 3.64.
The minimum velocity required for fluidization of a fixed-bed regime is ob-
served when gas velocity is low. In the ebullated bed, there is a smooth bed ex-
pansion with a well-defined bed surface. Higher velocity gives rise to bubbling
fluidization when gas bubbles grow from the distributor, coalesce with other bub-
bles and eventually burst at the surface of the bed. In small diameter reactors, a
further size in fluidization velocity results in a slug flow regime with the bubble
diameter approaching the reactor diameter. In a turbulent regime, the top surface
of the bed cannot be distinguished. Transport of particles and their recycle back
to the reactor is a feature of fast fluidization.
 3.6 Catalytic reactors 335

Fig. 3.64: Various flow regimes of fluidized bed reactors [53]. Reproduced with permission.

In three-phase fluidized beds, the solid volume fraction is much higher

(10–50%) compared to slurry bubble columns (< 5%).
Scaling-up of fluidized bed reactors is challenging as the mass transfer char-
acteristics of the fluidized bed are dependent on reactor dimensions. Moreover,
some large gas bubbles are formed that could pass the reactor without being in
contact with the catalyst, leading to lower conversion and requiring additional
amounts of catalyst. The riser reactor (Fig. 3.63C), however, approaches the plug
flow regime.
An example of a riser reactor usage will be discussed in detail in Chapter 4 in
relation to an oil refining process – catalytic cracking. Some other examples of fluid-
ized bed reactors usage in synthesis of basic organic chemicals could be mentioned,
such as oxychlorination of ethylene (Section 4.8) and synthesis of acrylonitrile
(Section 4.7.5.2). In the synthesis of acrylonitrile a single plant capacity can reach
100,000 tons per year. Oxidative ammonolysis of propene to acrylonitrile:

2CH2 = CHCH3 + 3O2 + 2NH3 ! 2CH2 = CHCN + 6H2 O

occurs at 430–480 °C, pressures 1.3–1.5 bar, and contact times of a few seconds.
Serious disadvantages of the fluidized bed reactors are attrition, loss of catalyst,
small variations in the residence times, which are typically low. Attrition of already
small particles produces fines, which are expensive to separate from the product
gas. It is, however, possible in principle to combine the catalyst with an attrition-
resistant binder or coat the catalyst with some attrition-resistant material.
In order to achieve the fluidization regime for catalysts with the size d the ve-
locity at which fluidization starts Ub is defined as:
 
v
Ub = Reb (3:256)
dp
336 3 Engineering reactions

where v is viscosity and

pffiffiffiffiffidp is the diameter of particles, while the Reynolds number
Reb = Ar/(1400 + 5.22 Ar) is related to the Archimedes number Ar:

gdp ρp − ρs
Ar = (3:257)
v ρs

Here ρp is the density of particles, ρp is the density of the reaction mixture, and g =
9.8 ms-2. The velocity at which carry over starts, Uc′ can be evaluated from

 
v Ar
Ub = Reb , Reb = pffiffiffiffiffiffiffi (3:258)
dp ð18 + 0.61 ArÞ

The operational velocities are thus in between these values. For small catalyst par-
ticles of ~60 μm the ratio between Uc and Ub is ~50, while for some large particles
(100 μm) it can be around 10. Because for real industrial catalysts there is a certain
particle size distribution, the values of velocities at which the fluidized bed should
operate are also confirmed experimentally.
As mentioned above, the formation of large gas bubbles (laminar fluidized bed)
is undesirable because it decreases the interphase mass transfer. In order to achieve
a desired turbulent regime the gas-flow rate should be sufficiently high, with
Reynolds numbers of 100 ± 200. The catalyst particle must have sufficiently low
density. Otherwise a transition from the bubble fluidized bed regime to the riser
transport regime (carry over) will occur above Ar > 100 with an increase of the gas
velocity.

3.6.4 Three-phase catalytic reactors

Catalytic three-phase (gas-liquid-solid) reactors are extensively used in industry in

oil refining, synthesis of petrochemicals, and various basic, specialty and fine
chemicals; these processes require different types of reactors. In addition to reactors
with fixed catalyst beds using up-flow and downflow of liquids and well as co- and
counter-current flows of gases and liquid, slurry reactors (bubble columns and
stirred tanks reactors, reactors with external loops) and fluidized bed reactors are
also applied (Fig. 3.65).
In some cases (for example hydrogenation of benzene or methanol synthesis)
both gas-solid and gas-liquid-solid operations are possible. In general, in compar-
ison with gas-solid reactors, three-phase reactors operate at low temperatures,
which is translated into energy savings, also preventing potential damage to ther-
mally unstable catalysts and support. In addition to better temperature control
and more design options, low operation temperature usually also affords better
 3.6 Catalytic reactors 337

Gas Gas + liquid

Level of
Liquid with catalyst
catalyst in suspension
suspension

Gas Gas + liquid

Fig. 3.65: Typical three-phase reactors (adapted from [29]). Left, bubble column; middle, stirred
tank; right, fluidized bed.

selectivity and higher catalysts effectiveness. Conversely, the introduction of an-

other phase boundary obviously results in an increase of mass transfer, while low
temperature leads to a decrease in the chemical reaction rate. Moreover, batch
slurry reactors are designed to operate in a limited pressure and (more impor-
tantly) a limited temperature range, which imposes some difficulties for catalyst
reduction and in situ activation where they are needed. In addition, in industry
the power input for impellers is limited and the same holds for the minimum cata-
lyst particle size, which is defined by available separation systems, and thus the
mass transfer cannot be neglected. In general, with slurry bed reactors the effi-
cient separation of catalysts, which are either suspended in the slurry by stirring
or bubbling gas up through the liquid, is difficult.

3.6.4.1 Fixed-bed multiphase reactors

Packed-bed reactors can operate in different modes with the downflow and up-
flow of the liquid and gas, and the gas can flow either co- or counter-currently.
There are certain advantages for the liquid in operating either up or down flow.
For example, the downflow option in fixed-bed multiphase reactors results in
lower-energy input, lower pressure drop and an absence of fluidization, while up-
flow operation leads to better liquid distribution, mixing, heat dissipation, surface
wetting, easier heat exchange, higher gas/ liquid mass transfer and effectiveness,
as well as slow aging.
In the downflow operation mode the liquid moves in a laminar flow down-
wards, wetting the catalyst in a trickling flow, thus the name “trickle bed reactors”
(Fig. 3.66). In these trickle bed reactors both the gas and liquid approach a plug
flow condition.
The flow map of different flow regimes observed in trickle bed reactors is illus-
trated in Fig. 3.67.
338 3 Engineering reactions

Liquid Liquid
Gas

Inlet distribution Gas

tray
Ceramic balls

Gas

Gas + liquid Liquid

Solid catalyst

Catalyst support Gas

Outlet collector
Gas +
Liquid
(A) liquid (B)

Fig. 3.66: Trickle bed reactors: (A) downward cocurrent, (B) counter-current gas flow. (Reproduced
with permission from [54]).

103
c f
e Bubbling

102 f T b
𝜌g 𝜌L 1/2 d
󰜆=(
λ𝜓GL/GG(-)

)
𝜌air 𝜌water d
1/3 Pulse flow
𝜎 󰜇 𝜌
𝛹 =𝜎W 󰜇L ( 𝜌W)2
L W L
101 Trickle
f
e b
c
Spray regime
100
10–2 10–1 100 101
GG/λε (kg/m2 s)

Fig. 3.67: Flow regime map in fixed-bed reactors with co-current gas–liquid flow [55].

The trickle flow regime is observed at low gas and liquid flow rates, with such
features as incomplete wetting, low pressure drop, low gas–liquid throughput, less
catalyst attrition and suitability for foaming liquids. Heat and mass transfer rates
 3.6 Catalytic reactors 339

are poor compared to other flow regimes in trickle bed reactors. Pulse flow regime
is observed at moderate gas and liquid flow rates. The liquid phase occupies the
entire flow cross section. Such regime has better utilization of the catalyst bed in
terms of wetting, as well as higher heat and mass transfer rates, which makes the
pulse flow regime potentially attractive for industrial application. In the bubble
flow regime at a low gas flow rate and moderate/high liquid flow rates, the volume
of the bed is occupied by the liquid with the gas phase flowing downward in the
form of bubbles. In this regime, the pressure drip is higher. Moreover, higher liquid
holdup leads to back-mixing, which depending on the reaction might not be suit-
able. When the liquid phase is a limiting component or the reaction is highly exo-
thermic, complete wetting of the bed and high heat and mass transfer rates can be
beneficial. The spray flow regime is observed at low liquid and high gas flow rates.
In this regime, droplets of the liquid are formed, while the gas phase is a continu-
ous phase. Subsequently this regime with a high gas–liquid mass transfer has a low
liquid holdup and low foaming ability.
Trickle bed reactors usually operate in two regimes. At low gas and liquid flows
[even as low as 0.1 kg m–2 s–1 in bench scale reactors and 0.6 kg m–2 s–1 in commer-
cial scale reactors] a trickle flow dominates, while at low gas and high liquid flows
the liquid-phase is continuous and gas bubbles flow through in a pulsed flow
regime.
In such a trickling flow, maldistribution and incomplete wetting (Fig. 3.68) and
losses caused by attrition counterbalance the advantages, while up-flow operation
requires a larger energy input. In order to avoid stagnant regions and excessive cat-
alyst loadings, more complicated geometrical shapes are preferable for trickle bed
operations.

Stagnant Liquid
cavity
Catalyst

Gas Gas

Catalyst Dry
zone
Liquid Gas

Fig. 3.68: Trickle bed reactors with maldistribution. (Reproduced with permission from [56]).
340 3 Engineering reactions

For up-flow reactors, in addition to the bubble flow (low gas and high liquid
flows) and spray flow (higher gas and low liquid flows), a slug flow with uneven
distribution of bubbles (high gas and low liquid flows) can also be developed,
which gives in the latter case a plug flow for the gas-phase, but partially back-
mixed liquid.
In summary, trickle bed reactors offer a simple structure, low investment costs
and closeness to plug flow. It should be mentioned that the ideal plug flow is the
one that is usually desired.
Catalytic monolith reactors, which were typically applied for two-phase systems
(gas abatement, automotive exhaust cleaning) with high gas velocity (as discussed
in Chapter 2), were proposed as an alternative to fixed-bed three-phase reactors
(Fig. 3.69).

Fig. 3.69: Monolithic reactor for three-phase system. (From [23].).

At high gas-flow rates the bubbles fill the cross section, affording very good mass
transfer. Special care should be taken to avoid annular flow at very high gas-flow
rates when the gas flow becomes continuous. Although the monolithic reactor can
allow very low pressure drop and negligible mass transfer as the wash-coating
thickness is small, the use of such reactors in industrial practice is prevented by
uneven liquid distribution and uneven metal profiling along the channels plus high
catalyst costs and limited experience.

3.6.4.2 Moving-bed multiphase reactors

In stirred tanks, cascades of stirred tanks and bubble columns, the back-mixing is
significant, broadening the residence time-distribution and making it more difficult
 3.6 Catalytic reactors 341

to achieve high yields. There are, however, several advantages of the reactors with
the moving catalysts beds, justifying their frequent application in industry.
Bubble columns (Fig. 3.70) often operating in a semibatch mode, i.e., continu-
ously for gas and as batch for liquid, provide good mixing and heat transfer char-
acteristics. Gas is distributed from the bottom of the reactor as bubbles, whose
size depends on physical–chemical properties, sparger design and superficial gas
velocity. The distribution of gas bubbles in radial direction is nonuniform, assist-
ing in internal circulation within the bubble column driven by buoyancy (an up-
ward force exerted by a fluid). Overall the back-mixing for the liquid and the solid
catalyst particles is more intensive, thus the bubble columns operate as isother-
mal reactors.

Gas outlet Gas outlet Gas outlet

G-L
separator
G-L
separator
Cooling Cooling Cooling
tube tube Riser tube

Down- Riser Downer

comer

Gas
Gas distributor
distri- Gas
butor Gas inlet distri-
Gas inlet butor Gas inlet
(A) (B) (C)

Fig. 3.70: Moving-bed multiphase reactors: (A) Bubble column, (B) Internal loop airlift reactor,
(C) External loop airlift reactor. (Reprinted with permission from [57]).

Mixing can be enhanced by introducing gas (air) lift, ejectors and circulation
pumps. The airlift reactor consists of risers and downcomers that provide the flowing
channel for global circulation of the liquid-solid slurry phase (Fig. 3.70B,C).
Bubble columns exhibit very complex hydrodynamics due to spatiotemporal
variations in interactions among gas, liquid and solid phases (Fig. 3.71).
Two major flow regimes exist in these reactors: the homogeneous and the het-
erogeneous ones. A homogeneous bubble flow regime is developed when bubbles
342 3 Engineering reactions

0.15

Slug flow regime Churn-turbulent regime

Gas velocity (m s–1)

0.10

Transition range
0.05

Homogeneous bubble flow regime

0
0.025 0.05 0.1 0.2 0.5 1
Column diameter (m)

Fig. 3.71: Flow map of a bubble column. (From [58], with the original figure from [59]. Reprinted
with permission).

are of the same size (ca. 0.5 to 2–3 mm), which is attained at low gas velocities. The
rate of bubble coalescence and breakup is lower in the homogeneous bubble flow
regime. With an increase of gas velocity for columns with large diameters there is a
certain bubble size distribution (heterogeneous flow regime). This flow regime is
also called churn–turbulent flow regime. Larger gas flow gives more inhomogeneity
in the bubble size and shape. In smaller columns at higher gas flow rates slugs are
formed because of wall restriction. The bubbles (mainly large bubbles with smaller
ones in between the liquid slugs) occupy the entire reactor cross section. Smaller
column diameters result in more sustainable slug and better control of back-mixing
and heat and mass transfer rates. Taking into account smaller diameters and thus
lower throughput, it is not surprising that these miniaturized bubble column reac-
tors can have advantages compared to other reactor options in production of fine
and specialty chemicals.
Mechanically agitated stirred tanks (with complete back-mixing profiles) can be
used in batch and semi-batch (continuous for the gas) modes, as well as CSTR with
a possibility of catalyst recycling. In CSTR (Fig. 3.72A) the mixing of the incoming
reactant is instantaneous, thus the concentration in the reactor is the same as the
concentration at the outlet. Such high degree of back-mixing is typically unfavor-
able for kinetics, resulting in a lower conversion of the reactants. Therefore, cas-
cades of CSTR are used in industrial practice (Fig. 3.72B).
Various options for temperature control in slurry reactors will be discussed in
Section 4.1.4. As already mentioned, the separation of fine catalyst particles can be
 3.6 Catalytic reactors 343

Feed Motor Feed

Cooling
jacket

Baffle

Agitator

(A) Mixed product (B) Product

Fig. 3.72: Stirred tank reactor. (A) CSTR [60]. (B) Cascade of stirred tanks.

challenging, therefore somewhat larger particles are used in industrial practice with
non-negligible diffusion resistance.

Many years ago I had a discussion with a plant manager and suggested, based on experimental
data in a laboratory reactor, that in order to avoid mass transfer limitations they should use
catalyst particles below 50 μm for hydrogenation of an aromatic compound. Unfortunately, the
filters available at the plant site could only handle catalyst particles that were several times
larger, leading unavoidably to the presence of internal diffusion.

The fluid flow in stirred tanks is highly complex and requires computational
fluid dynamics to get an exact solution of the partial differential equations asso-
ciated with the Navier-Stokes momentum, energy and continuity equations.
Tangential, axial and radial flows exist in stirred tanks depending on the impeller
used.
Application of baffles (three or four flat plates attached to the walls), occupy-
ing up to 10% of the tank diameter, can stop the tangential flow. Such baffles can-
not be recommended when the flow is laminar. Axial and radial flow impellers
could be used to generate flow parallel or perpendicular to the axis (shaft), re-
spectively. Pitched bladed turbines create a mixed flow. Axial flow impellers are
recommended for turbulent mixing. Although variations in the size of catalyst
particles mean that it is not possible to get suspensions that are completely uni-
form, the impellers are usually placed at one-quarter of the liquid level to achieve
acceptable mixing. In order to increase the gas-liquid mass transfer, the gas bub-
bles should be dispersed by the impeller. It should be also remembered that
because of the gas dispersed in the liquid, the volume of the liquid is increased
344 3 Engineering reactions

(gas hold-up). Heat transfer is typically unaffected by mixing and could be im-
proved by increasing heat transfer area (for example, installing coils in addition
to a heat transfer jacket).
Industrial reactors cannot operate at the same level of mixing efficiency as the
small scale reactors, thus simple geometrical similarities are insufficient in scaling-
up, and location of the feed points should be carefully considered. Some engineer-
ing consultants even advise for process development to rather scale down and use
laboratory reactors based on available large-scale reactors to mimic the industrial
conditions [61].
Finally, fluidized bed reactors should be highlighted as an option for conduct-
ing three-phase reactions. Such reactors typically operate in a concurrent mode
with up-flow of both gas and liquid. Conditions of high gas and low liquid flow (ag-
gregative fluidization) resemble two-phase (gas-solid) fluidized beds, while at low
gas flows the flow pattern (bubble flow) is obviously similar to fluidized beds with
only liquid- and solid-phase. A slug domain with uneven distribution of particles is
formed in between these two extremes.
Compared to bubble columns, larger particles are usually used in fluidized
beds because of higher liquid flow velocities. The flow pattern is more of back-
mixing type than plug flow.

3.6.4.3 Comparison between various reactor options

A general comparison between catalytic three-phase reactors was done by Trambouze
et al. [62] and is presented in Tab. 3.8.

Tab. 3.8: Comparison between three-phase reactors.

Application Suspension catalysts Three-phase fluidized Fixed bed

criteria beds

Catalysts related
Activity Highly variable, but Highly variable: intra- and Plug flow favorable
possible in many cases to extra-granular mass
avoid the diffusion transfers may significantly
limitations found in a fixed reduce the activity,
bed especially in a fixed bed
Back-mixing unfavorable
Selectivity Selectivity generally As for activity, transfers Plug flow often favorable
unaffected by transfers may decrease selectivity
Back-mixing often
unfavorable
 3.6 Catalytic reactors 345

Tab. 3.8 (continued )

Application Suspension catalysts Three-phase fluidized Fixed bed

criteria beds

Stability Catalyst replacement Possibility of continuous This feature is essential for

between each batch catalyst renewal: the fixed-bed operation: a plug
operation helps to catalyst must flow may sometimes be
overcome problems of nevertheless have good favorable because of the
rapid poisoning in certain attrition resistance establishment of a poison
cases adsorption front
Cost Consumption usually Necessarily low catalyst
depends on the impurities consumption
contained in the feed and
acting as poisons
Technologies characteristics
Heat Fairly easy to achieve heat Possibility of heat Generally adiabatic
exchange exchange exchange in the reactor operation
itself
Design Catalyst separation Very simple technology for
difficulties sometimes difficult: a downward concurrent
possible problems in adiabatic bed
pumps and exchangers
because of the risks of
deposit or erosion
Scaling-up No difficulty: generally System still poorly known, Large reactors can be built
limited to batch systems should be scaled up in if liquid distribution is
and relatively small sizes steps carefully arranged

(From [62].)

3.6.4.3.1 Fischer–Tropsch synthesis

It is instructive to compare different reactors mentioned in Tab . 3.8 for a particular
case of synthesis gas (mixture of hydrogen and CO with a stoichiometric ratio of 2:1)
transformation through the Fischer – Tropsch (FT ) synthesis (this section ) and to
methanol (section 3.6.3.4.2). In the former case the synthesis gas is converted to a
widerangeofhydrocarbonsinahighlyexothermicreaction:

nCO + 2nH2 ! ð − CH2 Þn + nH2 O ΔH = − 165 kJ=mole ðat 227 o C Þ

The molar hydrogen to CO consumption ratiois 2:1.The inlet ratio of H2 to CO in

industrial reactors is different from the stoichiometry thus, for example, in the
fixed-bed FT reactor the H2/CO ratio is changing along the reactor length.
WGS (Water-gas shift) reaction

CO + H2 O ! CO2 + H2 ðΔH = − 39.8 kJ =mole at 227 o CÞ

346 3 Engineering reactions

is significant at low-temperature FT (LTFT) synthesis (ca. 230 °C) giving an overall

stoichiometry

2CO + H2 ! ð − CH2 Þn + H2 ðΔH = − 204 kJ=mole at 227 o CÞ.

Subsequently, the hydrogen to CO consumption ratio is only 0.5. For high-

temperature FT (HTFT) synthesis the WGS reaction is less prominent.
Conditions of the FT process can be adjusted to the feedstock. When synthesis
gas is produced by coal gasification at high (> 1,000 °C) temperature, the generated
gas is rich in CO. To adjust the ratio between hydrogen and CO it is desirable to have
WGS activity, which is provided by iron catalysts. When the feed, such as syngas
from methane, is rich in hydrogen, no WGS activity is needed, and cobalt catalysts
active at low temperature can be used with a danger of promoting an undesired, in
this case, WGS reaction.
Utilization of iron catalysts at higher temperatures depending on the amount of
alkaline promotes can cause formation of carbon deposits and subsequent deactiva-
tion through direct coke formation H2 + CO → C + H2O and the Boudouard reaction
2CO → C + CO2 . Another potentially significant reaction is methanation CO+3H2 →
CH4 + H2O, which is thermodynamically more favorable than formation of higher
hydrocarbons. Consequently, applied catalysts should be selective toward forma-
tion of less thermodynamically preferred products.
An overview of FT technologies and reactor types used is given in Fig. 3.73.

Moving-bed Stationary-bed
reactors reactors
Low-temperature FTS
200–250 °C
• Three phase system: Multitubular
gas-liquid-solid fixed bed
Slurry bubble 6,000 b/d
• α = 0.85–0.95
column,
• Products:
with wax
waz, diesel, naphtha
24,000 b/d
• Catalysts:
supported cobalt or
precipitated iron

High-temperature FTS
320–350 °C
• Three phase system:
gas-solid
• α = 0.70–0.75
• Products:
gasoline; chemicals Fixed
• Catalysts: Circulating fluidized bed
fluidized bed 20,000 b/d
fused iron,K promoted
7,000 b/d

Fig. 3.73: Overview of Fischer–Tropsch technologies and reactor types used [63]. Capacities are in
barrels per day (b/d).
 3.6 Catalytic reactors 347

More expensive cobalt-based catalysts that operate at lower temperature

(200–250 °C) favor long-chain paraffins, are more robust and have low WGS activity
contrary to a cheaper alternative, iron. Iron needs higher temperature (220–350 °C
or higher for fluidized beds), possesses high selectivity to olefins and oxygenates, is
a WGS catalyst and readily deactivates.
FT process was originally developed for the production of synthetic fuels from
coal. The first FT plants began operation in Germany in 1938. There were nine
low-temperature cobalt-based plants with a total annual product capacity of
660,000 tons eventually closed down after the Second World War. In fact, coal-to-
liquid (CTL) had limited applications for many decades after WWII outside of the
Sasol company in South Africa because of a number of political and technical rea-
sons. A general overview of the process flow with coal as a feedstock is given in
Fig. 3.74.

Coal
Electricity
Prep Tail
Product gas
Synthesis gas FT Power
production process recovery generation
N2 O2 CO
H2 Hydrogen
Air sep. WGS Wax recovery
Air H2
plant
Liquid
Gas Wax
Fuels
treatment hydrocracking
Liquid
CO2 H2S fuels Transportation
fuels
Mid–distillate Diesel

Fig. 3.74: A general overview of the process flow [64].

The first Sasol plant in Sasolburg, South Africa, started operation in 1953 and had
annual production of million tons of FT products using coal as a feedstock and op-
erating five tubular fixed-bed (ARGE) reactors for wax production and three circu-
lating fluid bed reactors. A slurry reactor of the same production capacity replaced
five ARGE reactors in 1993. In 2004 natural gas reforming was introduced instead of
coal gasification at Sasol by transforming the plant technology from CTL to GTL pro-
ducing waxes and chemicals.
348 3 Engineering reactions

Further expansions by Sasol in Secunda, South Africa, were done in 1980s, uti-
lizing high-temperature Synthol reactors with improved heat exchange thereby
boosting threefold the capacity compared to the first-generation circulating fluid-
ized bed (CFB) reactors. The main focus of the production site of Sasol in Secunda
is on motor gasoline and diesel, as well as some chemicals. High-pressure distillate
hydrogenation section was also added to the tar refinery of Sasol III processing gas-
ification derived coal pyrolysis liquids from Sasol II and Sasol III. Sasol Advanced
Synthol reactors eventually replaced 16 second-generation CFB reactors with eight
fixed fluid bed (FFB) reactors decreasing the operation costs at the same capacity.
In the early 1990s, Mossgas started up a natural gas based 1 million ton per year
FT plant in South Africa using a high-temperature process with an iron catalyst for
making motor gasoline, distillates, kerosene, alcohols and LPG, while Shell put on-
stream 500,000 tons/year natural gas-based FT plant using the Shell Middle
Distillate Synthesis (SMDS) process for automotive fuels, specialty chemicals and
waxes. The strategy of the LTFT synthesis is to produce heavier products with a co-
balt catalyst, when formation of long-chain waxes is favored. The heavy alkanes are
converted through mild hydrocracking to the desired carbon number range with
subsequent product distillation.
Following this trend of renewed interest in using the FT process for the synthe-
sis of gasoline and diesel from primarily natural gas, the Sasol Oryx GTL plant oper-
ating in Qatar using a cobalt catalyst at low temperature was commissioned in 2007
producing 34,000 bpd of diesel fuel mainly and naphtha as by-product. The FT syn-
crude is similar to SMDS process and is processed in a similar way. Shell Pearl GTL
plant put onstream in 2011 in the same location in Qatar with a capacity of
140,000 bpd of petroleum liquids relies on the same LTFT technology to produce
distillate and base oils.
Worth mentioning is the world’s first coal to diesel conversion production line
of Shenhua Group with the diesel production capacity of 860,000 tons in 2017.
As mentioned earlier, high exothermicity is a typical feature of the FT process.
The choice of the catalyst (iron or cobalt) and process conditions (pressure, temper-
ature, hydrogen to CO ratio) influences the product molecular weight distribution.
The products in FT reactions are mainly n-paraffins, although terminal olefins and
alcohols could also be formed.
A well-defined range of products, for example middle distillates, cannot be di-
rectly synthesized; thus, the strategy of newer and more economic plants is to gen-
erate high-molecular-weight linear waxes for further hydroprocessing.
The main reaction in FT synthesis follows a polymerization-like mechanism
when a monomer CHx species (x = 1–2) is added stepwise to a growing aliphatic
chain (Fig. 3.75A).
Chain termination by desorption of unsaturated surface species and hydrogena-
tion with subsequent desorption of saturated species relative to chain propagation
determines the process selectivity. The weight fraction of a product with a carbon
 3.6 Catalytic reactors 349

(A) CO + H2 (B)
1.0
initiation 0.9 CH4

Termination 0.8
CH3 CH4 Wax
1–α 0.7
(C35–C120)

Weight fraction
Propagation α 0.6

0.5 Gasoline
C2H5 C2H6 (C5–C11)
1–α 0.4
Propagation α Diesel
0.3 C2 (C9–C25)
0.2
C3
Propagation α 0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
CnH2n+n CnH2n+2
1–α Probability of chain growth (α)

(C)
100
C3+C4 fuel gas
80 C3+C4 LPG
Percentage

60
Naphtha
40
Kerosene
20
Wax
Gas-oil
0
0.75 0.80 0.85 0.90 0.95
Probability of chain growth, α

Fig. 3.75: Illustration of chain growth, weight fraction of hydrocarbons and percentage of different
hydrocarbon products as a function of chain growth.

number n is defined through the Anderson–Schulz–Flory distribution Wn = n αn–1

(1–α)2n (Fig. 3.75B), where parameter α is the chain growth probability being the
ratio of the chain propagation to the sum of chain propagation and chain termina-
tion. This parameter is supposed to be independent on the carbon number.
As already mentioned, the product distribution depends on operating condi-
tions (temperature, pressure and CO-to-hydrogen ratio) as well as the catalyst type.
Cobalt gives more paraffins, while iron results in the product higher in olefins and
oxygenates (Fig. 3.76).
Typically, the degree of polymerization α ranges from 0.7 to 0.95. Even with
α = 0.95, a range of different products is generated with predominant formation of
high-molecular-weight linear waxes (Fig. 3.75C).
Low temperature (200–240 °C) and mid-pressure (2–3 MPa) are selected for FT
process along with active catalysts based on iron and cobalt to get high selectivity
to heavier products. Alternative utilization of nickel as a catalyst results in mainly
350 3 Engineering reactions

Olefins Alkanes Oxygenates Aromatics

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Co-LTFT Fe-LTFT Fe-HTFT

Fig. 3.76: Product distribution depending on the catalyst. Co-LTFT: cobalt low-temperature FT;
Fe-LTFT: iron low-temperature FT; Fe-HTFT: iron high-temperature FT.

methanation. In FT synthesis, essentially no aromatic compounds are formed ex-

cept for high-temperature processes (Fig. 3.76). The product is also free from sulfur
and nitrogen compounds.
Cobalt-based catalysts are generally preferred for natural gas-based syngas giv-
ing FT stoichiometric H2/CO ratio or close to it (Fig. 3.77).

Coal Naphtha

Vacuum
residue Fuel oil
Natural Natural
Asphalt gas gas and steam

Iron (Fe)

Cobalt (Co)

0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

H2/CO

Fig. 3.77: Ratios of H2/CO ratio for different feedstocks and catalysts [65].

Metallic cobalt which is considered to be the active phase in FT catalysts has low
WGS activity. Earlier, Co catalysts for fixed-bed FT were prepared by co-precipitation,
 3.6 Catalytic reactors 351

followed by shaping with extrusion, calcination and reduction. The novel generation
applied in slurry-phase reactors is mainly synthesized by impregnation of oxides with
aqueous or organic solutions of cobalt nitrates and other additives. Calcination of Co
nitrates results in formation of Co oxide, which is reduced in a hydrogen containing
gas.
During catalysis, Co metal crystallites are largely covered by active and inac-
tive carbonaceous species. Co catalysts are more expensive than iron-based ones,
at the same time possessing 10–20 times higher activity (calculated per weight for
promoted Co vs promoted Fe catalysts), high selectivity to long-chain paraffins
(C5+) and low selectivity to olefins and oxygenates, being also resistant to deacti-
vation. The metal loading is typically 35 wt% with the metal dispersion of approx-
imately 8–10%. A range of metal promoters (0.1–0.3 wt% Pt, Re or Ru) is used
to increase reducibility and dispersion of Co, improve stability against carbon
buildup and increase C5+ selectivity. Oxide promoters (i.e. 1–3% BaO or La2O3 or
other additives) are used to stabilize cobalt crystallites and the support and pro-
mote hydrocarbon chain growth. As a support δ-alumina (ca. 150 m2/g) stabilized
with lantana is used. The support should be chemically and physically stable dur-
ing catalyst preparation, activation, regeneration and reaction. The support can
be stabilized with some other oxides. Besides alumina, supports such as silica
and titania can be applied.
In comparison with cobalt, Fe-based catalysts are less expensive, display low
selectivity to long-chain paraffins, generate more olefins and oxygenates, exhibit
WGS activity and more prone to deactivation by coking.
Fe-based catalysts were used commercially in SASOL plants. They are generally
preferred for coal-based plants with lower hydrogen to CO ratio. The active phase is
Fe carbides (FexC, x < 2.5) covered by various carbonaceous species.
Fused iron oxide catalysts, for which alkali promotes are added during fusion at
high temperature to improve activity and selectivity (Fig. 3.78), were applied in the
HTFT synthesis using the fluidized bed (Sasol Advanced Synthol) reactors. Fusion is
followed by casting into ingots and cooling, milling and activation in hydrogen.

Iron oxide

1500 °C Molten magnetite Cooled rapidly

K2O (Fe3O4) Fused iron
Air Crushed in a ball mill

MgO or AI2O3

Fig. 3.78: Procedure for synthesis of fused iron oxide catalysts.

352 3 Engineering reactions

Precipitated iron oxide with addition of promoters (additives and modifiers)

during precipitation is more expensive to prepare and has less structural strength.
Such catalysts can be used in LTFT synthesis organized in trickle bed (ARGE) and
slurry reactors. The catalysts are prone to deactivation and gradual loss of activity
with a possibility for regeneration. Copper is added to the iron catalyst in order to
increase the rate of iron reduction and the catalyst activity. Selectivity toward lon-
ger hydrocarbons (waxes) is improved through addition of alkali (K2O). Silica is
mainly used as a support for precipitated iron catalyst affording the highest activity
and selectivity to waxes.
The catalyst is prepared by coprecipitation of iron and copper nitrates, followed
by washing and impregnation with the potassium and structural promoters.
Shaping of catalysts, done either by extrusion for fixed-bed or by spray drying for
slurry-phase reactors, is followed by calcination. Prior to application, the oxidized
iron catalysts are activated in hydrogen or synthesis gas.
Catalyst deactivation in FT synthesis occurs through poisoning of the active
metal by sulfur or nitrogen compounds. This can be prevented by desulfurization of
the syngas feed with formation of H2S, which is captured by ZnO installed upstream
an FT reactor. Prevention of fouling due to blockage of pores with hard waxes can be
achieved by operating at lower value of parameter α, control of hydrogen to CO ratio,
lowering temperature and in situ treatment of the deactivated catalyst in hydrogen at
temperature 10–20 °C higher than the reaction temperature. Hydrothermal sintering
of Fe happens at high pressure of steam. Keeping the latter below 0.5–0.6 MPa or
operating below 50–60% of CO conversion prevents sintering as well as formation of
iron oxides. Application of multiple reactors with intermediate removal of H2O and
stabilization of the supports with Ba, Zr or La oxides serve as a preventive measure
against deactivation. The same approach is used to prevent formation of cobalt ox-
ides. Generation of inactive cobalt carbides is minimized by keeping hydrogen to CO
ratio above 2.1 in all reactor parts.
Loss of catalytic material due to abrasion and erosion in the case of fluidized or
slurry reactors can be prevented by adequate preparation methods, including
sol–gel granulation and application of binders.
Several different reactor types (fixed bed, fluidized bed or slurry phase) are
used in the FT processes.
Details of operational conditions and the product composition for most impor-
tant FT facilities are given in Tab. 3.9.
Because of the exothermicity of FT reactions, heat removal is a major concern
in reactor design. In addition, selectivity to unwanted product – methane – is in-
creased with temperature, thus efficient temperature control is needed to achieve
desired selectivity. Several types of reactor systems presented in Fig. 3.79 (tubular
fixed-bed, circulated and fixed fluidized beds, slurry bubble columns) were and are
used in industrial practice.
 3.6 Catalytic reactors 353

Tab. 3.9: The most important FT plants [66].

Company Plant, Feed Technology/ Catalyst Conditions Products and

location, reactor capacity
date

Sasol Sasol I, Natural gas LT/slurry Precipitated – °C , bbl/d

Sasolburg, phase Fe/K Paraffin, waxes,
South Africa, distillate + oxygenates and
 multitubular fuel gas
fixed bed

Shell Bintulu, Natural gas LT/ Co/SiO  °C SMDS, ,

Malaysia, multitubular  bar bbl/d
 fixed bed LPG (–%),
naphtha
(–%),
distillate
(–%) and
oils (–%)

Sasol Oryx GTL Natural gas LT/slurry Co/Pt/AlO  °C ,  bbl/d
Ras Laffan phase  bar LPG, naphtha
Industrial distillate and distillate
City, (diesel blend)
Qatar, 

Shell Pearl GTL Natural gas LT/ Co/SiO  °C SMDS, ,
Qatar,  multitubular  bar bbl/d
fixed bed LPG (–%),
naphtha
(–%),
distillate
(–%) and
oils (–%)

Chevron Escravos Natural gas LT/ Co/Pt/AlO  °C , bbl/d

GTL, Nigeria, Similar to  bar LPG, naphtha
 Oryx and diesel
blend

Sasol Sasol  and Coal HT/fixed Fused Fe/K  °C , bbl/d.
 (Synfuels), fluidized bed  bar Fuel gas, oils,
South Africa, alpha-olefins,
 ammonia,
gasoline, jet
fuel, diesel
354 3 Engineering reactions

Tab. 3.9 (continued )

Company Plant, Feed Technology/ Catalyst Conditions Products and

location, reactor capacity
date

PetroSA Mossgas, Natural gas HT/ Fused Fe/K – °C , bbl/d.
Mossel Bay, circulating  bar LPG, gasoline,
South Africa fluidized bed diesel, fuel oil,
kerosene,
aromatics,
alcohols

A tubular reactor was the first to be used commercially. Although this is a sim-
ple design, easy to scale-up, construction was rather expensive because of the
large number of tubes needed for the industrial reactor. In the fixed-bed reactor,
the catalyst tubes (ca. 2,000) with approximately 5 cm in diameter and 12 m in
length are surrounded by boiling water, allowing efficient heat removal by
evaporation.
As mentioned above, replacement of catalyst is an issue, as iron catalysts
should be replaced periodically because of deactivation. Contrary to the lifespan
of iron, cobalt-based catalysts are more robust, with a lifetime of several years
and could be regenerated. The catalyst size is above 1 mm to avoid extensive pres-
sure drop, thus the effectiveness factor is certainly below unity and mass transfer
limitations are present. Possible temperature gradients in the tubes can lead to
sintering and deactivation.
Alternatives to tubular reactors are CFB (Fig. 3.79) reactors (Sasol Synthol reac-
tor) and fixed fluidized bed reactors (Sasol Advanced Synthol reactor). CFB reactors
provide better heat removal and temperature control and experience less pressure
drop problems than a tubular reactor. Moreover, catalyst removal and addition are
possible on-line. Fluidized bed reactors operate with iron catalysts at high tempera-
ture (330 °C–360 °C) giving products in the gas phase with the chain growth proba-
bility of 0.7–0.75. Such reactors can be used for gasoline production, as well as
olefins and oxygenates, which can be further processed to other types of fuels (e.g.,
oligomerization of olefins to diesel range hydrocarbons on HZSM-5 zeolites) and to
chemical products.
The major disadvantage of fluidized beds for FT applications is that a low mo-
lecular weight product (gasoline) is obtained, while the concentration of diesel
range products and waxes cannot be high because products must be volatile at the
reaction conditions. If non-volatile hydrocarbons accumulate on the catalyst, the
particle fluidization behavior is worsened as the particles stick to each other.
Scaling-up of such reactors is more difficult in comparison to tubular reactors.
 3.6 Catalytic reactors 355

Fixed fluidized bed Tubular fixed reactor

Product gases
Steam heater

Cyclones Steam
Gas inlet
collector
Steam outlet
Fluidized bed
Feed water
inlet
Steam Boiler
feed water

Tube
Inner shell
Gas bundle
distributor

Total feed Gas outlet

Wax outlet

Slurry phase reactor Circulating fluidized bed

Products

Slurry bed Product gases

Cyclone

Steam Boiler feed Catalyst Heat exchanger

water
Wax
Stand pipe
Gas
distributor
Gas and catalyst
Slide valve
Synthesis gas Gas feed
in

Fig. 3.79: Reactors for FT synthesis.

Slurry phase bubble columns (Fig. 3.79) are considered to be the best choice for
newer FT reactors with more active cobalt catalyst, which is suspended in a slurry.
The synthesis gas is bubbled through this slurry that contains hydrocarbon waxes,
liquid at reaction conditions, and catalyst particles of 50–80 mm in size. Such reac-
tor could weigh ~2,200 tons and be up to 60 m high with an outer diameter of 10 m.
Simpler heat removal and easier scaling-up are advantages of the slurry phase bub-
ble columns. For LTFT synthesis the chain growth probability is much higher than
for HTFT producing also liquid waxes. In the liquid slurry phase reactor, the cata-
lyst is suspended in the liquid product wax.
356 3 Engineering reactions

Such reactors provide good heat transfer and temperature control, low pressure
drop and are suited for the synthesis of higher boiling products. This could be an
advantage, as it gives more overall flexibility if there is a downstream unit for hy-
drocracking of the long-chain waxes, thus generating high-quality middle distil-
lates (diesel and jet fuel). The product of LTFT synthesis contains naphtha, which is
a feedstock for chemicals. The design of slurry phase bubble columns is rather sim-
ple, allowing easy addition and removal of catalysts. The catalyst, however, should
be highly mechanically stable and be resistant against attrition. The gaseous prod-
ucts are removed from the top, while there is a need to separate waxes from the
catalyst, which is send back to the reactor.
Table 3.10 presents a comparison between different three-phase reactors illus-
trating clearly that product composition and operation parameters are reactor
dependent.

Tab. 3.10: Comparison between different three-phase reactors for

Fischer–Tropsch synthesis (From [67]).

Application criteria Slurry bubble column Riser Fixed bed

Conditions
Inlet/Outlet T (K) / / /
Pressure (bar)   
H/CO ratio . . .
Conversion (%)   –
Products (wt%)
CH . . .
C . . .
C . . .
C . . .
C–C (gasoline) . . .
C–C (diesel) . . .
C+waxes . . .
Oxygenates . . .

As already mentioned the strategy of the LTFT synthesis illustrated in Fig. 3.80 is to
produce heavier products with a cobalt catalyst, when formation of long-chain
waxes is favored (α-value of 0.9 and higher). The heavy alkanes are converted
through mild hydrocracking to the desired carbon number range with subsequent
product distillation.
HTFT syncrude conducted in CFB or FFB reactors with Co catalyst has obviously
a higher naphtha yield, while application of LTFT using slurry (Co catalyst) or tubu-
lar (Fe catalyst) results in higher boiling point hydrocarbons. LTFT refineries are
less complex and typically have hydroprocessing (wax hydrocracking) steps and
 3.6 Catalytic reactors 357

Fuel gas (including LPG)

Syngas 470–490 K
20–30bar
570–620 K
30–50 bar Naphtha
Steam

Kerosene

BFW Diesel

Hydrogen

FT reactor Flash Hydrocracking Distillation

Fig. 3.80: Simplified flow scheme of the Shell Middle Distillate synthesis plant.

fractionation to produce naphtha and middle distillates . At the same time , LTFT re-
fineries are making fuel blending stocks rather than final fuels.On the contrary ,diesel
production can be achieved in HTFT process as the syncrude contains aro - matics ,
naphthenesgivingadieseldensityclosertotherequiredspecification.
Hydrocracking and hydroisomerization of the heavy product fraction are differ -
ent compared to similar processing in oil refineries (Chapter 4). More specifically ,
the feed after an LTFT consists of only paraffins without sulfur compounds and aro-
matics . This allows using a highly active hydrogenation metal (e.g., platinum ) on
an acidic support (silica –alumina or proton forms of zeolites ) with a danger of sec -
ondarycracking.

3.6.4.3.2 Methanol synthesis

Methanol is synthesized from CO and hydrogen according to the following revers
- ible exothermal reaction:

CO + 2H2 $ CH3 OH, ΔH = − 90.8 kJ =mol

Because the reaction is exothermal , equilibrium constant decreases with the in-
crease in temperature . Elevation of pressure results in shifting equilibrium toward
theproductside.
Another reaction leading to methanol is related to hydrogenation of carbon
dioxide

CO2 + 3H2 $ CH3 OH + H2 O, ΔH = − 49.6 kJ =mol

These two reactions are coupled by the WGS reaction

CO + H2 O $ CO2 + H2 , ΔH = − 41 kJ=mol
358 3 Engineering reactions

By-products in this process are higher alcohols and hydrocarbons. Formation of

dimethylether is also possible due to methanol dehydration. Application of active
catalysts based on copper (CuZn/Al2O3) allowed to decrease operation pressure of
25–35 MPa used in the classical gas phase processes with ZnO–Cr2O3 catalysts to
approximately 5–10 MPa. Selectivity toward the desired product in low-pressure
plants is above 99%. It should be kept in mind that the modern catalysts allow to
obtain such high selectivity toward the product, which is not the most thermody-
namically preferred, namely, methane is a more thermodynamically favored prod-
uct than methanol. Conversion of CO and CO2 to methanol is limited by chemical
equilibrium, thus a temperature rise being in principle beneficial from the view-
point of kinetics negatively influences thermodynamic equilibrium. In addition,
highly active catalysts are sensitive to temperature rise, which leads to catalyst
deactivation through irreversible sintering. Although initial activity declines sub-
stantially during operation as illustrated for different commercial catalysts in
Fig. 3.81, a careful catalyst design allows to extend the lifetime to 4–6 years and
even more.
Catalyst activity

Days on stream

Fig. 3.81: Dependence of catalytic activity in methanol synthesis with time-on-stream for different
catalysts.

A typical measure for counterbalancing deactivation in various catalytic processes

is to increase temperature restoring activity but often compromising selectivity. In
the case of methanol synthesis such approach cannot be easily applied as the tem-
perature should not exceed approximately 270 °C. Because copper is marginally ac-
tive below 230 °C, the temperature window for the process is rather low.
Increase in pressure is an alternative way of compensating for activity loss due to
sintering. At the same time, too high pressures of CO and CO2 (favoring conversion
from the thermodynamic viewpoint) increase equipment costs in the synthesis loop
and for syngas compressor. The synthesis loop is thus required as pressure in the
 3.6 Catalytic reactors 359

modern plants of 5–10 MPa gives only moderate conversion levels (15–30% in adia-
batic reactors). Unreacted gas is recycled back acting as a syngas quench cooler. The
ratio between the recycle gas and the fresh feed ranges from 3:1 to 7:1, which along
with the purge allows to prevent buildup of impurities (methane and argon) in the
loop.
Raw methanol containing water and impurities is condensed and sent to the
distillation unit, whose design depends on the desired product purity. Typically,
from one to three distillation columns are used with the first one (so-called topping
column) acting as a stabilizer for removal of dissolved gases (CO, CO2, H2, N2 and
CH4) and some of the light by-products (aldehydes, ketones and dimethyl ether). In
the downstream columns, raw methanol, containing besides water also minor
amounts of higher alcohols, is fractionated (Fig. 3.82). The heat input is optimized
in the three-column system.
An important theoretical and practical issue is related to a question which reac-
tant is leading to methanol. A long controversy surrounded this topic, and either
CO or CO2 or both were proposed as the true reactants. Methanol can be produced
from both H2-CO and H2-CO2 mixtures, while a mixture containing H2, CO and CO2
gives much higher yields of methanol. Isotopic labeling studies suggest that the
source of carbon in methanol is CO2, while CO is mainly converted to CO2 via a WGS
reaction. CO2 is also influencing the properties of the catalyst keeping an intermedi-
ate oxidation state of copper (Cuo/Cu+) and preventing reduction of ZnO. High con-
centrations of CO2 are, however, inhibiting methanol synthesis, in which rate drops
slightly up to 12 vol% of CO2 and thereafter more steeply.
A typical gas composition (can be different depending on the synthesis gas genera-
tion procedure) could be thus: 67.5% H2, 21.5% CO, 8% CO2 and 3% CH4 for high capac-
ity plants and 69% H2, 18% CO, 10% CO2 and 3% CH4 for lower and medium capacity.
Modern commercial catalysts for methanol synthesis are supplied as tablets and
contain above 55 wt% CuO, 20–25% ZnO with 8–10% Al2O3 and also catalyst pro-
moters, as well as catalyst binders (e.g., graphite). For such structure-insensitive reac-
tion as methanol synthesis the activity is dependent only on the total exposed copper
area and is not affected by the structure of the crystallites. This means that large load-
ing of copper (reaching 64% in some commercial formulations) and small cluster sizes
are in fact need for efficient catalysts. High metal dispersion as such is not enough for
successful industrial operation, as the catalyst should be stabilized against sintering. It
was mentioned earlier that thermal sintering is a key mechanism for deactivation with
temperature approaching 315 °C depending on the reactor type. Moreover, sulfur and,
in some cases, iron and nickel carbonyls introduced into the loop with fresh syngas
contribute to catalyst deactivation. ZnO used in the commercial formulations for many
decades is a textural and chemical promoter being introduced as small crystallites
(2–10 nm). It helps to stabilize copper against sintering, facilitating formation of small
copper clusters and also scavenging sulfur. Alumina needed in the catalyst to stabilize
both ZnO and copper oxide might in principle lead to formation of dimethylether;
 360
Synthesis gas

3 Engineering reactions
Compressor Separator
DME

Methanol 99%
Activated charcoal
Reactor
absorber

Off gas
Heat exchanger

Cooler Flash drum

iSE
www.inclusive-science- Heavy ends
engineering.com Low boiler column
Purge High boiler column

Production of methanol from synthesis gas

Recycle gas compressor

Fig. 3.82: Methanol production flow scheme with purification section [68].
 3.6 Catalytic reactors 361

however, presence of ZnO oxide neutralizes acidic sites of alumina. Other promoters
(such as MgO) were also introduced in commercial formulations.
Utilization of the commercial catalysts in the form of cylindrical pellets of 5–12
mm implies that diffusional limitations can be significant.
Several reactor designs and synthesis flowsheet arrangements for methanol pro-
duction can be utilized. Reactor choice depends on a plant size and the syngas genera-
tion method. In reactor selection, conversion temperature profiles should be optimized
being close to the equilibrium one and affording lower peak temperature. Moreover, in
addition to optimized temperature profile, proper mixing and reactant distribution
allow higher selectivity, thus diminishing amounts of by-products with substantial sav-
ings in product purification, as well as lower deactivation and longer catalyst life.
Mainly multibed (three to four) adiabatic fixed-bed reactors are applied in low-
pressure methanol processes with heat removal either by quenching with the cold
feed (quenching) or using heat exchangers. The temperature profiles are far from the
maximum rate curve as illustrated in Fig. 3.83 for a four-bed adiabatic reactor with
heat exchangers. Despite this apparent inefficiency, multibed reactors represent an
attractive low-cost reactor concept when there is no need for steam generation.

10
Methanol concentration (mol%)

Maximum rate curve

8
Methanol
equilibrium
6

0
180 200 220 240 260 280 300 320
Temperature (C°)

Fig. 3.83: Temperature profile for multibed adiabatic methanol synthesis reactor with heat
exchangers. Adapted from [69].

Not only cylindrical but also spherical adiabatic reactors (Fig. 3.84A) are applied
when the catalyst is located between two perforated spherical shells. Such reactor
type allows a decrease in the vessel wall thickness at a given pressure and thus af-
fords lower reactor costs. The flow in such reactors is organized from the outside of
the catalyst layer to the center of the spherical core. Pressure drop is minimized as
a relatively thin catalyst layer is used.
In a tube-cooled converter (Fig. 3.84B) the feed enters the reactor at the bot-
tom and flowing upward through the tubes with minimum thickness is preheated
362 3 Engineering reactions

(A) (B)

Feed gas
Tube-cooled
2
converter
(TCC)

BFW Steam
Syngas
from Circulator
compressor

BFW Steam 1 3
Purge

4
Heat
recovery
Condenser
Crude
Product gas methanol
to distillation

(C) (D)

Axial
1
steam-raising
converter Catalyst loading
Gas inlet
(A-SRC)

Steam
drum
Syngas
from Cooling Catalyst
2 compressor tube
Circulator
Central pipe

3
Purge
Inert balls
4
Steam outlet
Condenser BFW inlet
Interchanger Crude
methanol Gas outlet and
to distillation catalyst unloading

(E)
Manway Manway

Steam
outlet

Outer Catalyst
tube
BFW
inlet
Manway
Gas
Support
out grid
Diaphragm
Flexible hose Manway
One touch coupling Feed gas inlet

Fig. 3.84: Special reactor types used for methanol synthesis: (a) spherical adiabatic reactors,
(b) tube-cooled converter, (c) converter with axial steam rising [69], (d) Toyo’s MRF-Z reactor
adapted from [70] and (e) Mitsubishi superconverter, adapted from [71] .
 3.6 Catalytic reactors 363

by the product gas flowing downward through the catalyst bed. This way of ar-
ranging the heat exchange gives a temperature profile (Fig. 3.85) much closer to
the maximum rate curve than the case of a multibed reactor with heat exchangers.
The catalyst amount in such a tube-cooled reactor with the axial flow is limited by
pressure drop considerations. For large capacity plants, several reactors might be
needed.

10
Methanol concentration (mol%)

Maximum rate curve

8
Methanol
equilibrium
6

0
180 200 220 240 260 280 300 320
Temperature (C°)

Fig. 3.85: Methanol concentration profile in tube-cooled converter. Adapted from [69].

Near isothermal operation (Fig. 3.86) is provided in a tubular boiling water reac-
tor with axial flow (Fig. 3.84C), where the catalyst is located on the tube side.
Temperature is controlled by pressure of water, which is circulated on the shell side
generating steam at the maximum possible pressure without overheating the cata-
lyst. The temperature profile is close to the maximum rate curve and allows some-
what low temperatures than in tube cooler converters, still requiring a significant

10
Methanol concentration (mol%)

Maximum rate curve

8
Methanol
equilibrium
6

0
180 200 220 240 260 280 300 320
Temperature (C°)

Fig. 3.86: Concentration profile in a steam-generating multitubular reactor. Adapted from [69].
364 3 Engineering reactions

10

Methanol concentration (mol%)

Maximum rate curve
8
Methanol
equilibrium
6

0
180 200 220 240 260 280 300 320
Temperature (C°)

Fig. 3.87: Concentration profile in Toyo’s MRF-Z reactor. Adapted from [69].

recycle ratio. High investment costs for this reactor concept limit the maximum
plant size to approximately 1,500 tpd and require several reactors in series for larger
capacity.
A specific feature of Toyo’s MRF-Z reactor (Fig. 3.84D) is multistage indirect cool-
ing and a radial flow facilitating the capacity increase of methanol plants. This reactor
type generates steam of approximately 3 MPa, has a good approach to equilibrium
(Fig. 3.87), a small number of tubes and a low pressure drop (0.05–0.075 MPa) while it
can be in the range of 0.3–1 MPa for fixed-bed adiabatic reactors.
Mitsubishi reactor (Fig. 3.84E) for methanol synthesis can be viewed as an inte-
gration of interchange and steam rising. The design is rather complex consisting of
a large number of tubes, a manifold, two tube sheets. It generates approximately
4 MPa of steam, closely following the maximum rate line (Fig. 3.88) and thus allow-
ing high conversion per pass and a lower recycling rate.
Methanol concentration (mol%)

10
Maximum rate
curve
8
Methanol
6 equilibrium

0
180 200 220 240 260 280 300 320
Temperature (C°)

Fig. 3.88: Concentration profile in Mitsubishi superconverter. Adapted from [69].

 3.6 Catalytic reactors 365

The reactors presented above correspond to gas-phase methanol synthesis. An

alternative liquid-phase process for methanol production was developed by Air
Products and Chemicals.
The technology relies on a bubble slurry reactor (Fig. 3.89), in which an inert
hydrocarbon acts a reaction medium and a heat sink. As the feed gas bubbles
through the catalyst slurry forming MeOH, the mineral oil transfers the reaction
heat to an internal tubular boiler where the heat is removed by generating steam.
The reactor operates at isothermal conditions being able to handle CO-rich (in ex-
cess of 50%) syngas with wide compositional variations. Such operation mode al-
lows to reach much higher concentration of methanol (ca. 15%) than in the gas-
phase process increasing conversion from 15% to ca. 35%. This technology was
proven at the demonstration plant level [73] but has not yet been commercialized.
Few key scale-up-related issues were identified based on the experience with the
demonstration plant. A particular important in the context of this textbook was re-
lated to catalysis. Iron from residual construction debris caused severe poisoning of
the catalyst, requiring replacement of the full catalyst charge.

Unreacted syngas +
methanol product
(vapor)
Disengagement
zone
Vapor bubble
Vapor bubble

Catalyst
H2 CH3OH
Steam
CO

Liquid

Boiler feed
Water
Catalyst powder
slurries in oil

Syngas
Feed

Fig. 3.89: Bubble slurry reactor for liquid-phase methanol synthesis: (a) schematics [72] and
(b) column at the demonstration plant [73].

3.6.5 Reactor modeling

Nowadays many industrial companies realize that simulation and optimization of a

catalytic process require sound knowledge on reaction kinetics, transport phenom-
ena and hydrodynamics. The principles objectives of reactor modeling and design
include determination of the most technically and economically relevant reactor
366 3 Engineering reactions

type; its sizing by establishing the reactor and catalyst volume for grass root reac-
tors or estimation of throughput for available reactors; and determination of the op-
timum conditions and operation policy.

3.6.5.1 Reactor modeling and process development strategy

The chemical development, process development and process engineering phases
proceed both consecutively and in parallel. A typical interrelated workflow for the
development of chemical manufacturing processes is presented in Fig. 3.90.

Conceptual chemical
phase Chemical development
Conceptual process phase Commissioning and
development phase Process development start-up phase
phase Validation and
Process engineering optimization phase
phase

Fig. 3.90: Workflow for the development of a chemical manufacturing process. (From [74].).

The methodology of reactor modeling, on which the discussion below partly is

based, was presented by Dutta and Gualy [75].
Initially the reactions and separations of a process are studied thoroughly in
laboratory and bench scale. For catalytic reactions two types of process develop-
ment units are usually applied. In the first, the catalyst is used in the form of pow-
ders to generate reliable kinetic data; in the second, step conditions correspond to
those of a commercial plant in terms of catalyst particle size and hydrodynamics.
The key operating variables (space velocity or throughput, temperature, pres-
sure and composition) are usually carefully investigated at this stage. In particular,
a special attention should be paid to monitoring the temperature profiles, as this
parameter is the most dominant in reactor design and operation. It is important to
cover various combinations of variables, explore the broad range of them and get
information on each component reaction.
One of the most critical tasks is to establish reaction kinetics in a reliable fash-
ion. Otherwise the kinetic models are applicable only to a narrow range of condi-
tions. It is advisable in mechanistic studies to change one parameter at a time, at
least initially, and then cover several possible and diverse combinations of varia-
bles. An extremely important aspect is catalytic kinetics under non-steady-state
conditions, with transient experiments providing additional information necessary
in dynamic simulations. Non-stationary kinetics is also needed when a process in
organized in a non-stationary model (for example, reaction-regeneration cycles).
Experimental data generated in process development units should cover not
only the domain of pure kinetics, but also the region of significant mass transfer
 3.6 Catalytic reactors 367

resistance. For three-phase systems, after experimentation with the catalysts in

powder form (micron level), a catalyst basket containing pellets of mm size could
be placed inside an autoclave to determine pore diffusion.
An important question in the generation of catalytic data is related to the purity
of the feedstock and use of model compounds in test units. Preliminary laboratory
studies, aimed at initial screening of potential catalysts, evaluation of intrinsic ki-
netics and mass transfer, could probably be performed with a model feedstock. In
scaling-up batch operations the real feedstock can in principle be used at an early
stage, but this masks, however, intrinsic kinetics caused by the presence of various
impurities, which influence for instance catalyst deactivation.
The safety aspects should be studied carefully. This is important because labo-
ratory-scale continuous reactors are made of small diameter tubes that provide effi-
cient heat transfer to the environment. For large commercial reactors, adiabatic
conditions could be more prominent, leading to heat build-up.
The operating conditions are typically selected to be outside of explosion lim-
its, however, for fluidized bed reactors no hot spots are generated because of vig-
orous stirring, thus operation within the explosion limits is possible, provided
special care is taken about the way the reactants are introduced and removed
from the reactor.
In many cases an extensive pilot plant campaign is undertaken after laboratory
tests. In fact nowadays, extrapolation of laboratory results to industrial conditions
can be rigorously performed through adequate mathematical modeling.

A number of years ago at one of the conferences on chemical reactors I was approached by a
chemical engineer who worked at a petrochemical company in South America, and who very
enthusiastically acknowledged the efforts made by my colleague Prof. T. Salmi. It turned out
that the rigorous mathematical model for hydrogenation of aromatics, which included detailed
kinetics and transport phenomena, was directly used for building a commercial plant with an-
nual capacity of 300,000 tons.

Thus, from the modeling perspective, pilot plant experiments should be carried out
only when it is necessary to improve and validate models and when the passage from
the lab to production scale is too complex because of sophisticated reactor types.
However, in practice, piloting is still widely applied. The reasons for this are as follows.
The management, responsible for making decisions about building a large-scale plant,
needs to have more confidence in the process. The same holds for the potential cus-
tomers if licensing is an issue. Representative product samples might be needed for
marketing purposes and various application tests. Moreover, pilot units allow investi-
gating such parameters as the feed type, impurities, build-up of by-products, catalyst
deactivation, corrosion, etc. Stability tests are usually performed at a pilot scale with
the real feedstock. It is even possible to utilize side reactors (for example, in parallel
with industrial ones) for piloting as practiced by some large chemical manufacturers.
368 3 Engineering reactions

The complexity of the tasks, which a chemical engineer faces, justifies the ne-
cessity and the need for reliable kinetic and transport models, which are based on
physicochemical understanding of catalytic processes. The reactor models cover
several levels: an active site (mechanism and kinetics), a catalyst grain (mass and
heat transfer and thus knowledge of catalyst effectiveness factor, interphase mass
transfer, phase holdups) and the reactor itself (hydrodynamics, bed voidage profile,
axial and radial dispersion, bubble properties, wall heat transfer, etc.).
The degree of complexity of the reactor model is still a matter of debate. A very
complex model with a very large number of adjustable parameters might be found
impractical by some industrial companies.

Through my own industrial experience and collaboration with various companies I have seen
utilization of reactor models of different complexity levels ranging from very simple to extremely
sophisticated. On average, over simplified or over complicated models should be avoided, as
the former are applicable only to a narrow range of parameters, while the latter could be diffi-
cult to handle during parameter estimation.

A few reactors-related issues should be specifically highlighted. Changing geometri-

cal dimensions for fixed-bed reactors and trying to keep the same contact time
means that the ratio between the length of the catalyst bed and the linear velocity
should be kept constant. At the same time, it will certainly change the Reynolds
and Peclet numbers, which determine hydrodynamic conditions.
The importance of the regeneration model in systems such as fluidized bed
reactors should not be underestimated. In particular, the circulation lines should
be properly sized, which is important for cases when the catalyst itself acts as an
oxygen carrier and demands high solids circulation rates. Maleic anhydride syn-
thesis by n-butane oxidation is an industrial example of such a riser-regenerator
reactor arrangement. In this case, hydrocarbons and oxidants are not in contact,
which provides inherent safety to this technology. For alternative partial oxida-
tion technologies that use multi-tubular reactors, there are risks of explosions
and thus special care should be taken regarding the mixing zone and injection
procedures.
Design of the entrance zone and fluid (gas or liquid) distributors is, in general,
important as conversion is significant close to the rector inlet. In CFBs the entrance
and exit zone are also critical in influencing the overall reactor hydrodynamics.
On a more general level the reaction model should be clarified first before prop-
erly addressing reactor hydrodynamics. Although the actual hydrodynamics should
be modeled adequately, the impact of an inadequate reaction model is more serious
than an inadequate hydrodynamic model.
It is advisable during the model building to check the model consistency
against simple kinetic and reactor models where analytic solutions are available, as
the more complex model should be able to account for known simple systems.
 3.6 Catalytic reactors 369

Nowadays process simulators are commercially available with dynamic reactor

models, which are able to predict not only normal but also transient operations
(start-up and shut-downs). These simulators contain various reactor models and
make it possible to determine the best reactor configuration. Fig. 3.91 illustrates
such a multiple reactor model package.

Multitubular fixed
Adiabatic bed with varying
quench catalyst concentration Multitubular
reactor fixed bed
HC Air
gases

Mixed
HT fluid
gases
in HT fluid
? in
? ?
Outlet
gas
HC Fixed bed Outlet
gases design ? gas

Air ?
?
Inlet
Steam gas
Outlet
gas Radial flow
reactor
Adiabatic quench reactor
with split air flow BFW
Outlet
gas
? Double-wall heat
Inlet gas
exchanger reactor

Gas/solid Best
? Control, operational flexibilities, detailed safety,
multiple- technical
BFB/CFB catalyst mechanical checks and compatibility with
reactor model designs
development feasible? upstream and downstream sections
package

? Multistage BFB Products

Bubbling fluid Solids
bed (BFB) in
Multistage BFB
with split air flow
Air and temerature
programming
Solids out Gas out
?
HC gases
? Solids in
Fluidized ? Air
Product bed
BFB/BFB combo design
Reactor (BFB) ?
with catalyst as ? Solids out Eco-
oxygen carrier ? HC gases nomics
Stripping Flue gas
Products
gas
Regenerator
(BFB) Flue gas
Stripping The
Regenerator
gas final
Air (BFB)
design
Stripping
gas Reactor
HC vapor (CFB)
Air Circulating fluid
Stripping gas bed (CFB)
CFB/BFB combo with HC gases
catalyst as oxygen carrier

Fig. 3.91: Decision making with a multiple reactor model package. (From [75].).
370 3 Engineering reactions

3.6.5.2 Reactor models: mass and heat balances

The reactor type is defined by the physical configuration, the flow mode and
hydrodynamics.
The first model for discussion is catalytic packed-bed reactors. They can be
classified in two main categories: pseudo-homogeneous and heterogeneous mod-
els. In pseudo-homogeneous models the diffusional limitations (concentration
and temperature gradients) inside the catalyst are neglected and the fluid phase
(gas or liquid) and the solid catalyst are considered as a continuous phase. Mass
balance equations are thus only required for the fluid (bulk) phase. Even when
internal diffusion is prominent the pseudo-homogeneous model can be used
through use of a correction term, the catalyst effectiveness factor, extensively dis-
cussed in Section 3.5.
Heterogeneous models account for diffusion effects inside catalyst particles by
introducing separate balance equations for the fluid inside the catalyst pores and in
the bulk phase surrounding the catalyst particles.
Pseudo-homogeneous and heterogeneous models can be either one- or two-
dimensional. For one-dimensional models the radial concentration and temperature
gradients in the reactor tube are neglected and a chemical reaction is assumed to
proceed at the same rate in various radial positions in the reactor cross section.
When such radial concentration and temperature gradients are present, radial dis-
persion coefficients are utilized. Severe temperature gradients can appear in indus-
trial reactors, as the heat transfer area diminishes with an increase in the size and
the heat losses to the environment are not as efficient as in laboratory reactors.
In an idealized case, packed-bed reactors are described by a plug flow model,
which assumes a total absence of axial mixing within the reactor. Plug flow reactors
are typically long tubes with a high ratio of the length to the diameter.
In order to derive the mass balance for such reactors, we must first consider the
general mass balance for an infinitesimal volume element ΔV (Fig. 3.92):

V
Flow Flow
∙
∙
nA nA
n0A Fig. 3.92: General mass balance.

IN + GENERATION = OUT + ACCUMULATION

leading to the following equation in terms of moles:

dnA
n_ 0A + ηA rA ΔV = n_ A + (3:259)
dt

where ṅ0A and ṅA are the mole fluxes, ɳA is the catalyst effectiveness factor, which
takes into account mass transfer.
 3.6 Catalytic reactors 371

Equation (3.259) can be rewritten:

dnA
ηA rA ΔV = Δn_ A + (3:260)
dt

where nAin − nAout = Δn_ A · Replacing the amount in moles through concentration and
volume, considering an infinitely small volume nA = cAΔV, ΔV→ 0 one arrives at:
_ A dcA
dn
ηA rA = + (3:261)
dV dt

When the volumetric flow V(m _ Defining the residence time

_ 3/s) is constant n_ A =cA V.
in the reactor (τ, s) as τ = V=V_ 0 the dynamic mass balance equation is obtained:

dcA dcA
=− + ηA rA (3:262)
dt dτ

At steady-state the concentration is independent on time dcA/dt = 0, thus:

dcA
= ηA rA (3:263)
dτ

and the design equation is given by:

cðA ðα
dcA dα
τ= = c0A (3:264)
ηA rA − ηA rA
c0A 0

where α is conversion. For packed-bed reactors the balance is exactly the same as
for homogeneous plug flow reactors, except that the rate in the balance equation is
replaced by rA ρB, where ρB is the catalyst bulk density, leading to the steady-state
design equation:
cðA
dcA
τ= (3:265)
ρB ηA rA
c0A

As an example, let us consider an Eley-Rideal type rate equation:

kKcA
r= (3:266)
1 + KcA

When introducing this rate expression into eq. (2.265), and assuming that the cata-
lyst effectiveness factor is close to unity, the following design equation is obtained:
 
1 1 c0A 1
τ= In + ðc0A − cA Þ (3:267)
ρB kK cA k

The CSTR is the opposite of plug flow hydrodynamics, which assumes that the fluid
stream is completely mixed and the concentration in the reactor is the same as in
372 3 Engineering reactions

the reactor outlet. For a perfectly mixed CSTR at steady-state it holds that there is
no accumulation dnA/dt = 0 and thus:

n_ 0A − n_ A + ηA rA V = 0 (3:268)

Molar flows in eq. (3.268) can be substituted by concentration and volumetric

_ The volumetric flows in and out are assumed to be
flows: n_ 0A = c0A cA V_ 0 , n_ A = cA V.
equal (for example, the density is constant) V_ 0 = V. _ When introducing residence
time as the ratio of reactor volume to volumetric flow rate in the same way as for
the plug flow reactor, a design equation is then obtained:
cA − c0A
τ= (3:269)
ηA rA

and for heterogeneous catalytic reactions the design equation becomes:

cA − c0A
τ= (3:270)
αηA rA

with α = mcat/VL = ρB as the catalyst bulk density. For simple kinetics, analytical
solutions are possible, while numerical solutions provide a general approach to
model simulations and parameter estimation. For the Eley-Rideal reaction, and as-
suming ɳA = 1, the expression for the residence time becomes:
 
V 1 − 1 ðc0A − cA Þ
τ= = K + c0A x (3:271)
V_ kρB cA

Industrial reactors cannot be represented by these idealized plug flow and CSTR
models and axial dispersion concept is applied in all mentioned variations (one- or
and two-dimensional pseudo-homogeneous and heterogeneous cases).
The axial dispersion models add a second order spatial derivative to the mass
balances. For a steady-state, neglecting internal diffusion, the mass balance for a
compound i becomes:

d2 ci dðci wÞ
Da − + ri ρB = 0 (3:272)
dl2 dl
where l is the length, w is the fluid flow rate, Da is the axial dispersion coefficient.
When there is no axial dispersion Da ≈ 0 and a reactor approaches the plug flow
regime. For complete mixing Da ≈ ∞ and an axially dispersed reactor displays CSTR
behavior.
The values of the axial dispersion coefficient are difficult to predict theoreti-
cally. They are usually determined by experiments with tracers or through correla-
tions relating the Peclet number Pe = wL/Da (where L is the reactor length) with the
Reynolds number. When Pe > 100 the reactor behaves like a PFR, while values
below unity indicate a CSTR-like behavior. It should be noted that during scaling-
 3.6 Catalytic reactors 373

up hydrodynamic conditions are changed and thus the values for axial dispersion
coefficients cannot be kept constant.
Limitations in applying one single axial dispersion coefficient are related to
much more complicated flow patterns in industrial reactors. These patterns could
be of complex geometry, unusual shape or the reactors can have some compli-
cated internals influencing the flow pattern. For example, the presence of a stag-
nant zone leads to long tailing, which cannot be predicted by the axial dispersion
model.
Details on complex reactor models for different cases are presented in special-
ized textbooks on chemical reactors, for example, Salmi et al. [29], and will not be
treated here. As one example, a mass balance is given below for the liquid compo-
nent in a one-dimensional dynamic three-phase fixed-bed with axial dispersion:

dcLi dcLi d2 cLi

= − ðεL τL Þ − 1 + ðPeL εL τL Þ − 1 + ri ρ′B (3:273)
dt dz dz2
where ρ′B = ρB/ɛL, ɛL is the liquid hold-up, PeL is the liquid-phase Peclet number de-
fined by PeL = wLL/DL, DL is the liquid axial dispersion coefficient, wL is the liquid ve-
locity, L is the total reactor length, z is the dimensionless length coordinate (z = l/L), τL
is the space-time defined as τL = L/wL.
In the derivation of eq. (3.273) several assumptions and simplifications were
made, including, constant fluid velocity, applicability of the Fick’s law for gas-
liquid and liquid-solid fluxes, steady-state for the outer surface of the catalyst parti-
cle, absence of internal diffusion resistance in the pores, large values of liquid-solid
mass transfer coefficient and thus kinetic control, as well as non-volatility of the
organic liquid-phase component.
The axial dispersion model in eq. (3.273) is a dynamic one-dimensional model
with respect to the spatial coordinate, containing just two adjustable hydrodynamic
parameters, τL ɛL and PeL. This model can be used for an isothermal case without
the need to consider radial concentration gradients. For non-isothermal cases with
strong heat effects a two-dimensional model is required, consisting of a set of
coupled second order partial differential equations (PDEs). The strategy for solving
the problem is to discretize the reactor coordinate (z), i.e., to divide the area be-
tween the boundaries into grid points, where the values of derivatives are approxi-
mated, and to integrate the ODEs in time (t). The number of ODEs is the number of
components times the number of discretization points. Two methods are mainly ap-
plied for discretization – lines or orthogonal collocation. In the method of lines the
derivatives, which originate from the plug flow, should be described with backward
differences, whereas central differences should be used for the first and second de-
rivatives originated from diffusion and dispersion. In the orthogonal collocation
method the areas between collocation points are approximated by orthogonal
polynomials.
374 3 Engineering reactions

For gas-phase reactions in catalytic microreactors there are several differences

between the regular size reactors and microdevices. The flow in microchannels is
usually laminar, while in macrochannels it is turbulent. Because of the geometry of
microchannels and high surface-to-volume ratios, the diffusion paths for heat and
mass transfer are very small, surface effects dominate over volume effects, and the
surface heat transfer is important.
The Knudsen number (Kn), which represents the ratio between the main free
path of the gas molecules and the characteristic length of the flow domain (diame-
ter of the channel in this case), is, however, quite small in microchannels, as it is
several orders of magnitude below unity. When Kn > 1, which is realized in the
pores of zeolites (nm size compared to tens of microns in microchannels) the mole-
cules are most likely to collide with the walls rather than with each other. From the
modeling viewpoint Kn < 1 implies that the typical Navier-Stokes equations, which
are valid for macroflows, can be also applied for microreactors.
Depending on the thickness of the catalytic layer and on the reaction condi-
tions, diffusion limitation inside the microchannels might play a role. Moreover,
mass transfer limitation from the bulk phase to the surface of the coating could ap-
pear, mainly via molecular diffusion.
For fluidized bed reactors the ideal models of the plug flow and CSTR are very
crude approximations and the reactor design should be based on a hydrodynamic
model coupled with pilot scale experimentation.
Rather complex models for fluidized beds, such as the Kunii-Levenspiel model,
assume negligible gas flow in the emulsion phase and movement of the bubble
phase in the reactor following the characteristics of a plug flow. The cloud and
wake phases are also present, which are presumed to have similar chemical con-
tents. Fig. 3.93 illustrates the transport in these phases.

u0

Bubble Cloud Rest of

and emulsion
wake

Kbc Kcc

Fig. 3.93: The schematic structure of a fluidized bed

according to the Kunii-Levenspiel, where b, c and e refer
u0 to the bubble, cloud and emulsion phases respectively.
 3.6 Catalytic reactors 375

Details on this model are available in textbooks on chemical reactors [29].

For manufacturing of organic fine chemicals multipurpose semi-batch and
batch reactors are mainly used, equipped with necessary auxiliary equipment for
mixing, heating and cooling as well as for distillation or evaporation of solvents or
light reaction products. The reactions typically take place in the liquid-phase.
The simplified mass balance for the batch reaction does not contain any terms
related to in- or out-flow of compounds, thus:

dnA
ηA ρB rA V = (3:274)
dt

If the volume is constant dnA = d(cAV) = VdcA after some rearrangements, taking
into account that concentration is related to conversion α through cA = cA0 (1 – α)
and that at t = 0, α = 0 a design equation is obtained:
ðα
dα
t = cA0 (3:275)
− ρB ηA rA
0

which is the same as for a packed-bed reactor with the only difference that in eq.
(3.275) time is used instead of the residence time.
More complicated reactor models for three-phase systems include internal
diffusion:
0 dcLi s
1
dcLi d dr r
= εp− 1 @Dei + ri ρp A = 0 (3:276)
dt rs dr

where Dei is the effective diffusion coefficient of a species (i) in the catalyst particle,
r is the coordinate of the catalyst particle, s is a shape factor: s = 2 for spheres, s = 1
for long cylinders, s = 0 for slabs and ɛp–1 is the particle porosity.
Considerations above were related only to mass balances. However, design of
catalytic reactors should involve also energy balances. In general, it is enough for
catalytic reactor engineering to include only heat balances
8 9 8 9 8 9
> energy > > energy > > energy >
( ) ( )
energy < added = < leaving = < exchanged = energy
= − + +
accumulation > by > > by > > with > generation
: ; : ; : ;
convection convection surroundings
(3:277)

or more specifically

dQ . . X
= ðn cp TÞ0 − n cp T + UAðTc − TÞ + V rj ð − ΔHr, j Þ (3:278)
dt j
376 3 Engineering reactions

where Q is the heat (J), T the temperature of the reaction mixture (K), cp the
.
molar heat capacity (J/mol/K), n the molar flow (mol/s), U the overall heat trans-
fer coefficient (W/(m2/K) or J/s/m2/K), A the heat exchange surface area in the
reactor (m2), r the reaction rate (mol/s/m3), V the volume (m3), ΔH the enthalpy
(J/mol), t the time (s).
In batchwise-operated stirred tank reactors, the reaction mixture is homoge-
neous without any concentration and temperature gradients. The general heat bal-
ance is then

− − dT X
ðn cp + CW Þ = UAðTc − TÞ + V rj ð − ΔHr, j Þ (3:279)
dt j

−
The total heat capacity of the reactor CW is considered to be
−
temperature indepen-
dent. The same is assumed for the average specific heat Cp , which is additionally
assumed to be constant throughout the reactions. This heat balance equation to-
gether with the mass balance equation describes the batch slurry reactor behavior
and should be solved simultaneously. The time-dependent temperature profile is

dT UA da
= ðTc − TÞ + ΔTad (3:280)

dt n cp + C dt
W

With the adiabatic temperature rise defined as

Vð − ΔHr ÞcA, 0
ΔTad = (3:281)
 Þ
ðn cp + C W

where CA, 0 is the initial concentration of the key component. The temperature pro-
files are thus determined by the heat of the reaction, dependence of the reaction on
temperature, the adiabatic temperature rise and the heat removal rate. The latter is
influenced by the average temperature difference between the cooling(heating) me-
dium and reaction mixture as well as the heat exchange area A and coefficient U.
A constant coolant temperature implies a linear removal of heat with the tem-
perature in the reactor, while the exponential increase of heat generation follows
from the Arrhenius equation. Subsequently, extremely high-temperature peaks (so-
called hot sports) can be generated.
For an ideal CSTR without any concentration gradients the heat balance at
steady state is
− − X
V0 ρ0 cp0 T0 − V ρcp T = UAðTc − TÞ + V rj ð − ΔHr, j Þ = 0 (3:282)
j

where the subscript “0” corresponds to the reactor inlet.

For the ideal plug flow reactor at steady state, there are concentration and tem-
perature gradients along the reactor length without any gradients over the cross
 3.6 Catalytic reactors 377

section in the radial direction. Moreover, any axial mass and heat dispersion is
neglected.
The heat balance for a volume element in an ideal plug flow reactor is

. − dT X dA
n cp = rj ð − ΔHr, j Þ + UðTc − TÞ (3:283)
dV j
dV

The heat balance equation for a tube with a constant diameter dt takes the form

dT Uτ 4 rτ
= − ðTc − TÞ + ΔTad (3:284)
dZ ρ0 cp dt cA, 0

where the adiabatic temperature rise is defined in the following way:

ð − ΔHr ÞcA, 0
ΔTad = − (3:285)
ρ0 cp

Equation (3.284) should be solved simultaneously with the mass balance to deter-
mine the axial temperature and concentration profiles. It contains two terms, the
first one determined by the heat exchange rate and heat capacity of the mixture
and the second one is influenced by temperature dependence of the heat genera-
tion. Exponential dependence of the reaction rate in temperature implies poten-
tially prominent hot spots in case of exothermic reactions.

3.6.6 Catalyst handing in a plant

3.6.6.1 Delivery to a plant

Catalysts are usually supplied to the plant sites in steel drums and should be han-
dled as gently as possible to avoid additional attrition. In some cases, additional
sieving could be done prior to loading the catalyst into a reactor, using a simple
inclined screen without any vibrations, as such vibration screens might additionally
damage the catalyst and lead to losses of the material.

Sieving of catalysts sometimes is done after discharging them from a reactor. I was once in-
volved in techno-economical negotiations about the purchase of industrial sieves for the sepa-
ration of catalysts used in oxidation of SO2. Such screens are needed when the spent catalyst
from the top bed of a five-bed converter will be discharged after (for example) 1 or 2 years of
operation, then screened and utilized in a lower bed. The sieves can be of vibrating nature and
should be compact as they are placed at various levels of a converter. Such converters could
have catalyst loading of 250–300 m3 and be as large as a five-story building with a diameter of
10–13 m.

When the catalyst is sensitive to air moisture, the drums are kept closed until the
reactor is filled, which will be discussed in the following section.
378 3 Engineering reactions

3.6.6.2 Catalyst charging

Catalyst charging in adiabatic fixed beds (Fig. 3.94) and multitubular reactors is an
extremely important part of catalyst handling in the plant. Correct packing and
even distribution ensure a satisfactory gas flow through the bed, while poor pack-
ing can lead to an unsatisfactory performance because of uneven gas flow.

Charging tube

More larger particles- More smaller particles-

low pressure drop high pressure drop
Catalyst support Support grid

Fig. 3.94: Segregation of particles with one-spot charging.

Generally, when the catalyst is charged into a vessel it should not have a free fall
of more than 0.5–1 m. Some catalyst forms (spheres) withstand a drop better than
extrudates or soft pellets. Pouring a catalyst from a charging tube into a vessel
results in a particle hill in one spot (Fig. 3.94), which should be eventually
smoothed by raking or by a wooden plate. The charging tube is raised slowly to
allow a controlled flow of catalyst in the vessel.
The heap is segregated, with smaller particles and dust staying in the center.
This leads to a higher pressure drop and low flow velocity, while close to the
edge the gas velocity is higher because of larger particles. The variations in the
bed voidage could be ~10%. It should be noted that the uneven loading results
not only in the distribution of the gas flow through the bed and thus pressure
drop variations, but also in radial temperature gradients at the bed outlet. For
example, in methanol synthesis with a temperature of 220 °C at the inlet and the
average outlet temperature of 290 °C, a very pronounced hot spot corresponding
to the hill center is observed, with the maximum outlet temperature of 315 °C.
In a dense loading method (Fig. 3.95), a special rotating device is used to
spread the catalyst more evenly and diminish the void spaces and bridging. Higher
catalyst bed density allows to load more catalyst into the same reactor volume and
decrease channeling at the expense of a higher pressure drop.
 3.6 Catalytic reactors 379

Fig. 3.95: Dense loading method principle [76].

Obviously, deep narrow beds are easier to charge, averaging of irregularities,

while such even distribution might be challenging for wide beds (where charging is
done manually with a conveyor, Fig. 3.96) represented, for example, by some sulfur
dioxide oxidation reactors (Fig. 3.97) or radial flow beds.

Big bag
Chute or (catalyst)
conveyor belt

Bin

Platform
Outlet slide
Flexible hose

Converter Ground level

Fig. 3.96: Charging of a converter with a conveyor.

380 3 Engineering reactions

Fig. 3.97: The upper two beds of an SO2 oxidation reactor. Access points are indicated on the right.

For a wide and narrow reactor (as shown in Fig. 3.97), operators have to enter
the vessel through the side manholes during charging/discharging or when examin-
ing the bed at regular shut-downs. As walking on catalyst beds is not permitted,
planks should be used. Although, in principle, raking can be used to evenly distrib-
ute the catalyst, this creates fines, particularly close to the manhole. Therefore, for
such wide and shallow beds a small conveyor could be used, if it can fit into the
side manhole and is long enough to be used across the vessel.

The size of access points on Fig. 3.97 might be a bit exaggerated. I had to take samples of the
spent sulfuric acid catalyst by entering a reactor for SO2 oxidation through a side manhole 1 m2
while wearing overall protection and a gas mask, which was not very comfortable. The non-
uniformity in distribution and also the “waves” of catalysts, similar to those in Fig. 3.97, were
clearly visible as the bed had a diameter of 13 m and a catalyst bed height of ~0.5 m.

Special care should be also taken when charging catalysts in multitubular reactors
[Fig. 3.98, such as those described in detail in Chapter 4 for steam reforming of
methane and o-xylene oxidation)].

(A) (B)

Fig. 3.98: Loading of multitubular reactors: (A) sock loading (from [77]); (B) loading using a brush
(from [78]).
 3.6 Catalytic reactors 381

A sock method can be used for steam reforming, where the catalyst is first put in
socks of 1.5–2 m in height and with an outside diameter that should be 15 mm less
than the tube diameter, otherwise sliding will be troublesome (larger diameter) or un-
folding of the lower portion of the sock might be incorrect (smaller diameter). The
loose end of the sock is folded 20 cm from the bottom. The sock is introduced into the
reactor and a strong cord, which is hooked onto the top, is then pulled, unfolding
the loose end and releasing the catalyst. This is followed by topping up the tubes
with the catalyst to an exact level, careful checking of the catalyst level and measur-
ing the pressure drop in each tube. The pressure drop across each tube should be
within 5% deviation of the mean value. Higher values of pressure drop indicate the
presence of broken catalyst particles and require the tube to be discharged, the bro-
ken extrudates removed, and the tube to be recharged. When the pressure drops, low
gaps in the packing are present and vibration of the tube is required. In fact, the
tubes are vibrated periodically during charging either by electric vibrators or by
gently hammering the top flange with a soft-faced hammer. A more detailed descrip-
tion with illustrations is available in [79].
Tab. 3.11 contains data for temperature and conversion variations (slippage of
methane) obtained for a case of equal pressure drops, and for tubes showing pressure
drop (or flow rate) variations. As shown in Tab. 3.11, when there are no variations in
the flow rate, the methane slippage is 3.64%, while the methane slippage is higher
when deviations are significant. In the latter case approach to equilibrium, which is a
good indication of the catalyst activity, displays a high value of 9 °C, sometimes in-
correctly ascribed to poor catalyst activity rather than to unsuccessful catalyst load-
ing in the tubes. The catalyst performance is worsened further with operation time,
as activity of the catalyst in different tubes will decline in a different way.

Tab. 3.11: Temperature and conversion variations methane steam reforming with differently loaded
catalysts.

Tubes with gas-flow deviations No deviations

Flow rate
% of tubes      all tubes 
% of mean      
T at outlet (°C)       
CH slippage
[% (dry gas)] . . . . . . .
Approach to equilibrium       
(°C)

(From [80]).

For multitubular reactors, other methods of catalyst loading are used besides the
socks method. The SpiraLoad method, developed by Haldor Topsoe A/S, uses
382 3 Engineering reactions

loading equipment that consists of a number of equally sized tube sections with
spiral-shaped guide elements placed along the inner walls. The tube sections can
be assembled and inserted in the reformer tubes to form one long loading tube
(Fig. 3.99).

(A) (B)

Fig. 3.99: The SpiraLoad method: (A) general scheme, (B) industrial practice. (From [81].).

Among other methods that were used in the past for loading steam reformer tubes
(and that are still applied for other reactors), the use of ropes that contain knots
along the rope length after ~1 m is of note. The catalyst can be loaded directly from
drums. When the catalyst is experiencing free fall, the knots stop the fall and pre-
vent catalyst damage and the formation of fines. Another option is to use metal
brushes (Fig. 3.98B) attached to either a rope or a metal stick. Obviously, during
loading the rope should be continuously pulled.

The procedure of using a rope with knots might sound very primitive, however I was once in-
volved in the charging of a dehydrogenation multitubular reactor with 600 tubes, which was
successfully loaded with catalyst tablets using this method.

For loading of other multitubular reactors, such as those used in o-xylene oxida-
tion, special loading devices are available and are provided by the catalyst manu-
facturers. When complicated catalyst profiling along the tubes is required, as will
be discussed in detail in Chapter 4, thousands of reactor tubes are first loaded with
the reactor bottom type of catalysts, pressure drop is carefully monitored, the level
of this catalyst layer is adjusted and the procedure is repeated with all other types
of catalysts.
 3.6 Catalytic reactors 383

The justification for such a tedious procedure is that correct loading is very impor-
tant for successful catalyst performance. Catalyst loading can be therefore very time-
consuming and take up to several weeks depending on the catalyst and reactor types.

3.6.6.3 Catalyst activation, shutdown, restarts and discharging

The catalyst performance in a plant and its lifetime are influenced not only by the
way the catalyst is charged into a reactor, but also by pre-treatment, activation and
operation.
An activation procedure might involve, for example, sulfidation or reduction of
metal oxide. When hydrodesulfurization catalysts are charged directly as supported
oxides and put in contact with a feed containing sulfur compounds, the catalyst is
progressively transformed into sulfides. Sulfidation of the precursor oxide can be
performed either outside of the plant or specifically at the plant prior to regular
operation.
Similarly, reduction of metal oxides can be done on-site and should be carefully
controlled to achieve the best performance. For example, the reduction of palla-
dium catalysts used in selective hydrogenation of acetylenic compounds is done
after the first start-up and each regeneration. After regeneration by burning coke,
the reactor is blanketed from air by nitrogen. The heating is done under inert gas
flow up to the required temperature (~120 °C) with a specific heating gradient, not
exceeding 50 °C per hour. When the reduction temperature is reached, hydrogen is
slowly added to the inert gas. The temperature increase should be marginal to
avoid overheating. The reduction proceeds for 10–12 h. The catalyst is cooled down
to a normal operation temperature with the subsequent replacement of hydrogen
with nitrogen, which is then used when the reactor is in idle mode.
Another more detailed reduction procedure of iron oxide an ammonia synthesis
catalyst, is described in Chapter 4.

When I was working for a catalyst supplier almost two decades ago, I took part in the commis-
sioning of the revamp of an ammonia synthesis reactor from axial to radial flow. The long-
term performance of the catalyst was of major concern, not only for the plant operator, but
also for the catalyst supplier, as it was part of the warranty. Thus, catalyst reduction with con-
trolled temperature increase while monitoring the concentration of water exiting the reactor
was done. As uneven reduction could take place, which is especially pronounced in radial
flow converters because of shallow beds compared to axial flow reactors, special care was
taken with respect to the gas flow. During the reduction procedure the mantra of the catalyst
supplier was: “more flow”.

During shut-downs and restarts, damage to the catalyst should be prevented, espe-
cially when the catalyst is in the reduced state and is pyrophoric.
384 3 Engineering reactions

I had once to fly by private jet to a remote plant site, where there was an emergency situation
resulting in overheating of a WGS catalyst. Unfortunately, the catalyst had been damaged and
despite all the efforts of the plant personnel to make a careful shutdown by slow cooling, the
catalyst charge (which in that reactor was exceeding a hundred tons) had to be replaced.
In that case, the local overheating, which apparently exceeded 900 °C, was a result of an
accident.
Two other examples not from my personal experience, but also related to WGS, illustrate
more negligence during catalyst start-up and reduction. After loading the newly developed
high-temperature WGS catalyst based on iron and lead (the story is from 1947 and happened
in Novomoskovsk) in the reactor, there was a profound runaway when the catalyst was ex-
posed to a mixture of 50% CO and 50% H2. Apparently, a possibility of such runaway with a
more active than the previous generation Fe–Mg catalyst was unnoticed in lab-scale experi-
ments. The whole reactor was damaged, as not only the catalyst melted, but the same hap-
pened with the catalyst grid and thermocouple sockets.
In another example of a profound overheating, now from ammonia plant in Rovno in 1974,
the low-temperature WGS catalyst based on copper experienced temperature of 1,350 °C (in-
stead of 250 °C). During a start-up of the reactor with 80 tons of catalyst, after reaching 200 °
C under an inert gas (nitrogen), two plant operators decided to bring some innovation into the
catalyst reduction procedure. Against all instructions instead of adding 0.3% of hydrogen,
they decided that 15% of hydrogen would be more efficient to shorten the reduction time.
Obviously, after the accident the whole catalyst charge had to be replaced and the damaged
walls of the reactor had to be repaired.

For some catalysts that gradually deactivate, regeneration is required from time to
time. As already mentioned, a palladium catalyst that is active in selective hydro-
genation of triple bonds needs to be regenerated after ~6 months of operation.
Regeneration can be undertaken with a steam/air mixture at temperatures between
400 and 450 °C. After blocking the process gas flow and depressurizing the reactor,
it is heated under an inert gas (nitrogen) at a certain temperature ramp (~40 °C h−1).
At ~150 °C the inert gas is continuously replaced by overheated steam while heating
to 400 °C. At this point, air is added to the steam and the temperature might in-
crease slightly. Regeneration is carried out while controlling the outlet concentra-
tion of CO2. At a certain value of CO2 slip (~500 ppm), which indicates almost
complete burn-off of carbon deposits, the air is switched off and cooling is carried
out under steam to ~200 °C. Below this temperature cooling continues with
nitrogen to the required reduction temperature (~120 °C, as mentioned above).
Thereafter the catalyst is reduced. Reduction procedures, available from catalyst
suppliers, typically include detailed information about exact values of tempera-
tures and flow rates of steam, nitrogen, air, hydrogen, etc.
As air should not be in contact with the catalyst, an inert atmosphere (nitrogen
or process gas) is kept in the plant where it is off-line. Nitrogen can be considered a
suitable gas, as the oxygen content in it is below 10 ppm. Blanketing of the reduced
catalyst is typically done under slight pressure with static nitrogen. When the reac-
tor must be opened under a nitrogen blanket, the rule is to have one opening at a
 3.6 Catalytic reactors 385

time. In some cases natural gas can be also applied for blanketing, however, it rep-
resents a fire hazard.
Discharging a catalyst that is not going to be used further is usually a simple
process, where the discharge is done by gravity through the bottom of a vessel.
Alternatively, suitable chutes, conveyors or vacuum equipment could be used to
carry the catalyst out into drums or trucks.
Special care should be taken when a pyrophoric catalyst is discharged, as it
should be done under a nitrogen blanket to prevent oxidation, overheating and po-
tential damage to the reactor vessel. When the catalyst does not react with water,
resulting in a build-up of hydrogen, filling of the vessel with water and draining it
several times can be done prior to discharge as an alternative to discharge under
blanketing.
Top discharge for a vessel with a large top opening could even be done by shov-
eling the catalyst into buckets attached to ropes and pulled out of the reactor.
Obviously, the proper ventilation should be done prior to discharge and the person-
nel should have appropriate protective clothing.
Another possible method is for a top discharge is to use industrial vacuum
cleaners.

Sometimes catalyst discharge with vacuum cleaners is complicated by the required precision
regarding the exact amount of catalyst to be removed. In one industrial example that I was in-
volved in, one-third of the catalyst charge had to be withdrawn from a multitubular reactor be-
cause of the high-pressure drop, with a target of minimizing deviations in pressure drop in
each tube after partial catalyst discharge.

3.6.6.4 Metal recovery

Catalysts containing nickel, copper, zinc and precious metals (PMs) may be sent for
the recovery of these metals. The economics of such refining depends on the
amount of metal, the plant location, as well as current metal prices. Thus the value
of precious metals in catalysts easily exceeds the price of the catalyst and signifi-
cantly exceeds the recovery cost. Moreover, PM recovery provides a safe and en-
vironmentally responsible disposal of contaminated catalysts. Recovery and recycle
of PMs also minimizes the effect of metal market price fluctuations.
One of the most important economical issues is an accurate, transparent and reli-
able determination of PM content in the spent catalyst. A 2% deviation from this dur-
ing the catalyst handling, weighing and/or sampling corresponds to a 2% deviation
in PM credit, which for a catalyst with Pt content of about 0.3% and a price of $2100/
troz is translated into a US $400,000 discrepancy for 10 tons of spent catalyst.
The refining process can be either pyro-metallurgical (i.e., smelting) or a hydro-
metallurgical recovery (i.e., a wet chemical process). The modern smelters offer a
broad range of applications as, for example, refining Pt, which is not dissolved in
386 3 Engineering reactions

sodium hydroxide or sulfuric acid, the same can hold for such insoluble supports
as carbon or zeolites.
A flow diagram of the supercritical water oxidation process, based on the
Chematur Engineering supercritical water oxidation process, is presented in Fig. 3.100.
Water is pressurized to ~240 bar and then heated to ~385 °C to ensure formation of the
supercritical phase. Heterogeneous catalysts are fed in a slurry form from the feed tank
through an economizer prior to oxygen treatment. Homogeneous catalysts are injected
into the reactor after the oxygen points to avoid the pyrolysis of elements such as phos-
phorus during preheating.

Steam CO2, N2, O2

Steam Pressure reduction

generator
Separator

Cooler

Clear water +
precious metal
Economizer oxide
Boiler water Heater
feed pump

Oxygen High
pressure
Hetero- pump
geneous
catalyst Homogeneous
+ water catalyst
Feed Reactor
tank

Centrifugal Quench water

pump

Fig. 3.100: Johnson Matthey-Chematur Engineering process flow scheme. (From [82]). Copyright
Johnson Matthey.].

References
[1] Kusema, B.T., Mikkola, J.-P., Murzin, D.Yu. (2012) Kinetics of L-arabinose oxidation over
supported gold catalyst with in-situ catalyst electric potential measurements. Catal. Sci.
Technol. 2: 423–431.
[2] Murzin, D.Yu., Salmi, T. (2016) Catalytic Kinetics. Science and Engineering, Amsterdam, The
Netherlands: Elsevier.
[3] Murzin, D.Yu. (1995) Modeling of adsorption and kinetics in catalysis over induced
nonuniform surfaces: surface electronic gas model. Ind. Eng. Chem. Res. 34: 1208.
 References 387

[4] Kiss, V., Osz, K. (2017) Double exponential evaluation under non-pseudo –first order
conditions. A mixed second-order process followed by a first-order reaction, Int. J. Chem.
Kinet., 49: 602–610.
[5] Lente, G., (2015), Kinetics of irreversible consecutive processes with first order second steps:
analytical solutions, J. Mathem. Chem., 53: 1172–1183.
[6] Murzin, D.Yu., Simakova, I.L., Wärnå, J. (2019) Selectivity analysis for networks comprising
consecutive reactions of second and first order, Int. J. Chem. Reactor Eng, 17: https://doi.
org/10.1515/ijcre-2018-0161.
[7] Boudart, M. (1969) Catalysis by supported metals. Adv. Catal. 20: 153–166.
[8] Murzin, D.Yu. (2010) Size dependent heterogeneous catalytic kinetics. J. Mol. Catal. A. Chem.
315: 226.
[9] Murzin, D.Yu. (2010) Kinetic analysis of cluster size dependent activity and selectivity.
J. Catal. 276: 85.
[10] van Hardeveld, R., Hartog, F. (1969) The statistics of surface atoms and surface sites on metal
catalysts. Surface Sci. 15: 189.
[11] Lazman, M.Z., Yablonskii, G.S. (2008) Overall reaction rate equation of single-route complex
catalytic reaction in terms of hypergeometric series. Adv. Chem. Eng. 34: 47.
[12] Glasstone, S., Laidler, K., Eyring, H. (1941) The Theory of Rate Processes. New York: McGraw
Hill.
[13] Eyring, H. (1935) The activated complex in chemical reactions. J. Chem. Phys. 3:107.
[14] Temkin, M. (1938) Transition state in surface reactions. Acta Physicochimica. URSS, 8:141
[15] Campbell, C. T., Arnadottir, L., Sellers, J.R.V. (2013), Kinetic prefactors of reactions on solid
surfaces, Z. Phys. Chem. 227: 1.
[16] Dumesic, J.A., Rudd, D.F., Aparicio, L.M., Rekoske, J.E., and Trevino, A.A. (1993) The Micro-
Kinetics of Heterogeneous Catalysis. ACS Professional Reference Book, American Chemical
Society, Washington, DC.
[17] Kiselev, V. D., Miller, J. G. (1975) Experimental proof that the Diels-Alder reaction of
tetracyanoethylene with 9,10-dimethylanthracene passes through formation of a complex
between the reactants, J. Am. Chem. Soc., 97: 4036.
[18] Kiselev, V. D., Konovalov, A.I. (2009) Internal and external factors influencing the Diels–Alder
reaction, J. Phys.Org. Chem. 22: 466.
[19] Han, X., Lee, R., Chen, T., Luo, J., Lu, Y., Huang, K.-W. (2013) Kinetic evidence of an apparent
negative activation enthalpy in an organocatalytic process, Sci. Rep., 3: 2557.
[20] Wei, J. (1996), Adsorption and cracking of n-alkanes over ZSM-5: Negative activation energy
of reaction. Chem. Eng. Sci., 51: 2995.
[21] Moulijn, J., Diepen, van A.E., Kapteijn, F. (2001) Catalyst deactivation: is it predictable? What
to do? Appl. Catal. A-Gen. 212:3.
[22] Cornish-Bowden, A. (2001) Detection of errors of interpretation in experiments in enzyme
kinetics. Methods 24: 181.
[23] Kapteijn, F., Berger, R.J., Moulijn, J.A. (2008) Rate procurement and kinetic modeling. In: Ertl,
G., Knözinger, H., Schüth, F., Weitkamp, J., editors. Handbook of Heterogeneous Catalysis.
Weinheim, Germany: Wiley-VCH, p.1693.
[24] Gianetto, A., Silveston, P.L. (1986) Multiphase Chemical reactors. Theory, Design and Scale-
Up. Hemisphere Publishing Corporation, Washington.
[25] Hájek, J., Murzin, D.Yu. (2004) Liquid-phase hydrogenation of cinnamaldehyde over Ru-Sn
sol-gel catalyst. Part I. Evaluation of mass transfer via combined experimental/theoretical
approach. Ind. Eng. Chem. Res. 43: 2030–2038.
[26] Murzin, D.Yu., Konyukhov, V.Yu., Kul’kova, N.V., Temkin, M.I. (1992) Diffusion from surfaces
of suspended particles and specific mixing power in shaker reactors. Kinet. Catal. 33:728.
388 3 Engineering reactions

[27] Hemrajani, R., Tatterson, G.B. Chapter 6. Mechanically Stirred Vessels. In: Paul, E.L., Atiemo-
Obeng, V.A., Kresta, S.M. Handbook of Industrial Mixing: Science and Practice, DOI: 10.1002/
0471451452.ch6. John Wiley and Sons. Hoboken, NJ, USA.
[28] Sherwood, T.K., Pigford, R.L., Wilke, C.R. (1975) Mass Transfer. New York: McGraw-Hill.
[29] Salmi, T.O., Mikkola, J.-P., Wärnå, J.P. (2019) Chemical Reaction Engineering and Reactor
Technology. Second Edition. Boca Raton, FL: CRC Press.
[30] Kolodziej, A., Lojewska, J. (2013), Engineering aspects of catalytic converter designs for
cleaning of exhaust gases. In: S.L. Suib, A.V. Sapre, R.J. Katzer (Eds.), New and Future
Developments in Catalysis: Catalysis for Remediation and Environmental Concerns.
Amsterdam, p.257.
[31] Dittmeyer, R., Emig, G. (2008) Simultaneous heat and mass transfer and chemical reaction.
In: Ertl, G., Knözinger, H., Schüth, F., Weitkamp, J., Editors. Handbook of Heterogeneous
Catalysis. Weinheim, Germany: Wiley-VCH, pp. 1727–1784.
[32] Moulijn, J.A. (2003) NIOK course on catalysis, Lecture notes.
[33] Ioffe, I.I., Reshetov, V.A., Dobrotvorskii, A.M. (1985) Heterogeneous catalysis. Leningrad,
Chimia.
[34] Sandelin, F., Oinas, P., Salmi, T., Paloniemi, J., Haario, H. (2006) Dynamic modelling of
catalytic liquid-phase reactions in fixed beds-Kinetics and catalyst deactivation in the
recovery of anthraquinones. Chem. Eng. Sci. 61: 4528–4539.
[35] Leveneur, S., Wärnå, J., Eränen, K., Salmi, T. (2011) Green process technology for peroxycar-
boxylic acids: Estimation of kinetic and dispersion parameters aided by RTD measurements:
Green synthesis of peroxycarboxylic acids. Chem. Eng. Sci. 66: 1038–1050.
[36] Aris, R. (1976) The Mathematical Theory of Diffusion and Reaction in Permeable Catalysis,
vol.1. Oxford: Clarendon Press.
[37] Mears, D.E. (1970) Diagnostic criteria for heat transport limitations in fixed bed reactors.
J. Catal. 20: 127.
[38] Mears, D.E. (1971) Tests for transport limitations in experimental catalytic reactors. Ind. Eng.
Chem. Proc. Dd. 10: 541.
[39] Anderson, J.B. (1963) A criterion for isothermal behavior of a catalyst pellet. Chem. Eng. Sci.
18: 147.
[40] Weisz, P.B., Prater, C.D. (1954) Interpretation of measurement in experimental catalysis. Adv.
Catal. 6: 143.
[41] Villermaux, J. (1993) Génie de la réaction chimique – Conception et fonctionnement des
réacteurs. Tec&Doc Lavoisier. Paris.
[42] Weisz, P.B., Hicks, J.S. (1962) The behaviour of porous catalyst particles in view of internal
mass and heat diffusion effects. Chem. Eng. Sci. 17: 265.
[43] Koros, R.M., Nowak, E.W. (1967) A diagnostic test of the kinetic regime in a packed bed
reactor. Chem. Eng. Sci. 22: 470.
[44] Madon, R.J, Boudart. (1982) Experimental criterion for the absence of artifacts in the
measurement of rates of heterogeneous catalytic reactions. Indust. Eng. Chem. Fund. 21: 438.
[45] Temkin, M., Kul’kova, N. (1969) Ideal-displacement laboratory reactor. Kinet. Katal., 10: 461.
[46] Mestl, G. (2012). High throughput development of selective oxidation catalysts at Sued-
Chemie, Combinat. Chem High Throughput Synth. 15: 144.
[47] http://www.wraconferences.com/wp-content/uploads/2016/11/ILS-ERTC-Presentation-V3b-
2.pdf
[48] Temkin, M. I., Kiperman, S. L., Luk’yanova, L. I. (1950) Flow-circulation method of
investigation of the kinetics of heterogeneous catalytic reactions. Dokl. Akad. Nauk SSSR 74:
763.
 References 389

[49] Stitt, E.H. (2002) Alternative multiphase reactors for fine chemicals: a world beyond stirred
tanks? Chem. Eng. J. 90: 47.
[50] Bartolomew, C.H., Farrauto, R.J. (2006) Fundamentals of Industrial Catalytic Processes. Wiley.
Hoboken, NJ.
[51] Eigenberger, G., Ruppel, W. (2012) Catalytic Fixed-Bed Reactors, Ullmann’s Encyclopedia of
Industrial Chemistry. 10.1002/14356007.b04_199.pub2.
[52] http://www.uhde.eu/en/competence/technologies/hydrogen/129/133/methanol-reactors.
html.
[53] Gunjal, P.R., Ranade, V.V. (2016) Catalytic Reaction Engineering, in Industrial Catalytic
Processes for Fine and Specialty Chemicals, http://dx.doi.org/10.1016/B978-0-12-801457-
8.00007-0, Elsevier.
[54] Mederos, F.S., Ancheyta, J. (2007) Mathematical modeling and simulation of hydrotreating
reactors: Cocurrent versus countercurrent operations. Appl. Catal. A: Gen. 332: 8.
[55] Sato, Y., Hirose, T., Takahashi, F., Toda, M. (1972), Performance of fixed-bed catalytic reactor
with cocurrent gas liquid flow, Pac. Chem. Eng. Cong, Sess. 8, Pap. 8–3: 187–196.
[56] Mederos, F.S., Ancheyta, J., Chen, J. (2009) Review on criteria to ensure ideal behaviors in
trickle-bed reactors. Appl. Catal. A: G. 355: 1.
[57] Wang, T., Wang, J., Jin, Y. (2007) Slurry reactors for gas-to-liquid processes: a review. Indust.
Eng. Chem. Res. 46: 5824–5847.
[58] Kantarci, N., Borak, F., Ulgen, K.O. (2005) Bubble column reactors. Proc. Biochem. 40: 2263.
[59] Deckwer, W.-D., Louisi, Y., Zaidi, A., Ralek, M. (1980) Hydrodynamic Properties of the Fischer-
Tropsch Slurry Process. Ind. Eng. Chem. Proc. DD. 19: 699.
[60] Post, T. (2010) Understanding the real world of mixing. Chem. Eng. Proc. 25–32.
[61] Daniel A. Hickman, D.A., Holbrook, M.T., Mistretta, S., Rozeveld, S.J. (2013), Successful scale-
up of an industrial trickle bed hydrogenation using laboratory reactor data, Ind. Eng. Chem.
Res. 52: 15287–15292
[62] Trambouze, P., van Landeghem, H., Wauquier, J.P. (1988) Chemical Reactors – Design/
Engineering/Operation. Paris: Editions Technip.
[63] Loosdrecht, van de, J., & Niemantsverdriet, J. W. (2012). Synthesis gas to hydrogen,
methanol, and synthetic fuels. In R. Schlogl (Ed.), Chemical Energy Storage. Walter de Gruyter
GmbH, pp. 443–458. DOI: 10.1515/9783110266320.443.
[64] http://what-when-how.com/energy-engineering/coal-to-liquid-fuels-energy-engineering/
[65] van der Laan, G.P. (1999) Kinetics, Selectivity and Scale Up of the Fischer-Tropsch Synthesis,
PhD Thesis, University of Groningen.
[66] Advanced liquid biofuels synthesis. Adding value to biomass gasification. ECN-E–17-057 –
February 2018, www.ecn.nl
[67] Moulijn, J. A., Makkee, M.I., van Diepen, A.E. (2013) Chemical Process Technology, 2nd
Edition. Chichester, Wiley.
[68] http://www.inclusive-science-engineering.com/wp-content/uploads/2012/01/Production-of-
Methanol-from-Synthesis-Gas.png
[69] http://www.gbhenterprises.com
[70] http://www.gbhenterprises.com/methanol%20converter%20types%20wsv.pdf
[71] http://www.slideshare.net/GerardBHawkins/methanol-flowsheets-a-competitive-review
[72] Salehi, K., Jokar, S.M., Shariati, J., Bahmani, M., Sedghamiz, M.A., Rahimpour, M.R. (2014)
Enhancement of CO conversion in a novel slurry bubble column reactor for methanol
synthesis, J. Natur Gas Sci. Eng. 21: 170.
[73] https://www.osti.gov/servlets/purl/823132
[74] Tirronen, E., Salmi, T. (2003) Process development in the fine chemical industry. Chem. Eng.
J. 91: 103.
390 3 Engineering reactions

[75] Dutta, S., Gualy, R. (2000) Reactor models. Chem. Eng. Prog. N10, 37–51.
[76] https://www.catalystseurope.eu/images/Documents/ECMA1004q_-_Catalyst_handling_
best_practice_guide.pdf
[77] http://www.crealyst.fr/lang/en-us/activities-2.
[78] http://www.catalysthandling.com/content/services/catalyst-loading-services.
[79] Twygg, M. (1997) Handbook of Industrial Catalysis. Manson Publishing.
[80] Farnell, P.W. Methods of charging and discharging of steam reforming catalysts, ICI Katalco,
410W/037/0/RUS.
[81] http://www.topsoe.com/business_areas/hydrogen/~/media/PDF%20files/Steam_reform
ing/ Topsoe_spiraload_technology.ashx.
[82] Grumett, P. (2003) Precious metal recovery from spent catalysts. Platin. Met. Rev. 47: 163.

Brought to you by | The University of Texas at Austin

Authenticated
Download Date | 2/15/20 10:09 AM
4 Engineering technology

4.1 General structures of chemical processes

This chapter will discuss the practical implementation of heterogeneous catalytic

processes. The general structure of chemical processes is presented in Fig. 4.1.

Raw
materials Physical
treatment

Reactor

Physical Products
treatment

Fig. 4.1: General structure of chemical processes.

As can be seen in Fig. 4.1 the main features of the chemical processes could be
identified as: feed purification, reaction, separation and product purification.
Chemical technology is the science of those operations, which convert raw ma-
terials into desired products on an industrial scale, applying one or more chemical
conversions. A new technology can succeed only if it is robust, reliable, safe, clean,
cheap, easy to control, and if it provides significant gains over existing processes.
Moreover, technology should be based on sound economical considerations, safety
requirements and labor conditions.
It is important to note that in chemical technology the process should be
viewed in its whole complexity, rather than a combination of individual steps. For
example, the performance of a reactor unit can depend on the performance not
only of the units located upstream to reactor vessel, but also downstream.
Obviously the upstream units would influence the inlet composition or the purity of
the feedstock, and thus have an impact on the reactor. Selectivity changes made by
adjusting the catalyst performance would influence separation. A loss of pressure
downstream from the reactor could lead to an increase of pressure in the reactor
and might damage the catalyst support grid. Improvement of one unit (reactor) usu-
ally improves the overall performance. Thus, a slight improvement in catalyst selec-
tivity would sometimes result in very high savings in separation.

https://doi.org/10.1515/9783110614435-004
392 4 Engineering technology

It should be also mentioned that optimum conditions for a system element are
not necessarily optimum for the system as a whole. Thus the optimization of a par-
ticular chemical technology should include not only the reactor unit, but other
units as well.
Process technology design includes a number of aspects to be considered: the
reactions involved, thermodynamics, operating parameters such as temperature and
pressure, kinetics, conversion levels and thus a need for recycle, separation of prod-
ucts, types of catalysts (homogeneous vs heterogeneous), catalyst stability, deactiva-
tion and regeneration if needed, safety aspects, environmental issues, corrosion, etc.
Such process design involves quite complicated flow charts and is not a straight-
forward application of the disciplines on which chemical technology is based (chem-
istry, physical transport, unit operations, reactor design), but rather the integration
of this knowledge.
Complications also arise from having to choose from many possibilities, taking
into account product markets, geographical location, social situation, legal regula-
tions, etc. and that the final result must be economically attractive.
As an example, we can consider the synthesis of caprolactam from benzene
(Fig. 4.2). In fact, various options are available: hydrogenation of benzene in the
liquid or gas phases to cyclohexane; partial hydrogenation of benzene to

OH

Oxidation
OH (5%)
Hydrogenation, Pd,
O (95%)
Oxidation
Adipic acid +

AH salt Distillation O
H2O
OH
Polyamide 6.6 –H2

NH2OH O

NOH NH

H2O2 + NH3

Oxime Caprolactam

Fig. 4.2: Several routes for caprolactam synthesis.

 4.1 General structures of chemical processes 393

cyclohexene; formation of phenol from benzene first with subsequent hydrogena-

tion to cyclohexanone or cyclohexanol. Several options are available for
Beckmann rearrangement, synthesis of cyclohexanone oxime and even the pro-
duction of hydroxylamine. For example, Beckmann rearrangement of cyclohexa-
none oxime to caprolactam can be done in the liquid phase using sulfuric acid as
a catalyst or alternatively in the gas phase over a zeolitic catalyst.
In addition to the routes sketched in Fig. 4.2 other options are also possible,
thus for instance BASF-Du Pont technology relies on synthesis of adiponitrile
(ADN) from butadiene by two-step hydrocyanation, partial hydrogenation of ADN
to hexamethylenediamine and 6-aminocapronitrile, with this being converted to
caprolactam through a cyclization reaction.
Many of the routes presented in Fig. 4.2 are used commercially, emphasizing
the fact that there is not only one way to produce caprolactam, and many variants
for a particular technology are possible.
Regarding the requirements for catalysts that were mentioned in Chapter 1 – they
should be active, selective and stable. For reactors, a requirement of high selectivity
means low capacities and thus savings in capital costs. High selectivity also implies
few separation steps for product purification. Catalyst stability is an important issue
and will be discussed below in greater detail.

4.1.1 Safety in design

The need for adequate design of a particular process, including safety issues,
should not underestimated. It is helpful to consider the worst industrial disaster of
the twentieth century, which occurred on December 2nd 1984 at the Union Carbide
Corporation, in the city of Bhopal, in India, which had a population of about one
million people. Over 40 tons of methyl isocyanate (MIC), as well as other lethal
gases including HCN leaked from the plant side to the city. There are different num-
bers available in the literature about the casualties, but according to the Bhopal
Peoples Health and Documentation Clinic, 8,000 people were killed in immediate
aftermath of the leak, and over 500,000 people suffered from injuries.
On the night of the disaster, water that was used for washing the lines entered the
tank containing methyl isocyanate through leaking valves. The refrigeration unit de-
signed to keep the methyl isocyanate close to 0 °C had been shut off in order to save
on electricity costs. The entrance of water to the tank, which was full of methyl isocya-
nate at an ambient temperature, initiated an exothermic runaway process and the sub-
sequent release of the gases. The safety systems, which were not properly designed to
handle such runaway situations, were non-functioning and under repair.
Unfortunately workers ignored early signs of disaster, as gauges measuring
temperature and pressure in the various parts of the unit, including the methyl iso-
cyanate storage tanks, were known to be unreliable.
394 4 Engineering technology

It was assumed that the methyl isocyanate would be kept at low temperatures by
the refrigeration unit; however, this unit was shut off. In addition the gas scrubber,
meant to neutralize methyl isocyanate if released, had been shut off for maintenance.
In any case, the design was inappropriate, as the maximum designed pressure
was only 25% of that actually reached during the disaster.
Moreover, the flare tower, which was installed to burn off escaping methyl iso-
cyanate, was not in operation awaiting the replacement of a corroded piece of pipe.
Even if it had been in operation it would have only been able to process a fraction
of the gas released. There were some other reasons for the disaster, such as a too
short water curtain, a lack of effective warning systems and a failure of the alarm
on the storage tank to signal a temperature increase. Overfilling of the storage tank
beyond the recommended capacity and methyl isocyanate filling a reserve tank,
which was supposed to be empty, are added to the overall picture.

I remember the day in 1984 that the news about the disaster broke, as I was doing some experi-
ments in the lab, synthesizing substituted carbamates using alkyl isocyanates.

In the case of the Bhopal disaster, there are many reasons for such unfortunate
events, including poor maintenance, design and inventory excess.
It would have been possible to prevent the release of methyl isocyanate if the
technology had been organized in a different way. As illustrated in Fig. 4.3, methyl
isocyanate was formed by a reaction of methylamine with phosgene, with a subse-
quent reaction with 1-naphtol. If the process had been designed in another way, for
example, phosgenation of naphthol, the release of methyl isocyanate could have
been avoided.

4.1.2 Conceptual process design: examples

Before discussing conceptual process design, a few words should be mentioned

about flow diagrams in general. These diagrams should provide a clear simple out-
line of the steps involved in the process and must cover all the major steps, includ-
ing the steps before and after the main chemical processing (Fig. 4.4).
Fig. 4.4, for example, demonstrates cyclohexanol dehydrogenation in a multi-
tubular reactor. An interesting feature is a heat exchanger upstream reactor, which
is used by the products to warm-up the reactants. The scheme also comprises a dis-
tillation column. A stream of unreacted cyclohexanol from this column back to the
reactor points out incomplete conversion during the reaction.
Another example of heat integration is shown in Fig. 4.5 for catalytic incinera-
tion of volatile organic compounds, comprising an efficient heat exchanger in addi-
tion to a burner for preheating the gases and a catalyst chamber with monolithic or
pelletized catalyst.
 4.1 General structures of chemical processes 395

CH3NH2 + COCl2 CH3N=C=O + 2 HCl

Methylamine Phosgene Methylisocyanate

O
OH OCNHCH3

CH3N=C=O +

1-Naphthol Carbaryl

O
OH OCCl

+ COCl2 + HCl

O O
OCCl OCNHCH3

+ CH3NH2 + HCl

Fig. 4.3: Alternative pathways for synthesis of carbaryl. Upper scheme: process at UC Bhopal plant,
lower scheme: alternative design.

A block type diagram is usually sufficient for understanding and evaluating the
process flow and complex engineering drawings (Fig. 4.6) are not required.
Let us consider the treatment of V.S. Beskov and V.S. Safronov in [2] as an ex-
ample of conceptual process design synthesis of nitric acid. The process consists of
several steps. Initially, ammonia is combusted to form nitric oxide:

4NH3 + 5O2 ! 4NO + 6H2 O (4:1)

This reaction is highly exothermic and occurs at high temperatures of 850–900 °C

over platinum catalysts. The residence time should be minimized (using high flow
rates) to prevent side reactions, such as:

NH3 + NO ) N2 + H2 O (4:2)

In order to achieve this, rather atypical Pt gauzes are utilized. As the reaction rate is
fast, the influence of external diffusion at such conditions can be prominent. The
influence of reaction parameters on this reaction is not straightforward. With an in-
crease of temperature the reaction rate increases at the expense of selectivity. An
increase of pressure similar to temperature enhances the reaction rate, but leads to
396 4 Engineering technology

350 °C

Flue gas
450 °C
Dehydrogenation of cyclohexanol
Gas
Light
320 °C

Cyclohexanone

160 °C
Oxidation
Cyclohexanol
Distillation

Heavies

Fig. 4.4: A diagram of cyclohexanol dehydrogenation.

Combustion
blower
Process exhaust Exhaust to
inlet atmosphere Natural gas
fired burner

Catalyst chanmber
(monolith or pelleted)

Supply fan High–efficiency

heat exchanger

Fig. 4.5: A diagram of heat integration in oxidation of VOC [1].

higher metal losses. Because of severe conditions the catalyst lifetime is limited
usually to one year.
The gauzes gradually deactivate because of platinum volatilization in the form
of platinum oxide (Fig. 4.7a).
The fresh gauze is initially activated for several hours, leading to formation of
crystallites with more active crystallographic planes and increase in the surface
area of palladium. Thereafter, gradual deactivation of the gauzes occurs because of
platinum oxide volatilization. Elevation of the severity of operation from 800oC and
 4.1 General structures of chemical processes 397

(A) (B)

Fig. 4.6: Examples of (A) engineering drawings and (B) complex piping networks.

(A) (B)

Fig. 4.7: Catalysts for ammonia oxidation: (A) comparison between the fresh and deactivated 90%
Pt–10% Rh commercial catalyst [3], and (B) an assembled pack of platinum–rhodium gauzes [4].

atmospheric pressure to 900oC and 8 bars leads to four- to eightfold increase of the
lost palladium amount. Recovery of palladium was first realized by placing a woven
Pd-rich alloy (platinum–palladium) gauze immediately below the oxidation gauze,
giving 70% recovery. Further improvement was introduction of 90% platinum/10%
rhodium alloy not only reducing platinum losses and increasing lifetime but im-
proving as well selectivity to more than 94%. Gauzes are changed when the amount
of Rh is increasing to 12% with a subsequent decrease of platinum content. As there
is a certain activation of the catalyst after the start-up, the fresh low-activity gauze
is always packed below the used gauze. Apart from Pt/Rh gauzes also 90% Pt + 5%
398 4 Engineering technology

Rh + 5% Pd alloy is used. Such gauzes are typically woven or knitted using wire
diameters between 0.06 and 0.12 mm.
A pack installed in the reactor (Fig. 4.7b) can contain up to 40 platinum–rho-
dium gauzes with the catalysis occurring in just few gauzes, while the rest helps to
mitigate deactivation and improve heat and flow distribution.
Monolith nonplatinum catalysts for oxidation of ammonia have been intro-
duced in commercial nitric acid plants (Fig. 4.8) for partial substitution of Pt-Rh
gauzes to decrease platinum losses.

Fig. 4.8: Monolith nonplatinum catalysts for oxidation of ammonia in the production of nitric
acid [5].

After the catalytic reaction in the subsequent step, oxidation of nitric oxide to nitro-
gen dioxide is done:

2NO + O2 ! 2NO2 (4:3)

This is a gas-phase, reversible exothermic reaction, which does not require the pres-
ence of any catalysts. Thereafter nitrogen dioxide is absorbed in water to form nitric
acid in a heterogeneous gas-liquid process:

4NO2 + O2 + 2H2 O ! 4HNO3 (4:4)

This reaction is more complicated as other components can react (NO, N2O3, N2O4,
etc.) in addition to NO2. In fact the main reactions happening in an absorption
tower could be written in the following way:

N2 O4 + H2 O ! HNO2 + HNO3 (4:5)

3HNO2 ! HNO3 + H2 O + 2NO (4:6)

A preliminary flow diagram for nitric acid synthesis is given in Fig. 4.9.
 4.1 General structures of chemical processes 399

Cooler Water
S10
Reactor S11
S5
Mixer 1
S8 2
S3 S4 S7
Air S1
E1 SS1
M1 S6 R2 S9 HNO3
R1
Oxidation
S2

NH3

Fig. 4.9: Flow diagram of nitric acid synthesis (first iteration).

The flow diagram in Fig. 4.9 consists of a mixer, where air and ammonia are
mixed, followed by an adiabatic reactor, and then a cooler to make the oxidation
reaction as complete as possible. This is important because reaction 4.4 has an un-
usual temperature dependence and is more active at low temperatures. This reac-
tion represents an example of tri-molecular reactions, whose temperature
dependence can be explained within the framework of the transition state theory,
taking into account temperature dependence of the partition functions for the tran-
sition state and reactants.
Several improvements should be made however in Fig. 4.9 as it suffers from ob-
vious deficiencies. For example, air will not flow to a mixer without a blower or
compressor, which should be added to the flow scheme. Fig. 4.9 shows that ammo-
nia enters the mixer as a liquid. Dosing of low-boiling liquids is difficult and thus
ammonia evaporation should also be included. As mentioned above, the oxidation
of ammonia is an exothermic reaction. The reaction temperature is 850–900 °C and
the adiabatic heat release is equivalent to 720 °C, thus the reactor inlet temperature
should be ~130–180 °C. This implies that a heat exchanger should be installed up-
stream to the catalytic reactor.
Furthermore, NO oxidation is also exothermic, and thus gases are heated dur-
ing this reaction. As a result, the absorption of NO2 is worsened. In order to circum-
vent this a cooler upstream absorber would be required.
In addition, the heat integration in Fig. 4.9 is far from optimized. At least high-
energy gases after the reactor could be used for steam generation, and thus a steam
reclaimer downstream to the reactor could be applied.
Finally, an optimum ammonia/oxygen ratio in the reactor should be 1:1.8, but
two volumes of O2 per one volume of NH3 are required for the acid production, and
therefore extra air could be added to the absorber. With these changes, the next it-
eration of the nitric acid flow scheme is illustrated in Fig. 4.10.
This scheme can be still improved by considering the following issues. The cat-
alyst is sensitive to impurities; however, no cleaning of air is used in Fig. 4.10. Air
400 4 Engineering technology

Air
Cooler
S17 S10
Mixer Water
S15 S5 Oxidation
S1
Reactor S11
S21
S3 1
S20 2
C1 S4 S7
M1 HNO3
S13 S18 E1 SS1
S16 R1 R2 S9
E2 S6 E3
S8 S19
Steam
Steam
S14 S12
NH3
Water
Evaporator

S2

Fig. 4.10: Flow diagram of nitric acid synthesis (second iteration).

filtration should be then installed in the third iteration scheme (Fig. 4.11). The heat
exchanger network in Fig. 4.10 is not rational and could be significantly improved,
for example the gases coming from the reactor could be used to heat the reactants
and a heat exchanger upstream to the reactor could be arranged in a similar way to
that in Fig. 4.4.

Filter S7
Air S22
U
A
S Cooler
S10 Absorber
Mixer Water
S15 S5 Oxidation
S1
Reactor
S3 S21 S11
S20 1
C1 S4 2
U1
M1
S13 E1 S18 SS1
S16 R1 R2 S2
E2 S6 E3 S19
S8
S9
Steam U
A
S17
S14 S S23
NH3 Water U2
F1 S12
Bleacher HNO3
Evaporator

S24

Fig. 4.11: Flow diagram of nitric acid synthesis (third iteration).

As water is produced in ammonia oxidation, low concentration nitric acid already

in the heat exchanger upstream absorber could be produced by water condensa-
tion. This in turn means that an extra stream could be introduced to the absorber.
Nitric acid will contain some NOx resulting in a low quality product with a
yellowish color, thus a bleach (stripper) should be introduced to treat unwanted
emissions of NOx through contacting the product acid with air.
An essential question to be addressed regarding the scheme in Fig. 4.11 is at
which pressure the nitric acid process should be operated, as absorption of NOx
 4.1 General structures of chemical processes 401

should preferably be performed at high pressure, while atmospheric pressure is

beneficial for oxidation. In fact, several options exist (Tab. 4.1), including carrying
out the whole process at atmospheric pressure or combining oxidation at low pres-
sure and absorption at high pressure. In the latter case a compressor should be
added between the ammonia conversion stage and the absorption stage.

Tab. 4.1: Modern nitric acid plants [6]*.

Atmospheric pressure Intermediate pressure High pressure

(AOP) (IOP) (POP)

NH concentration (vol%)   

Gauze temperature (°C) – – –

Burner Pressure (atm)   

Absorber pressure (atm)   

Number of gauzes – – –

Gauze diameter (m) – – –

Life of gauze (months) – – –

Pt loss (g/ton HNO) . . .

Acid strength (%) – – –

Conversion efficiency (%) – – –.

*Ammonia concentration below 12% is used to avoid explosive limits.

The technological scheme developed by Krupp Udhe GmbH (Fig. 4.12) is obviously
more complicated than Fig. 4.11, but still contains the same essential features as in
a simplified diagram.
Another example worth considering when discussing conceptual process de-
sign is the synthesis of methyl tert-butyl ether (MTBE). It can be achieved by react-
ing isobutene with methanol.
As a raw material, isobutylene from a C4 fraction could be used either from
steam cracking (20–30%) or fluid catalytic cracking (10–20%).
As the tertiary carbon atom in i-C4H8 is much more active than primary and sec-
ondary carbon atoms in other C4 hydrocarbons, only small quantities of other esters
will be obtained as by-products. The reaction in Fig. 4.13 is in equilibrium, thus in
order to obtain high product yields it should be shifted to the right side. The ether-
ification reaction in Fig. 4.13 is an exothermic one, therefore it is thermodynami-
cally favored at low temperatures.
One of the questions that arises when devising a flow scheme is in which phase
the reaction should be carried out, for example, the liquid- or the gas-phase. Non-
 402
Air condenser

CCW CCW
Ammonia air mixer
Tail gas
Tail gas heater 1
AC Reactor Tail gas heater 2

4 Engineering technology
NOx < 50 ppm
Secondary air
HP-
Air compressor steam STH
Tail gas- CW Tail gas
turbine
Process CW Process water
Air filter gas WB
cooler
Air Cooler
Tail gas NO gas condenser 2
CW
Ammonia tail gas mixer
WB
HP steam
NOx- Absorption
compressor Economizer
STL/CPL
Tail gas Boiler
feedwater Bleacher
reactor and dryer
Ammonia
gas filter
LP steam CW
CCW Tail gas
CPL heater 3 CW Nitric acid
Ammonia product
evaporation Ammonia gas
and
superheating CCW CW
LP steam Auxiliary Acid condensate
AW CPL ammonia CPLLP
Ammonia liquid evaporation Cooler condenser 1 and
feedwater preheater Steam acid heater

Fig. 4.12: Krupp Uhde dual pressure plant. (From [7].).

 4.1 General structures of chemical processes 403

CH3 CH3
CH3–C=CH2 + CH3OH CH3–C–O–CH3
CH3

Fig. 4.13: Synthesis of MTBE.

ideality of the mixture between methanol and hydrocarbons increases the activity
coefficient of methanol with respect to the vapor mixture. This leads to a much
higher equilibrium concentration of the product in the liquid-phase at the same T.
The process should thus be preferentially performed in the liquid-phase. This type
of etherification reactions can be catalyzed by sulfonic acid resins, which are suffi-
ciently active at rather low temperatures (40–50 °C). Requirements of low temperatures
inevitably mean that pressure is needed (around a few bars) to keep the reactants in
the liquid-phase. This results in a rather high equilibrium concentration (90%).
This high concentration (along with other parameters) determines the input-
output structure (Fig. 4.14). Two reactants (C4 and methanol) will result in the prod-
uct. Unreacted C4 hydrocarbons other than isobutene (remember that the feedstock
is coming from steam cracking or FCC and contains a mixture of hydrocarbons) and
unreacted isobutylene will be a part of the outlet stream.

C4 Unreacted C4
Reaction Separation
Methanol MTBE
Recycle methanol

Fig. 4.14: Input-output structure in MTBE production.

Recycling of this unreacted substrate is not worthwhile at such high conversion levels.
In fact this will be a recycle of normal butenes, leading to their build-up. Therefore no
recycling is done and unreacted C4 is used in other refinery processes. The equilibrium
could, however, be shifted by applying an over than stoichiometric ratio between
methanol and isobutene, and thus only methanol is recycled (Fig. 4.14). A high excess
of methanol is not usually used at it is slightly over stoichiometric.
Fig. 4.14 is somewhat simplified as it does not contain by-products, which are in
fact formed because of the dehydration of methanol to dimethylether CH3–O–CH3
and the dimerization of C4 to C8 hydrocarbons. The first reaction is a non-equilibrium
one and is favored at high T, high methanol concentration and low space velocity.
The product is a poison in downstream alkylation of C4 hydrocarbons. The second
reaction is also a non-equilibrium one, being favored at low methanol concentra-
tion and must be limited because C8 hydrocarbons leave with MTBE and decrease
its purity.
404 4 Engineering technology

In addition, the hydration of isobutene to tert-butyl alcohol can also occur,

while its equilibrium is limited by low water content. As mentioned above the reac-
tion of methanol with normal butenes is another side process which is not that im-
portant if MTBE is used for gasoline blending. Several options could be considered
for the reactor. A plug flow reactor should afford the highest reactant concentra-
tions. As the reaction is an exothermic one, being favored at low T, thus a low inlet
T is beneficial. In order to avoid unnecessary overheating a multitubular reactor
with external water cooling could be applied, which is obviously not a purely iso-
thermal reactor. A hot spot exists, as heat generated close to the inlet by a high re-
actant concentration cannot be totally transferred by water. With operation time
the catalysts gradually deactivates, which is associated with the fact that the inlet
streams contain impurities with deactivated catalyst. As purification is considered
to be not economical, gradual deactivation means that the hot spot also moves
gradually towards the reactor exit.
A Huels design of the MTBE process is illustrated in Fig. 4.15.

Raffinate

Reaction and n-C4

separation steps
for n-C4 production Methanol

MTBE
Column 1 Column 2
Reactor cascade MTBE purification n-C4 production

Fig. 4.15: Huels design of MTBE synthesis. ([8]). Reproduced with permission.

In the reactor cascade, i-butene is completely converted with methanol to MTBE.

The effluent from the reactors contains MTBE, linear butenes and methanol, which
was in the excess. As methanol and C4 form an azeotrope the bottom product in
column 1, operating at 6 bar, consists of ~97 wt% MTBE and 3 wt% methanol. In
column 2, MTBE with over 99 wt% is obtained. The MTBE/methanol azeotrope at
the top of column 2 is recycled into the reaction zone.

4.1.3 Conceptual process design: general comments

As illustrated in these two examples, quite a number of issues should be considered

in conceptual process design. In addition to the chemistry of the main reaction
 4.1 General structures of chemical processes 405

including its stoichiometry, stoichiometry of the parallel or series reactions should

be also determined. Temperature, pressure and concentration domains should be
carefully considered and a decision should be made on which phases are present in
the reaction system.
Analysis of kinetics and thermodynamics makes it possible to address composi-
tion at chemical equilibrium as well as to calculate how feasible the reaction rates
are, and which type of catalysts might be needed.
The application of a catalyst eventually raises a question of its stability, possi-
ble deactivation and whether it is necessary to perform regeneration.
The desired reactant and product purity at the available quality of feedstock
would also impose requirements on the feed purification and separation of products.
Obviously, processing constraints such as explosivity, corrosivity, toxicity and
overall safety should be carefully evaluated.
Thus, conceptual process design should answer a number of questions, some of
which are presented below:
– Is continuous or discontinuous processing to be preferred?
– What are the optimal regions of process conditions?
– Which process conditions are dangerous?
– How is the reaction T reached?
– Which type of reactor is preferred?
– Is pre-treatment of the reactor feed necessary?
– How is the reaction mixture processed?
– Are there any special measures in relation to the co-products and waste required?

For example, if conversion is not complete there is an option of recycling (Fig. 4.16)
that should be carefully considered, as recycling at rather high conversions is not
economical. Moreover, a simple recycling could lead to the build-up of impurities,
which might be present in the feedstock, thus the introduction of a purge stream is
necessary.

Main product
Feedstocks Process
By-products

Purge

Fig. 4.16: Input-output structure with purge.

Common sense rules should be used in the phase of conceptual design, such as
those for separations: avoid unnecessary separations; do not separate fuel and
waste stream products any further; do not separate and then remix.
406 4 Engineering technology

Basic engineering design based on the conceptual one will then answer the
question is the production process technically feasible in principle. More detailed
calculations should address the issue of economic attractiveness and the level of
the risk in economic and technological terms.
Technical risks are associated, for example, with a need to use technically non-
established equipment or equipment exceeding the conventional limits such as very
high distillation columns. A particular company might be not familiar with a certain
technology requiring for example the application of high pressure or fluidized bed
technology. Pharmaceutical companies might not be willing simply to switch to con-
tinuous processes from conventional batch ones. Use of units, which are difficult in
scaling-up, such as in general solid processing, might also be a risk. With solids even
apparently simple operations as bunkering and transportation can be difficult.

A couple of decades ago during a visit to an ammonia plant, a plant manager, who was previ-
ously involved for many years in production of fertilizers, told me that it was so nice to deal
with only gases and never in his life he would like to return to production of any solids.
Once upon a time, my father-in-law was working at a plant producing phthalic anhydride
from naphthalene. Later the feed was changed to o-xylene. Naphthalene is a solid with a melt-
ing point of approximately 80oC. Once a railroad car was delivered to the plant site and naph-
thalene had to be melted and pumped to the storage facilities during the winter time when the
outside temperature was approximately –30oC. Switching the feed to o-xylene (liquid at room
temperature) was considered a blessing.

There could be risks of operational failure of a single-train plant when there is a

strong coupling between steps when some streams are fed back to earlier stages.
Decoupling by inserting buffer tanks for intermediates can be an option for risk mit-
igation, although this approach is not practical for larger product quantities of
product or gaseous intermediates.
When a new technology is introduced, it can be anticipated that the operational
safety is lower. It is possible to diminish the risks by increasing expenditure on re-
search and development and developing failure scenarios.

4.1.4 Reactor selection

Reactor selection is an important issue in process design. An economic option is to

use fixed adiabatic beds when the temperature rise corresponds to the conversion.

When I was working in industry, one of the experienced ammonia plant managers mentioned to
me that for some reactions happening in ammonia production (to be discussed later), such as a
water-gas shift reaction, he did not even need to wait until the analysis of the product mixture, as
the temperature difference between the inlet and outlet is an excellent indication of conversion.
 4.1 General structures of chemical processes 407

As already mentioned in Section 3.6, for adiabatic fixed-bed reactors when too high
temperatures should be avoided, several beds are applied with interbed cooling ei-
ther using heat exchanges or quenching with cold reactants or an inert gas. The
synthesis of ammonia or hydrotreating of various streams in oil refining are typical
examples of this approach. In the case of ammonia synthesis, driven by equilib-
rium, quenching is a viable method to shift equilibrium. In other cases, the down-
side of quenching, resulting in an increase of the gas mixture volume, decrease of
the reactants partial pressure and lower reaction rates, can overweight the benefits
of interbed cooling. For various hydrogenation or hydrotreating reactions, quench-
ing is done by hydrogen.
For adiabatic reactions, a particular care should be taken of the first bed.
Higher concentrations of reactants imply higher rates, and thus larger temperature
gradients than for other beds. Subsequently, the height of the bed has to be de-
creased, which in turn negatively influences gas flow distribution in the radial di-
rection across the bed. An option would be to dilute the first bed with an inert or
introduce a less active catalyst.
For very strongly exothermic or endothermic reactions too many beds would be
required in order to control temperature rise, thus hundreds (for benzene hydro-
genation, or methane steam reforming) or thousands of catalysts tubes filled with
solid (ethylene oxide or phthalic anhydride synthesis) are arranged parallel to each
other, with cooling or heating in between the tubes (Fig. 4.17). Because of better
control temperature, such approach might either prevent excessive deactivation
and/or is needed to improve selectivity.

2 2

1 1

1 1

2 2

Fig. 4.17: Different arrangements of heat media circulation in between the tubes. 1, Heat medium;
2, reactants. (From [9]).
408 4 Engineering technology

Deactivation is in fact an important issue that often determines the type of reac-
tors used in industry. For example, for gas-solid catalytic reactions if deactivation is
not that profound (months to years) then packed beds of catalysts can be used.
Poisons present in the feed can be removed if necessary by installing guard
beds, which can be done either by use of a separate adsorbent, or by over sizing the
catalyst where this additional volume of catalyst is used to adsorb impurities.
Usually if the catalyst lifetime is sufficiently long (several years) then no regen-
eration is done and the catalyst is simply removed when it is considered uneconom-
ical to continue with it. For example, in the synthesis of ammonia the lifetime of
catalysts is usually 14–15 years. In fact, as impurities do not influence the catalyst
performance in this case and no carbon deposition occurs, the lifetime could be
even longer. However, ammonia synthesis requires the use of high pressures, there-
fore according to legislation the reactor vessels should be regularly inspected. For
this reason, after 14–15 years the catalyst charges are unloaded from the reactor.
Other catalysts in the same ammonia train, such as catalysts for natural gas pri-
mary and secondary reforming, high temperature and low-temperature shift, can
operate for several (2–6) years without any regeneration. An interesting example
in the same process is hydrodesulfurization, containing NiMo or Co-Mo catalysts.
The lifetime of such catalysts depends heavily on the presence of sulfur in the natu-
ral gas.

In 1999, I visited an ammonia production plants in Russia that utilized natural gas from Siberia,
and I asked about the situation with the hydrosulfurization (HDS) catalysts. It turned out that
the plant was put on stream in 1972 and since that time there were no changes in this catalyst,
which was obviously connected to low sulfur content in the feed. Similar plants, for example in
the Persian Gulf, change HDS after ~5 years.

The example of HDS demonstrates a case where it is not sufficient to just use a bed
with adsorbent or install a bigger volume of catalyst in order to remove impurities
in the feed (for example, mercaptanes in natural gas). In fact a separate reactor is
used upstream to the main one, where steam reforming of natural gas on supported
nickel catalysts is performed. Mercaptanes should be removed because they are poi-
sons for nickel.
In an HDS reactor the following reaction occurs:

RSH + H2 ! H2 S + RH (4:7)

Additional hydrocarbons, which are formed during this reaction undergo transfor-
mation to hydrogen along with methane that is exposed to steam over nickel cata-
lysts. H2S should be, however, removed upstream of steam reforming and this
is done by contacting it in a separate reactor with zinc oxide, which reacts (non-
catalytically) to zinc sulfide, which is then is discharged when all zinc oxide is
consumed.
 4.1 General structures of chemical processes 409

After several years in operation the activity of catalysts in adiabatic or isother-

mal fixed-bed reactors could decline. To compensate activity losses, an engineering
practice during industrial operation is to increase temperature (Fig. 4.18).

200

195

190
Temperature (°C)

185

180

175

170

165

160
0 10 20 30 40 50 60 70 80
Time-von-stream (months)

Fig. 4.18: Compensation of activity losses by temperature increase.

Although the policy of temperature increase can allow the constant level of produc-
tion output, the obvious drawback with this operation policy is that with a tempera-
ture increase side reactions are becoming more prominent, thus further deteriorating
the catalyst performance.
It is, however, possible that deactivation increases too strongly above a certain
T. For example, copper catalysts are very sensitive to sintering, which prevents a
policy of temperature increase in methanol synthesis over copper-containing cata-
lysts. In such cases the pressure could be gradually increased to compensate the
activity decline with operation time.
When deactivation is more profound because of coking and carbon deposition
it becomes necessary to regenerate the catalysts after several months. Fixed-bed
reactors could be applied and regeneration (by coke burning) is done while the
reactor is offline. As regeneration should be done very carefully in order to pre-
vent, for example, catalyst sintering, this regeneration process could be rather
time-consuming. Let us consider, as an example, a hydrogenation process where
the catalyst deactivates. As direct exposure of the catalyst to air (or oxygen) is
very dangerous and can lead to explosions, because the catalyst could still con-
tain some hydrogen, the reactor should be first purged with an inert gas, and
thereafter the coke from the catalyst should be carefully oxidized thus controlling
the amount of oxygen in the feed. If it is not properly done, heat released during a
highly exothermic coke oxidation can promote catalyst sintering and irreversible
410 4 Engineering technology

losses of catalyst activity. After burning of coke and subsequent purging with an
inert gas, the catalyst should be once again activated.
These lengthy procedures can significantly influence production output, thus
an additional reactor is usually installed in parallel. This allows the first reactor to
be isolated and regenerated as the feedstock is rerouted to the second reactor, al-
lowing the plant to operate continuously. After regeneration the first reactor re-
mains in a stand-by mode.
If the catalyst is active for several days or weeks, an option is to use a moving-
bed reactor with continuous catalyst regeneration.
Catalytic reforming should be mentioned in this context. This process helps to
transform heavy naphtha feedstock into high octane reformate for gasoline blend-
ing, as well as to generate high purity hydrogen for use in hydrotreating and hydro-
cracking (HC) applications.
In the CCR reforming process developed by UOP (Fig. 4.19), catalyst flows
through the reactors in series instead of remaining static in fixed beds of individual
reactors. Spent catalyst is continuously removed from the last reactor, transferred
to the regeneration section, where it is regenerated in a controlled way and trans-
ferred back to the first reactor. Such frequent regeneration allows CCR reforming to
operate in more severe conditions than for fixed-bed reforming.

Stacked
reactor
Naphtha feed Net H2-rich gas
CCR from treating Net gas
compressor Fuel
regenerator gas
Combined Recovery
feed ex- section
changer
Sepa-
rator Light ends
Stabilizer
Regenerated
catalyst
Fired heaters
Spent
catalyst
Aromatics
rich reformate

Fig. 4.19: CCR platforming from [10].

Deactivation could be even more severe, as will be discussed later on for fluid cata-
lytic cracking. A fluidized catalyst bed could be used when very small catalyst par-
ticles are fluidized in the flow of feed gas and there is continuous flow of the
deactivated catalyst to a regenerator and back to the reactor. It was already mentioned
in Chapter 3 that fluidized beds of catalyst allow operation at a uniform temperature.
A similar concept of separating the reactor and the regenerator can be applied
when a riser is used instead of a fluidized bed reactor.
 4.1 General structures of chemical processes 411

It became apparently clear from the discussion in Chapter 3 and also in this
chapter, that catalyst handling, including installation, activation (reduction, oxida-
tion, sulfidation) and regeneration, is extremely important as modern plants use
hundreds of tons of different catalysts, representing a significant cost.

I was once involved in a pre-scheduled supply of 100 tons of catalyst to a plant operator as the
result of the catalyst deactivation because of a maloperation. The catalyst was substantially
overheated by more than 200 °C, which resulted in an activity loss after just 1 year of operation,
while the expected lifetime was 5 years. The plant operator suffered significant financial losses
caused by an unscheduled purchase of catalyst, as well as a need to discontinue production for
a substantial time. Closing down such a large plant usually is very expensive.

Example of catalyst deactivation, time scales and the consequences for processes
are given in Tab. 4.2. As could be seen from the table, a range of reactions involving
organic compounds are prone to deactivation by coking, while the time scale of de-
activation depends on a particular reaction and can vary from seconds to days,
weeks and even years. Regeneration by coke combustion, as mentioned above,
could be made either continuously when the catalyst is recycled between a reactor
and regenerator or discontinuously.
Another option to prevent deactivation by the deposition of carbon is to operate
under conditions that will minimize sintering, for example, using excess of steam
in steam reforming of natural gas, which will be discussed below.
Other types of catalyst deactivation (discussed in Chapter 3), such as poisoning,
could be irreversible and thus catalyst regeneration is not an option. In those cases
the technological scheme should include a purification section, as exemplified
above for the hydrodesulfurization of natural gas prior to steam reforming.

One interesting historical example in this context is the nickel-based catalyst for synthesis of
so-called protective atmosphere for different metallurgical processes. In the mid-1960s, a truck
producer ZIL headquartered in Moscow with the annual capacity of approximately 200,000
trucks per year and approximately 100,000 employees working in approximately 20 locations
across the former Soviet Union (20,000 only in metallurgical plants) had to temporary shut
down production. The reason was severe deactivation of nickel catalysts, which were used for
generation of protective atmospheres containing hydrogen from reforming of natural gas with a
mixture of steam and CO2. For a long time, the factory operated with a relatively sulfur-free nat-
ural gas originating from Samara and the Soviet Central Asia. Due to expansion in the needs of
natural gas in rapidly growing city of Moscow, an alternative supply was found. The natural gas
from Orenburg had a much higher content of sulfur. Obviously, absence of any natural gas hy-
drodesulfurization resulted in severe deactivation of nickel catalysts.

Another issues that should be considered while selecting a reactor are injection and
dispersion strategies, which will be discussed below, following the approach intro-
duced by Krishna and Sie [12].
 412
Tab. 4.2: Deactivation in different industrial processes. (From [11]).

Process Catalyst Main deactivation Time scale of Consequences for Regeneration Consequences for
mechanism deactivation catalyst process

4 Engineering technology
FCC Zeolite Coke Seconds Regeneration on s Coke combustion Recirculation catalyst
scale between reactor and
regenerator

Oxidative Various Coke Seconds idem Similar schemes as

dehydrogenation oxides for FCC

Catalytic reforming Pt/γ-AlO Coke, Cl loss Months Days Alloying Coke combustion Cl Fixed-bed, swing
supply redispersion operation, moving-
bed

Hydrotreating Co/Mo/S/ Coke Metal sulfides Months Days Once-through Coke combustion Fixed-bed, slurry,
AlO catalyst Adapted moving-bed
porosity

Methanol Cu/ZnO/AlO Sintering (Cl) Years Stabilization Feed purification

Water-gas shift Cu/ZnO/AlO Poisoning (S, Cl) Years Stabilizers (ZnO) Feed purification

Three-way catalyst Pt, Pd Sintering, loss of Years Noble metals. Rejuvenation by

active components. Stabilized alumina leaching
deposits (Zn, P from (La, Ba)
lubricants)

Steam reforming Ni/AlO Coke, whiskers K. Mg Gasification Coke combustion Excess steam
catalysts

Dry reforming Ni Coke S-doping Coke combustion Excess steam

Diesel soot Cu-Cl Evaporation Minutes Hours Select other catalyst Add catalytic
additives to fuel (Ce)

DeNOx VO/AlO Formation surface Months Select other carrier

salts

Wacker oxidation Pd, Cu Catalyst deposit Low pH

Xylene oxidation Co, Mo, Br Mo, Co deposits Add new catalyst Deposits in reactor
and downstream

Styrene Iron oxide Coke, sintering Structural promoters Coke gasification in

movement promoters steam

4.1 General structures of chemical processes

413
414 4 Engineering technology

For example, reactants can be introduced in a one-shot mode as for batch reactors
or in a step function mode as for continuous reactors. Staged injection is an intermedi-
ate case and is applied in semi-batch reactors. Another option is to apply pulsed feed,
as for flow reversal types of reactors or semi-batch ones. Energy supply could be also
envisaged in many ways. In adiabatic reactors it can de done through quenching by
the cold reactant or by introducing intermediate heat exchangers. The application of
fluidized bed reactors can be an option for exothermal reactions, for example, selective
(or partial) oxidation of alkanes, i.e., n-butane to maleic anhydride (Fig. 4.20).

3.5 O2 [cat.] 4 H 2O
+ O +
400 °C

Fig. 4.20: Butane to maleic anhydride oxidation.

This reaction is the complex one, requiring extraction of 8H atoms and insertion of
three oxygen atoms to butane along with a ring closure. A high concentration of
butane is advantageous, but also dangerous because butane and oxygen could be
within explosion limits, thus the concentration of butane in air is ~2.5%. The appli-
cation of fluidized bed reactors allows the concentration of butane in the feedstock
to be increased by 50–100% compared to that in fixed-bed operations.
In this particular case of butane oxidation there is also a possibility to utilize a
transport reactor [13]. The advantage of such system, when a metal oxide oxidation
(in this case V2O5) is changing during the reaction (from V+5 to V+4), is that dona-
tion of oxygen from the catalyst lattice to the substrate with subsequent reduction
of V+5 to V+4 is separated from oxidation of V+4 to V+5. The latter process is con-
ducted in a separate reactor, which prevents butane from being in contact with air,
making the process thus much safer (Fig. 4.21). Operating temperature is similar to
the conventional processes (Tab. 4.3), while the pressure is somewhat higher. The
throughput can be increased by 10% with 20% higher yield [6]. The catalyst micro-
spheres of approximately 40–150 μm are made by spray drying of (VO)2P2O7 with
5% of polysilicic acid at pH 3 resulting in a hard porous shell of silica over the sur-
face of catalyst particles.
A number of oxidation reactions are conducted in a batch mode in slurry reac-
tors. For such reactors, several options for energy removal could be considered, in-
cluding evaporative cooling (Fig. 4.22).
 4.1 General structures of chemical processes 415

Maleic anhydride
Reduced catalyst

Off gas
(COx, H2O, etc.)

Main reaction
O
O2
O
V+5 O

Regenerator

Catalyst redox Inert gas

O2 O2 Air Riser
V+3 V+4 V+5
HC HC
Butane feed gas
Reoxidized catalyst

Fig. 4.21: Maleic anhydride synthesis in a transport reactor-regenerator tandem.

Tab. 4.3: Oxidation of butane to maleic anhydrate in fluidized beds [6].

Alusuisse DuPont

Design Conventional fluidized bed Circulating fluidized bed reactor with no

with separate feed and air oxygen supply and a separate regenerator
injection with air addition

Operating Ca.  Ca. 

temperature, oC
Butane  
concentration in the
reactor (vol.%)
Conversion Ca.  Ca. 

Other options for slurry reactors could be to have an external heat exchanger
(Fig. 4.23) or arrange heating through a double jacket or internal coils (Fig. 4.24).
The methods of energy input could be also different. In addition to those men-
tioned above, energy removal could be arranged through programmed temperature
cooling.
From the point of view of concentration profiles, it could be more attractive to
have a continuous plug flow reactor with no mixing of reactants. Conversely, such
types of arrangements for exothermal reactions lead to hot spots (Fig. 4.25), which
416 4 Engineering technology

Fig. 4.22: Slurry reactor with evaporative cooling.

External heat
exchanger

Pump Fig. 4.23: Slurry reactor with a heat exchanger.

not only determine conversion and selectivity, but also catalyst lifetime, reactor ma-
terials and safety of the whole process.
A specific method of energy removal could be to load catalysts with different
activity along the tube. In this way the amount of the active phase on a support can
be profiled along the tube length. This counterbalances the excessive temperature
increase and hot spots through deliberately minimizing the activity of the catalyst
layer close to the reactor inlet. Another possibility is to keep the amount of metal or
metal oxide on a support the same, but dilute the catalyst with the support at a dif-
ferent ratio changing along the reactor length (Fig. 4.26).
During reactor selection for a particular technology a decision should be also
made on whether in situ product removal will be beneficial for the process.
 4.1 General structures of chemical processes 417

2
3

4
(A) (B)

Fig. 4.24: Slurry reactors with (A) a double jacket and (B) a double jacket and internal coil. 1,
motor; 2, hood; 3, reactor walls; 4, discharge valve; 5, internal coil; 6, shaft with impeller.

275
235 °C
238 °C
239 °C
241 °C
T (°C)

250

0 1 2 3 4 5 6 7
l (m)

Fig. 4.25: Temperature profiles in a multitubular reactor along the length for different heating
media temperatures (ethylene oxidation to ethylene oxide, from [9]).
418 4 Engineering technology

Fig. 4.26: Catalyst profiling along the bed.

There are examples of reactions driven by equilibrium when in order to shift

equilibrium, the products should be withdrawn. A typical example is the produc-
tion of sulfuric acid by oxidation of sulfur dioxide to sulfur trioxide over vanadium
pentoxide catalysts.

SO2, O2, N2 98.5%

100 Bes 4
Bes 3
710 K 80
Bes 2
SO2 conversion (%)

865 K
Cooling
710 K 60
770 K
Cooling
705 K 40
Bes 1
Cooling 720 K
695 K 20
700 K

0
600 700 800 900 1000
(A) SO3, SO2, O2, N2 (B) Temperature (K)

Fig. 4.27: Adiabatic fixed-bed reactor for SO2 oxidation.

The reaction is limited by equilibrium (Fig. 4.27) and when performed in an adiabatic
fixed-bed reactor after the fourth bed the conversion is 98–99%, which still must be
improved to afford better SO2 utilization and diminish SO2 emissions, For a plant pro-
ducing 1 million tons per year the annual decrease of SO2 emissions could be as high
as 14,000 tons if the overall conversion is increased from 97.5 to 99.5%.
It should be also mentioned here that the most demanding job is done in the first
bed, and the temperature difference between the bed inlet and outlet could be as high
as 200 K with the catalyst bed of, for example, 0.5 m, while in the last bed the tempera-
ture difference could be just be 1 K. Note that the bed height of the last bed is usually
higher, than the first bed. The lifetime of catalysts in different beds thus varies dramati-
cally from 1 year for the top layer of the first bed to 10–15 years in the last one.
The intermediate removal of SO3 from the gas stream, usually after the third
bed and a quench, which can be done upstream of bed 4 (Fig. 4.28) or bed 5, allows
a shift of equilibrium leading to conversion above 99.5%.
 4.1 General structures of chemical processes 419

Boiler feed water Steam

a p

c m n o
b q

e e Tail gas
l
f h j
h k h h
Air g i i i
Product
acid
Process water

Fig. 4.28: Sulfur-burning double-absorption sulfuric acid process (Lurgi). (a) Steam drum; (b) Sulfur
furnace; (c) Waste heat boiler; (d) Main blower; (e) Mist eliminator; (f) Drying tower; (g) Air filter; (h)
Cooler; (i) Acid pump tank; (j) Intermediate absorber; (k) Final absorber; (l) Candle filters; (m) Steam
superheater; (n) Boiler; (o) Economizer; (p) Converter; (q) Intermediate heat exchanger (From [14]).

In Fig. 4.28 SO2 gas from the blower passes through heat exchangers before en-
tering the converter. After the first bed the partially converted gas is cooled in the
tube side of the hot heat exchanger and then flows to the second bed. The same is
done with the gases leaving the second bed. Hot gases leaving the third converter
bed and still containing unreacted SO2 are cooled and routed to the first absorbing
tower, where SO3 is efficiently combined with sulfuric acid. Unreacted gas is directed
to the fourth and subsequently to the final bed (if there are five beds). Thereafter the
gas leaving the converter passes the second absorption tower. This double-absorption
technology affords high overall SO2 conversion and lowers SO2 emissions.
In fact, process efficiency can even be improved if the catalyst with a cesium pro-
moter is employed instead of potassium. The main components of the conventional
catalyst include: SiO2 as a support, vanadium (V), potassium (K) and various other
additives. The reaction occurs within a molten salt consisting of potassium/cesium
sulfates and vanadium sulfates, coated on the solid silica support. Vanadium is pres-
ent as a complex sulfated salt mixture and not as vanadium pentoxide (V2O5).
Cesium-promoted catalysts were developed specifically for lower temperature
operations that can lead to greater SO2 conversions and hence lower SO2 emissions
to the atmosphere. The cesium salt (Cs) promoter reduces the required operating
temperature for the sulfuric acid catalyst by as much as 40 °C. The cesium/vana-
dium catalyst can be used in the first bed to reduce the bed inlet temperature
420 4 Engineering technology

(saving energy and start-up time). The Cs-catalyst can be also used in the final cata-
lyst bed (at a low inlet temperature) to maximize the SO2 conversion and reduce
emissions.

Once a plant manager told me that he wanted to have Cs promoted catalysts in the final bed,
because in addition to the interests of his company to save on payments for SO2 emissions, he
wanted to have a better environment for his daughter living nearby the plant. The same plant,
even if expressing an interest in applying Cs promoted catalyst in the first bed, in fact had prob-
lems with cooling gases to the required temperature after sulfur burner.

The example above was related to shifting equilibrium by removal of product. A

similar approach could be applied for other reactions, limited by equilibrium. As an
example, in an esterification reaction the removal of water (Fig. 4.29) through a
membrane can drive the reaction to completion.

Butanol Acetic acid Water Butyl acetate

H2 H2 O O
H2 H2
C C O
H3C C O H + H3C C H H + H C C C C O C CH
H2 O H 3 3
H2

O
H O H 2 C CH 3
H C C2 H H2 C O
3 2 H3C C
C C C C
H O O H H2
2 H H 3C Catalytic layer
H
O
H

Selective layer
O

H
H

O
H

H O Support layer
O

H H
H

Fig. 4.29: Esterification reaction with water removal. (From [15] copyright Elsevier.).

Another example of a reaction/separation combination using a membrane reactor

is worth mentioning even if it is related not with equilibrium, but rather with safety.
The synthesis of ethylene oxide from ethylene over silver catalysts (Fig. 4.30) has
some limitations associated with the fact that at certain concentrations ethylene
and oxygen could be within explosivity limits, thus an excess of ethylene is applied
in industry.
An interesting concept would be to utilize a catalytic membrane reactor (Fig. 4.31).
Reactive distillation along with a membrane reactor could be also applied to sepa-
rate products from the reaction mixture for equilibrium-limited reactions such as esteri-
fication mentioned above. Conversion can be increased far beyond the equilibrium
because of the continuous removal of reaction products from the reactive zone.
 4.1 General structures of chemical processes 421

C C
k1
O
C C
k2
CO, CO2
Fig. 4.30: Oxidation of ethylene.

Reactor wall

Catalytic membrane
Cooling tube

Ethene oxide,
Ethene
unconverted ethene

Oxygen

C=C EO
Catalytic layer
Porous layer
O2

Fig. 4.31: Catalytic membrane reactor for ethylene oxide synthesis.

Heterogeneous reactive distillations could be performed in distillation columns

illustrated in Fig. 4.32: the reaction section is the middle section containing a solid
catalyst, while above and below there are rectifying and stripping sections.
A clear advantage of combined separation and reaction is that a single piece of
equipment is used, making a considerable cost saving.
A similar strategy is applied in synthesis of ethylbenzene by alkylation of ben-
zene with ethylene (Fig. 4.33).
C2H5

AlCl3 or
+ CH2=CH2
Zeolite

Benzene Ethylene Ethylbenzene (4.8)

Benzene is fed to the top of the alkylation reactor, while ethylene is fed as a vapor
below the catalytic distillation section, making a counter-current flow of the
422 4 Engineering technology

Cooling water

Product

Feed

Catalyst

Steam

Bottoms

Fig. 4.32: Reactive distillation. (From [16]).

Catalytic Transalkylator Ethylbenzene PEB

distillation column column
column
Benzene

Inerts Ethylbenzene

Ethylene

PEB recycle Heavies

Fig. 4.33: Flow scheme of CDTECH EB technology. (From [17]).

alkylation reactants. In the catalytic distillation section vapor-liquid equilibrium

(VLE) is established, with ethylene being mainly in the gas-phase. The reaction
heat provides the necessary vaporization to influence distillation. The bottom of the
reactor/separator operates as conventional distillation columns.
 4.3 Fluid catalytic cracking 423

An important issue in the reactor selection is the injection strategy, for example
directions of flow, as discussed in detail for downflow and up-flow options in fixed-
bed multiphase reactors in Chapter 3.

Selection of an operation mode could be far from straightforward. Several years ago, at a con-
ference after a presentation about hydrotreating reactors, I posed a question about which oper-
ation mode is preferential. A very famous Dutch chemical engineering professor replied “Wise
men do downflow”. A manager from one of the leading European chemical companies answer-
ing the same question mentioned that all pilot plant reactors at that company operate in up-
flow mode.

4.2 (Petro)chemical industry

The chemical industry comprises the companies that produce 70,000 different
chemical products starting from basic raw materials, such as oil, natural gas, air,
water, metals, and minerals.
It is certainly beyond the scope of this book to describe various processes that
are currently applied in industry. There exists several excellent textbooks, handbooks
and encyclopedias that cover various aspects of chemical process technology in gen-
eral and processes where catalysts are applied.
Among the major processes that require application of heterogeneous catalysts,
fluid catalytic cracking (FCC), hydrocracking, stream reforming of natural gas and
synthesis of ammonia as well as a range of oxidation reactions including oxychlori-
nation were selected for more detailed description in this textbook.
FCC catalysts account for a significant part of all refinery catalysts, creating a
demand of approximately 650,000–750,000 Mt annually. Worldwide annual HC ca-
pacity was approximately 200 million tons some years ago. The synthesis gas cata-
lysts (including ammonia and methanol synthesis) market is also very large.
These examples and others, which will be discussed in detail below, are not
only related to important chemicals and fuels generated in these processes, but
also illustrate different types of reactors and catalysts, which were addressed in the
previous sections.

4.3 Fluid catalytic cracking

The thermal cracking process was patented in 1891 and substantially modified
later. Thermal cracking is still used to upgrade very heavy fractions or to produce
light fractions or distillates, burner fuel and/or petroleum coke.
Catalytic cracking as introduced by E. Houdry in 1928 is a flexible process to
reduce the molecular weight of hydrocarbons. The first full-scale commercial fixed-
424 4 Engineering technology

bed catalytic cracking unit began production in 1937. In the first plants cyclical
fixed-bed operations were used, which were replaced by a more efficient operation
with continuous catalyst regeneration in 1941.
The reasons for such modifications were associated with the fact that during
the catalytic cracking process the catalysts deactivated after a short time because of
coke deposition. Although coke can be removed and regenerated by burning, the
regeneration time is relatively long compared to the reaction time. Application of a
fixed-bed reactor with very frequent shut-downs for regeneration is not an economi-
cally viable option, while use of two reactors, one for hydrocarbon cracking and an-
other for catalyst regeneration along with moving of catalyst between these two, is
much more efficient. The first continuous circulating catalyst process used a bucket
elevator.
Although this technology of the moving-bed solved the problem of moving the
catalyst between efficient contact zones, the catalyst beads used still were too
large. Large catalyst particles result in large temperature gradients within catalyst
particles, which restrict regenerator temperatures and result in a large regenerator
and catalyst hold-up.
The next step in the development of catalytic cracking was introduction of the
fluid catalytic cracking. This process uses fine powdered catalysts that can be fluid-
ized. The first commercial circulating fluid bed process was put on stream in 1942
and by the 1970s FCC units replaced most of the fixed and moving-bed units.
Initial process implementations were based on synthetic low activity silica-alu-
mina catalysts and a reactor where the catalyst particles were suspended in a rising
flow of feed hydrocarbons in a fluidized bed.
A dramatic increase of catalyst activity and gasoline selectivity was achieved
with the introduction of zeolite cracking catalysts modified with rare earths in 1962
by Mobil. Thus, conversion and gasoline yield could be increased from 56% and
40% respectively for silica-alumina gels to 68–75% for conversion and 52–58% for
gasoline yield depending on the catalyst.
A significant development in the 1980s was the introduction of ZSM-5 zeolite
into the catalyst matrix as an octane enhancer.
Even today, FCC remains the dominant conversion process in petroleum refin-
eries, producing a high yield of gasoline and liquefied petroleum gas (LPG).
Approximately 1,500 tons per day were consumed worldwide in 722 refineries ac-
cording to Bartolomew and Farrauto [11].

4.3.1 Feedstock

In the 1950s kerosene gasoil fractions were mainly applied as the feedstock, while
later vacuum gas oil and even heavier feedstocks started to be used. The boiling range
 4.3 Fluid catalytic cracking 425

of the feedstock could be 540–550 °C with a molecular mass 1.5 times larger than for a
lighter feedstock. It also means that vaporization of the feedstock is not complete.
A switch to heavier feedstock also required efforts aimed at a decrease of the
coke yield, which for vacuum gas oil can reach 5–6%. The trend of using heavier
metal-containing feedstock, such as atmospheric residues and synthetic crude, is
continuing and will only accelerate. The atmospheric residue can contain Ni and V
in the amounts of up to 10 ppm and even higher, while the content of these metals
is below 1 ppm in vacuum gas oil. Undesirable effects of nickel and vanadium are
associated with increased gas and coke selectivity for Ni and loss of activity for V.
In addition, this heavier feedstock can contain sulfur. One of the options of han-
dling it on a catalyst level is to incorporate an additive into the catalyst, which will
capture SOx formed during regeneration. Other possibilities might include a hydro-
treating unit upstream FCC.
Typical FCC unit feeds and operating conditions are presented in Tab. 4.4.

Tab. 4.4: Typical FCC unit feeds and operating conditions.

Operation

Feedstock Vacuum gas oil Atmospheric residue

API gravity . .

Sulfur (wt%) . .
Nickel (ppm) . 
Vanadium (ppm) . .
Conradson carbon (wt%) . .
Conversion (vol%)  
Fuel gas (wt%) . .
Total C (vol%) . .
Total C (vol%) . .
C + gasoline (vol%) . .
LCO (vol%) . .
Slurry (vol%) . .
Coke (wt%) . .
Reactor temperature (°C)  
Regenerator temperature (°C)  

(From [6].)

The amount and the quality of FCC products depends on the feedstock characteris-
tics as well as the process parameters leading to fuel gas, gasoline and light cycle
oil (LCO). LCO, when used as a blending component in heating oil, can have a
greater value than that of gasoline. An example of the product distribution is given
in Tab. 4.5. Under such circumstances, many refineries adjust their FCC unit opera-
tion to increase LCO yield at the expense of gasoline.
426 4 Engineering technology

Tab. 4.5: Product distribution of FCC unit [18].

Product Yield, vol% Yield, wt%

Light gases (C−) .

Propane . .
Propylene . .
Isobutane . .
n-butane . .
Butenes . .
Gasoline (C+) . .
Light cycle oil (LCO) . .
Heavy cycle oil (HCO) . .
Coke .
Total . 
Cycle oils . .
Total LPG . .

Capacity 25,000 bbl/d or 65.61 m3/h. Conversion of gas oil = 72 vol%.

LCO compared to diesel fractions has lower cetane number and higher sulfur con-
tent. The cetane number depends on the feedstock and operation temperature.
Obviously an increase of LCO yield can be achieved by diminishing FCC unit crack-
ing severity, eventually leading to higher yields of heavy products (light cycle oil,
heavy cycle oil, and clarified oil) and lower yields of light products (gasoline, LPG,
and gas) as well as coke. Reducing catalyst activity lowering reactor temperature,
and reducing the catalyst/oil ratio can lead to increase LCO yield.

4.3.2 Reactions/mechanism

Many reactions happen during catalytic cracking, such as isomerization, cracking, pro-
ton transfer, alkylation, polymerization, cyclization, dehydrogenation, condensation,
coking, etc. Some of them are primary reactions, while the majority are secondary.
A somewhat simplified overview of the reactions is given in Fig. 4.34.
During the cracking of normal paraffins the cracking reactions are the dominant
ones. The products contain mainly paraffins of lower molecular weight and olefins:

H2 H2 H2 H2 H
C C C C CH3 + C
H3C C C CH3 H3C C H3C CH2
H2 H2 H2

The yield of the latter increases with the increase of the feedstock molecular mass.
Heavier fractions are less stable and can be cracked more easily than light fractions.
Isoparaffins crack easier than paraffins.
 4.3 Fluid catalytic cracking 427

Long-chain alkanes

Smaller alkanes Branched alkenes and

and cycloalkanes branched alkanes

Aromatics
Smaller alkenes and
branched alkenes

Fig. 4.34: Diagrammatic example of the catalytic cracking of petroleum hydrocarbons. (From [19]).

Naphthenes are cracked at the side chain. Alkylnapthenes or aromatics with

shorted side chains are formed first as they are stable, especially with methyl and
ethyl substituent. Cracking of olefins, which is a secondary reaction, leads to
smaller olefins:

Initiation of the cracking process of olefins occurs (as for many other reactions of ole-
fins) as a result of formation of carbocations, catalyzed by Brønsted and Lewis acids. A
carbenium ion is a positively charged tricoordinated carbon atom, while a carbonium
ion is a positively charged pentacordinated carbon atom. If a carbenium ion is large
enough (C6+) it can crack, forming alkene and another carbenium ion, which isomer-
izes if possible into a secondary or tertiary ion. When the carbenium ion is not large
(C3–C5), it will be terminated into an olefin through proton abstraction to the catalyst
or another olefin, or through termination to a paraffin via addition of a hydride.
The mechanistic scheme for primary catalytic cracking is presented in Fig. 4.35.
Branched polycyclic aromatics are often dealkylated
428 4 Engineering technology

H3C CH2 CH2 CH2 CH2 CH2 CH3 n-Alkane

Initation
+
H3C CH CH2 CH2 CH2 CH2 CH3 Classical carbenium ion

H3C CH CH CH2 CH2 CH3 Protonated cyclopropane

H+ CH2 Hydride shifts + Isomerization
C–C bond breaking
+ +
H3C C CH3 + H2C CH CH3 H3C CH CH CH2 CH2 CH3
CH3 n-Alkene CH3
etc.
Hydride transfer

H3C CH CH3
CH3
iso-Alkane

Fig. 4.35: Mechanistic scheme for catalytic cracking [20].

or also can be cracked in the side chain:

H2 H2 H2
C C C
C CH3 CH3
H2 + CH2=CH2

H
C
CH2
+ CH3– CH3

Carbon-carbon bond rapture occurs at the ring and is not very profound when an
alkyl chain contains less than three carbon atoms. Polycyclic naphthenes are trans-
formed to monocyclic naphthenes and alkenes. Naphthenes can react with alkenes
through hydrogen transfer, leading to aromatics and alkanes:

+ C=C–C + C–C–C

A large amount of aromatic compound is obviously also produced from paraffins,

whose structures allow cyclization.
Straight-chain alkenes can undergo acid-catalyzed skeletal isomerization reac-
tions, leading to branched alkenes:
 4.3 Fluid catalytic cracking 429

CH3
H2 H
C C C CH3
H3C C CH3 H3C C
H H

This is an important reaction because it increased the octane rating of gasoline

range hydrocarbons. In addition to skeletal isomerization, double-bond migration
can also occur. In addition, dehydrogenation can take also place, which together
with cyclization gives polyaromatics, which are adsorbed on the catalyst surface,
causing catalyst deactivation through coke formation.
The secondary reactions also produce unwanted light gases in addition to the
formation of coke, which could be very substantial (as shown in Tab. 4.4). In a con-
ventional operation around 2–5% coke is typically deposited on the catalyst.
Among secondary reactions, ring opening of cycloalkanes could be highlighted,
as well as alkyl groups transfer:

H3C CH3 CH3 CH3

+ +

and dehydrocyclization:

H
C H2 H H2
CH2 C C C CH3
+ H3C C C C H2 + H2
H H2 H2 C CH3
C C
H2 H2
CH2
H3C

An overview of cracking reactions and the products is provided in Tab. 4.6.

The handbook by Lloyd [6] gives a very thorough description of FCC as well as of a number of
other important industrial reactions and it is strongly recommended to those interested to learn
more about these processes.

Dehydrogenation reactions does not usually play a significant role during the crack-
ing of high molecular-mass paraffins, while for low molecular weight paraffins
these reactions could be important as they result in the formation of valuable com-
ponents – olefins and from low value ones – paraffins.
Coke, which usually has an atomic ratio between H and C from 0.3 to 1 and
spectral characteristics typical for polycyclic aromatic compounds, can be classified
into several groups:
430 4 Engineering technology

Tab. 4.6: Catalytic cracking reactions.

Hydrocarbon Initial products Further products

Paraffins Branched paraffins and olefins mainly Olefins crack and isomerizes and are also
in the C–C range. saturated by hydrogen transfer to give
paraffins. Olefins also cyclize to
naphthenes.

Naphthenes Crack to olefins. De-hydrogenate to Further dehydrogenation to aromatics, by

cyclic olefins. Isomerize to smaller hydrogen transfer.
rings.

Aromatics Alkyl groups crack at the ring to form Further dehydrogenation and
olefins. condensation forms coke.

Dehydrogenation and condensation to

polyaromatics.

Typical Light gas (%) H, CH, CH, CH, CH, CH
products LPG (%) CH, CH
(approximate) Naphtha (%) Light,  °C– °C,
LCO (%) Heavy  °C– °C
HCO (%) Jet fuel  °C– °C
Coke (%) Kerosene, diesel, heating oil Recycle.
Higher than  °C

(From [6].)

– catalytic coke formed on acid sites

– dehydrogenated coke, which results from dehydrogenation reactions over met-
als coming with the feedstock and deposited on the catalyst
– chemisorbed coke, which is a result of irreversible chemisorption of high boil-
ing point polycyclic arenes
– reversible coke, resulting from incomplete desorption after exposure to steam

Catalytic coke is directly related to cyclization of olefins, condensation, alkylation

and hydrogen transfer. The highest yield of dehydrogenated coke occurs over such
metals as Co, Ni, Cu and to a lesser extent V, Mo, Cr and Fe.
One of the criteria for determining the significance of the secondary reactions is
the ratio between gasoline and coke. If the ratio is high it shows the dominance of
the desired reactions, provided that the octane number of gasoline is high enough.
Isomerization, hydrogenation, cyclization and mild aromatization of olefins could
be considered as desired reactions. They lead to high yields of paraffins, including
branched ones, as well as aromatics with a boiling range close to that of gasoline
and to a favorable ratio between iso- and normal paraffins. Excessive cracking, de-
hydrogenation and polymerization of olefins, as well as condensation of aromatic
 4.3 Fluid catalytic cracking 431

hydrocarbons results in high yields of hydrogen and coke, low yields of olefins, giv-
ing a heavier product with lower yields of gasoline and a lower octane number. The
amount of coke is dependent, both on properties of the catalyst and the feedstock
but also on the kinetic parameters of the process, which will be discussed below.

4.3.3 Kinetics/process variables

Because of the high complexity of cracking reactions and hundreds of them occur-
ring at the same time, it is not realistic to develop a detailed model for cracking,
therefore for modeling purposes empirical models are usually applied. Simplified
kinetic models are applied when several compounds are lumped together. For ex-
ample, a simple three lump model given in Fig. 4.36 consists of a) one feedstock
lump (gas oil, VGO or any other heavy feed) and two product lumps: b) gasoline
and c) coke + light gases. The gasoline lump contains the fractions between C5 up
to the hydrocarbons with a 220 °C boiling temperature. The coke + light gases lump
contains (in addition to coke) C4 and lighter than C4 hydrocarbons.

Gasoline
k1

k3
Gas oil

k2 Light gases and coke

Fig. 4.36: The three lump model.

As example of this reaction mechanism is gasoline formation from gas oil and it is
expressed in the following way:

rgo to gasoline = k1 ’1 Cgo n (4:9)

where φ1 is the deactivation function and n is the reaction order in gas oil.
Although the first-order reactions are mainly considered in the literature, some-
times second order is also assumed. In order to explain the effect of coking on the
catalyst decay two main approaches have been used. The first one relates deacti-
vation with the catalyst coke content, while the second approach links catalyst
deactivation to time-on-stream.
Another approach to kinetic modeling is to lump different products based on
their chemical structure, for example, separately lump paraffins, olefins, naph-
thenes, and aromatics. This approach also gives an option to include the reaction
type (cracking, hydrogen transfer, isomerization, etc.) and stoichiometry.
432 4 Engineering technology

Although a detailed kinetic analysis is challenging, it is however worth consid-

ering various kinetic parameters and their influence on the catalytic cracking.
Catalytic cracking is conducted in the absence of any heat supply and is thus con-
sidered adiabatic. The main process parameters are temperature, pressure, the ratio
between the feedstock and the catalyst, as well as the circulation ratio.
In catalytic cracking the catalyst is circulated continuously between the two reac-
tors and acts as a vehicle to transport heat from the regenerator to reactor. The spent
catalysts, which are partially covered with oil, are sent to a steam stripping unit to
remove the adsorbed oil before entering the regenerator. The stripped oil vapors are
sent to a fractionation tower for separation into streams with desired boiling range.
Catalytic cracking is usually conducted at 450–550 °C decreasing from bottom to
top in a riser as the cracking reactions are endothermic. Typical temperature profiles
for catalysts with different amounts of rare earth metals are illustrated in Fig. 4.37.

+40 °C Mix. temp.

Low-rare-earth-exchanged zeolite
+20 °C High-rare-earth-exchanged
zeolite

Riser terminal temperature

Feed
injection

Fig. 4.37: Typical riser reactor temperature profile [21].

After stripping, the catalyst, which still contains typically 0.8–1.3 wt% coke, is fed
to the regenerator where the coke is burnt off. The temperature in the regenerator is
540–680 °C being carefully controlled as high temperatures can deactivate cata-
lysts. The residence time is in the order of minutes. Pressure is 1.6–2.4 bar in the
reactor and 1.3–3.1 bar in the regenerator. Because steam generated by combustion
has an undesired effect of dealuminating the framework of the zeolite and degrad-
ing zeolite crystallinity, the fresh catalyst is fed to the unit continuously to partially
replace a spent catalyst. Thus, in FCC the catalyst (coined as “equilibrium catalyst”)
is a physical mixture of the fresh catalyst and regenerated catalyst with a broad age
distribution circulating within the FCC column.
Temperature in the riser has a following influence on the reactor performance.
With temperature increasing to 470–480 °C the gasoline yield is increasing, passing
through a maximum at 490 °C as secondary cracking of formed hydrocarbons is be-
coming more prominent and the yield of gaseous products as well as coke is in-
creasing. Moreover, with a temperature increase the octane number is increasing,
along with the C1 to C3 in the gases, while C4 is decreasing. With the pressure in-
crease there is a decrease in the gasoline and C1–C3 yields, as well as the total
 4.3 Fluid catalytic cracking 433

amount of olefins and aromatics. The coke yield is almost independent of pressure.
More catalysts are introduced to the riser than the feed: the ratio of the catalyst to
oil is 5–30, typically approximately 10. Increase of this ratio increases conversion
and diminishes the coke content on the catalyst surface, while the yield of the de-
sired products (gasoline) is increasing.
Another important parameter is the flow rate (16–20 h−1 in a riser) or the con-
tact time (few seconds). The process usually operates at 75–80% conversion, which
also determines the contact time. Longer contact times will result in secondary
cracking and undesired low molecular weight gases and coke.
In summary, the yield of the main products can be influenced by the following
parameters. Dry gas (an undesirable product) can be diminished by decrease in
temperature and residence time; metal content and aromatic content of the feed.
The yield of the desired product is increased by: increasing the catalyst to oil
ratio; operation at the maximum possible temperature (without over cracking)
and increasing the catalyst activity. Production of LCO can be maximized by de-
creasing reaction temperature, decreasing catalyst to oil ratio and a decrease in
catalyst activity.

4.3.4 Catalysts

Cracking catalysts are solid acids, therefore amorphous silica-aluminas were ini-
tially used for this process. Acidic strength is usually higher for crystalline zeolites
compared to amorphous ones, also displaying higher activity, which is essential for
the yield of the desired product. Because of such features, compact riser reactors
with zeolites replaced rather large fluidized bed reactors.
The highest acid strength was found in zeolites containing the lowest concen-
trations of AlO4 tetrahedra, for example, ZSM-5 and Y. Some early zeolite catalysts
contained X-zeolite, probably because it was cheaper to manufacture, however,
later on it was demonstrated that Y zeolite was more stable under typical operating
conditions. One of the reasons could be that zeolite X is hydrophilic (contrary to Y
zeolite, which is hydrophobic).
It should be also noted that if a catalyst is too active this might lead to higher
yields of gas and coke, and during regeneration substantial amounts of heat would
be released leading to the collapse of the zeolite framework. Undiluted zeolites are
too active for use in existing units and would be immediately deactivated as coke is
deposited on the surface.
Thermal stability increases with increasing Si/Al ratio, thus ultrastable zeolite
Y (prepared by ion exchange with rare earth ions, La3+ and Ce3+ or by dealumina-
tion through treatment with steam), found widespread application in fluid catalytic
cracking. Note that the regeneration of zeolites includes treatment with steam, thus
hydrothermal stability is essential.
434 4 Engineering technology

In addition to the modification of acidity and stability as described above in

order to achieve better heat removal (~10%) the catalyst also contains a matrix in ad-
dition to zeolite. Initially the matrix was simply a diluent and binder to form porous
particles strong enough to resist attrition while circulating continuously between the
reactor and the regenerator. As this matrix is also used to transport reactants, large
pore materials are applied. Nowadays, aluminosilicate is applied as an amorphous
matrix, also providing the required mechanical stability. Typical commercial cracking
catalysts are thus a mixture of Y zeolite and SiO2–Al2O3. Alumina solutions made by
dissolving pseudoboehmite in a monobasic acid, such as formic acid, had previously
been used to bind silica/alumina catalysts. Other binders now include silica sols and
aluminum chlorhydroxide. A fresh FCC catalyst consists of spray-dried spherical, flu-
idizable, attrition-resistant particles, typically containing 20–40% ultrastable Y (USY)
zeolite, a binder, a catalytically active acidic matrix and various additives.
The use of cracking catalysts containing zeolite increased rapidly following
their introduction in 1962, and by 1972 they were being used in at least 90% of the
FCC units in the United States. The most stable and successful catalysts for produc-
ing high yields of gasoline were made from rare earth zeolites. Various manufac-
tures prepared FCC catalysts with different amounts of rare earths depending on
operational severity and product objectives.
The flow scheme of the rare earth zeolite Y preparation is given in Fig. 2.121 and
the synthesis of microspherical catalysts is presented in Fig. 2.122. Spatial distribu-
tion of the zeolite Y in the matrix is addressed in Section 2.2.15.6.
The most important achievement of the application of zeolites compared to alu-
minosilicates was not the activity increase, but rather an increase of selectivity to-
wards some products. For example, the amount of olefins is much less over
zeolites, which is associated with a higher rate of hydrogen transfer on zeolites
than on aluminosilicates, and stronger adsorption of olefins compared to paraffins.
Because olefins have a longer residence time on the active sites, they are able to be
saturated with hydrogen, which is generated because of naphthene dehydrogena-
tion to aromatics. Yields of gasoline are higher on zeolites, while the yields of coke
and gas (<C4) are not less than on aluminosilicates.

In my first discussion with the leader of the research group in a company where I used to work
in the 1990s, it was mentioned that selectivity improvements by just 1% are extremely impor-
tant for a chemical company dealing with large production volumes, and that a substantial
amount of work in the company was just about that.
A similar quote from Robert S. Langer’s Priestley Medal address (2012), when he was de-
scribing his experience in 1970s, is: “One job interview made quite an impression on me. I went
to this interview at Exxon in Baton Rouge, La., and one of the engineers there said to me that if I
could increase the yield of a particular petrochemical by about 0.1%, wouldn’t that be wonder-
ful? He said that would be worth billions of dollars”.
In very many cases selectivity issues are the main focus of industrial research and develop-
ment work, although activity and stability to deactivation should not be undermined.
 4.3 Fluid catalytic cracking 435

During catalyst preparation, the initial exchange with rare earth and ammonium chlo-
ride removes sodium ions from the supercages. When calcined the rare earth oxides
and hydroxides decompose and migrate into the sodalite cages, where they can ex-
change for more sodium ions. In the calcination process, some of the rare earth ions are
converted to cationic polynuclear hydroxy complexes that provide additional acid sites.
Because the demand for gasoline grew with time, the level of rare earths in the
catalyst formulation refiners tended to increase, being on average ~3%. A recent
spike in the prices of rare earth metals has put an additional pressure on catalyst
manufacturers as well as refiners to diminish the role of rare earths for FCC.
Although better conversion and an increased yield of gasoline could be achieved
with rare earths Y zeolite catalysts, it was found that octane levels were lower be-
cause fewer olefins were produced. As gasoline quality could be improved by either
increasing cracking severity or using antiknock lead that contained compounds such
as tetraethyl lead, this lower octane level was not a problem before new constraints
were enforced.
In the mid 1970s, the application of small pore ZSM-5 (0.55 × 0.5 nm channel)
as a co-catalyst with rare earths Y zeolite increased the octane number of gasoline.
Straight-chain C6–C10 olefins produced by normal cracking were further processed
in a shape-selective cracking over ZSM-5, resulting in an increase in the gasoline
octane number with the same dry gas, heavy oil, or coke production at the expense
of decreasing gasoline yield by up to 2%. This loss could be compensated because
propylene and n-butylene are used in an alkylation unit. ZSM-5 in an inert matrix is
added along with Y zeolite on a daily basis to FCC units. ZSM-5 framework is ex-
tremely stable, affording less deactivation than typical Y zeolites.
Low hydrothermal stability of ZSM-5 leads to relatively fast dealumination and
the relative activity loss if no special measures are taken. Modification with phos-
phorus is used to limit this relative activity loss.
The feedstock contains small amounts of metals even after hydrotreating (in par-
ticular Ni and V), which deactivate zeolites, acting as dehydrogenation catalysts and
giving more coke, thus passivators as Sb, P, Sn and B compounds are added because
they react with the impurities. Nickel, which is a concern when its concentration on
the equilibrium catalysts exceeds 800 ppm, may be passivated by the addition of low
levels of antimony or bismuth to the feed. A negative side of this method is an in-
crease of NOx emissions and potential bottoms fouling.
A more permanent solution is to use a material within the catalyst that is able
to deposit Ni. In the current residue FCC catalysts a special alumina is integrated in
the catalyst throughout the catalyst microsphere in order to trap nickel forming
nickel aluminate. More efficient use of alumina could be achieved if it is located
only at the outer layer of the catalyst. Such catalysts were recently introduced into
the market and the production technology is presented in Fig. 4.38.
Further development of more stable catalysts was related to introduction of
boron, which migrates within the catalyst by solid-state diffusion to passivate Ni.
436 4 Engineering technology

Step 1: Inner stage microsphere (MS) Step 2: Outer stage microsphere

Clay slurry Clay slurry

H2O + alumina H2O
Dryer Dryer
+ inner
stage MS
Alumina free
Calciner inner stage MS Calciner Two-stage MS
(~65 μm) (~80 μm)

Step 3: Zeolite crystallization and after-treatment

Microsphere
caustic
silicate Base Rare-earth Multistage
Calciner
exchange exchange reaction catalyst
Na-Y

Fig. 4.38: Steps in preparation of multi-stage reaction catalyst. (From [22]).

This innovation helps to increase the liquid product yields with heavy residue feeds
by better metal passivation and lower hydrogen and coke production (Fig. 4.39).

Operating gasoline vs. conversion

54
Cat A BoroCat
Operating gasoline (wt.%)

53

52

51
70 72 74 76
Unit conversion (wt.%)

Fig. 4.39: Comparison of boron containing catalyst with a reference commercial catalyst (From [23]).

A decrease of the amount of carbon on the regenerated catalyst and an improvement

in burning CO in the regenerator can be achieved by adding a CO combustion pro-
moter, such as Pt, Pd, Ir, Ru, etc, in small amounts (0.01 to 50 ppm) to the catalyst of
the regenerator.
 4.3 Fluid catalytic cracking 437

Another key property of FCC catalysts is attrition resistance, as it determines

the catalyst addition rate (typically less than 1% per day) and the catalyst fluidity.
This property increases with increasing bulk density and decreasing pore volume.
Conversely, catalyst activity and resistance to deactivation by the plugging of pores
increases with a decrease in bulk density and increased pore volume. The applica-
tion of microspherical catalysts (Fig. 4.40) in aluminosilicate amorphous matrix
(10–150 μm) improved abrasion resistance and diminished the formation of fines.

Fig. 4.40: Microspherical catalyst for FCC.

4.3.5 Technology

In its simplest form, the FCC unit (Fig. 4.41) consists of three sections: reactor sec-
tion, fractionation and gas concentration (Fig. 4.42).

Regenerator Reactor Main column

Overhead vapors Fuel gas to treating

To and unstablized
flue gas system gasoline Gas
C3–C4 to treating,
concentration
unit alkylation or polymerization

Debutanized gasoline to treating

Heavy naphtha product
Light cycle oil product
Heavy cycle oil product

Main column
Air Fresh feed bottoms
product

Fig. 4.41: Flow scheme of a FCC unit [24].

438 4 Engineering technology

Products

Flue gas

ca nt
Regenerator

e
t
Sp
Stripper
Temperature: Temperature:
650–760 °C 493–554 °C

ECAT Riser
FRESH
Feed (VGO or residue)
Re
Air blower g
Furnace ca en
t

Fig. 4.42: Fluid catalytic cracking with reactor and regenerator (modified from [25]).

In the riser reactor design introduced by UOP in 1971 the pre-heated raw oil
feed, together with hot catalyst from the regenerator, enters at the lower part of the
reactor (Fig. 4.42). Vaporization of oil (vaporizing temperature depends on the feed,
usually between 350 °C and 450 °C) occurs because of heat provided by the catalyst.
In fact, the catalytic cracking unit (Fig. 4.43) also has a furnace (Fig. 4.43b), which
is required during the start-up or insufficient heat supply from the regenerator.

Reactor

Regenerator

Fractionation
column

Fig. 4.43: Photo of (a) the reactor section with regenerator, (b) furnace for heating the feed
and (c) cyclones.

The cracked hydrocarbons, separated from the catalyst in cyclones (Fig. 4.43c),
leave the reactor overhead and go to the fractionation column for separation into
fuel gas, LPG, gasoline, naphtha, LCOs used in diesel and jet fuel, and heavy fuel
oil. A large part of cracking occurs within a few seconds (2–5 s). Minimum gas
 4.3 Fluid catalytic cracking 439

velocity should be at least 10 times higher than the minimum fluidization velocity
of the catalyst particles. Because of the increase in volume inside the reactor, the
gas velocity increases axially and the highest gas velocity at the outlet of the riser
can be as high as 25 m s–1.
The spent catalyst first falls down into the stripping section within the reactor
vessel. Steam (3–5 kg of steam per 1,000 kg of the circulating catalyst) removes
most of the hydrocarbon vapor and the catalyst then flows down a standpipe to the
regenerator. If these hydrocarbons are not removed, temperature in the regenerator
will be increased beyond that required for smooth operation limits. The residence
time of the catalysts in the stripper is 0.5–1 min.
The spent catalyst mixes with air and clean catalyst at the base of the regenera-
tor, where the coke deposited during cracking is burned off, thus regenerating the
catalyst and additionally providing heat for the endothermic cracking reactions.
The temperature in the regenerator can rise from 500 °C to 650 °C because of the
exothermicity of the coke burning. Gases leaving the regenerator contain substan-
tial amounts of CO and have significant heating value, which is used to generate
steam, thus the flue gas is treated in a CO boiler where CO is transformed to CO2.
Catalysts are partially deactivated in each cycle. Moreover, because of catalyst
transport in the reactor and the regenerator and catalyst attrition, losses of catalyst
occur. To compensate for this, make-up catalyst is added to the process continuously,
with a replacement rate of about 1–3% of the total inventory every day. Catalyst attri-
tion in the unit is directly related to the catalyst circulation rate. Increasing superfi-
cial gas velocity (with increase of feed rate, reactor temperature, quench steam or
decrease of operation pressure) increases catalyst entrainment to the cyclones and
results in increased catalyst losses.
More than 2,000 tons of catalysts are replaced every year in a FCC unit with an
inventory of 200 tons of catalyst and a replacement rate of 3% a day.
Because of the constant addition of catalysts FCC units operate continuously
and are very seldom closed down to replace the catalyst inventory completely. Such
shutdowns would obviously lead to significant decreases in production capacity.
Several types of FCC units are currently in operation, using various designs, not
only the one presented in Fig. 4.42. For example, the R2R process (Fig. 4.44) was
initially developed by Total to process feedstock with a high residue content and is
now licensed by Axens/IFP and Stone and Webster. In addition to a riser, a stripper,
a disengager and two standpipes; in this process there is a regeneration system
composed of two regenerators linked by a lift.
The first regenerator acts as a mild pre-combustion zone at a temperature not
higher than 700 °C to achieve 40–70% of the coke combustion. The partially regen-
erated catalyst with less than 0.5 wt% coke is transported to the elevated second
regenerator where complete regeneration is achieved at almost 900 °C with slight
air excess and under a low steam partial pressure. Such conditions allow almost
complete removal of coke.
440 4 Engineering technology

External
Withdrawal well
cyclones

Second-stage Reactor
regenerator Riser termination
Combustion air ring device

Packed stripper
First-stage
regenerator Reactor riser
Combustion air ring MTC injection
Lift air Feed injection

Fig. 4.44: R2R FCC process. (From [26] copyright Elsevier.).

Rather recently, UOP has started to market a new concept of millisecond cata-
lytic cracking (MSCC), which is designed to operate at very short contact times.
Short contact times could have advantages, as product selectivity is higher over
freshly regenerated catalysts than with the catalyst that was in contact with feed. In
particular, short contact times favor the formation of desirable liquid product and
minimize the formation of gas and coke. The process (Fig. 4.45) is organized in a
way that instead of widespread and more costly riser reactors, the feed is injected
perpendicular to a falling curtain of the regenerated catalyst. The feed, which is va-
porized, and the catalyst move horizontally across the reaction zone and immedi-
ately enter the separation section.

Fig. 4.45: Scheme of millisecond catalytic cracking: 1,

2 3 regenerated catalyst; 2, steaming section; 3, to separation; 4,
feed.
 4.4 Hydrocracking 441

4.4 Hydrocracking

4.4.1 Overview of hydrocracking

Hydrocracking (HC) process was developed to produce high yields of distillates with
better qualities than can be obtained by FCC (fluid-bed cracking cracking), which is
described in the previous section. FCC is performed under more severe conditions
than HC and is mainly aimed for gasoline production. Milder conditions of HC allow
also to produce middle distillates, thus making the process more flexible.
An example of how HC can be incorporated in an oil refinery with some other
catalytic processes is presented in Fig. 4.46.

Fig. 4.46: An example of utilization of hydrocracking (HDC) in oil refining for fuel production [27].

Importance of HC, especially in Europe and the USA, is illustrated by the fact that
currently HC of vacuum gas oil represents approximately 7–8% of processes in oil
refining in terms of volume. Capacity of hydrocrackers can be up to 3–4 million tons
per year. In the USA, hydrocrackers were traditionally used to produce naphtha from
low-value aromatic streams including LCO product from fluid catalytic cracking;
however, more units are shifting to production of middle distillates.
442 4 Engineering technology

HC feedstocks are typically flashed distillates from vacuum distillation, from

catalytic and thermal cracking, or deasphalted oils. Operating conditions in the re-
actor section of hydrocrackers are usually about 400 °C and 8–15 MPa.
Major advantages of the process include the full-scale production of high-qual-
ity products, that is, no low-grade fuel oil remains as residue, and the high flexibil-
ity toward product yields, that is, possible gasoline maximization, or gasoline–
kerosene, or kerosene–gas oil maximization.
The reactions in hydrocrackers take place on metal sulfide catalysts in the pres-
ence of hydrogen resulting in hydrotreating, in which impurities such as nitrogen,
sulfur, oxygen and metals are removed from the feedstock, and HC per se when car-
bon–carbon bonds are cleaved with hydrogen addition over bifunctional catalysts.
Besides these reactions, there are several others such as hydrogenation of olefins
generated during cracking and skeletal isomerization of alkanes giving equilibrium
composition of normal and isoalkanes. From the viewpoint of catalyst stability, HC
catalysts operating under milder conditions than in FCC and in the presence of hy-
drogen are less prone to catalyst deactivation.
The process is carried out in fixed catalytic reactors at 280 and 475 °C and hydro-
gen pressure between 3.5 and 22 MPa depending on the feedstock properties and the
products desired. Heavier feedstock requires higher temperatures and pressures and
longer residence time. Lower pressures (3.5–8 MPa) correspond to mild HC.
Typically, cracking conversion is between 30% and 70% per pass to achieve re-
quired selectivity to naphtha or middle distillates, and can be increased by recy-
cling of the unconverted feed close to 100%.
A schematic view of HC reactions is presented in Fig. 4.47.

Polyaromatics
Phenanthrenes
Hydrogenation

fluorenes Naphthalenes
tetrahydro-
phenanthrenes

Dinaphthene Tetralins Paraffin

Alkyl-benzenes Dealkylation
benzenes and indanes cracking

Multiring 2-Ring Single-ring Paraffins

naphthenese naphthenese naphthenese
Ring opening (followed by dealkylation)

Fig. 4.47: Schematic hydrocracking chemistry.

 4.4 Hydrocracking 443

The first step in the mechanism of HC is dehydrogenation of normal alkanes to al-

kene intermediates at very low near-equilibrium concentrations, which react on
Brønsted-acid sites giving alkoxides. Subsequent acid-catalyzed isomerization and
beta-scission cracking is followed by desorption from the acid sites and shuttling to
metal sites, where isomerized and cracked species are hydrogenated. In parallel to
this route, alkenes can form coke. Keeping low concentrations of alkanes allows
to maintain activity of HC catalysts for several years, while FCC catalysts as pre-
sented in Section 4.3 deactivate in seconds.
One of the design parameters for HC units is hydrogen pressure, whose eleva-
tion diminishes coking and extends time between regeneration. Obviously, in bi-
functional catalysis with consecutive reactions the rates of dehydrogenation/
hydrogenation on metal sites and reactions on acidic sites should be harmonized.
The same holds for location of these functions, which, in an ideal HC catalyst for a
long time, were supposed to be in a relatively close proximity after the initial hy-
pothesis of P. Weisz. Such proximity ensures that the equilibrium composition of
alkenes is maintained at the acid sites.
An example of the product distribution of vacuum gas oil at different temper-
atures is given in Tab. 4.7, while compositions of the product depending on the
main target of HC is illustrated in Tab. 4.8,

Tab. 4.7: Yield of vacuum gas oil hydrocracking products at different reaction
temperatures.

Product – oC – oC

HS . .

C–C . .
C–C . .
C–C . .
Gasoline (b.p. up to  oC) . .
Diesel (– oC) . .
Recycle (above  oC) . .
Hydrogen consumption, vol % . .

Occurrence of two main types of reactions, hydrotreating and HC, allows to split
processes in modern hydrocrackers into units with two reaction stages – a hydro-
genation step or desulfurization, denitrogenation and deoxygenation using co-
balt–molybdenum catalysts and an HC step using nickel–tungsten catalysts.
Hydrotreating catalysts are used upstream of conventional HC resulting in higher
yields of gasoline. There is also a possibility to use milder conditions for HC will
lower catalytic activity and higher yields of middle distillates.
444 4 Engineering technology

Tab. 4.8: Yield of vacuum gas oil hydrocracking depending on the desired main
target.

Yield, vol% Gasoline Jet Diesel

Butane . . .

Light naphtha . . .
Heavy naphtha . . .
Jet fuel – . –
Diesel – .
Hydrogen consumption, m/m of feed   

4.4.2 Hydrocracking catalysts

HC catalysts (more than 200 available) combine acid and hydrogenation compo-
nents; therefore, during HC, acid-catalyzed isomerization and cracking reactions as
well as metal-catalyzed hydrogenation reactions occur resulting in the products
with lower aromatic content, more naphthenes and highly branched paraffins. As
mentioned earlier, olefins are completely hydrogenated.
Hydrogenation function is provided by noble metals and combinations of cer-
tain base metals. Either platinum and palladium or sulfided forms of molybdenum
and tungsten-promoted nickel or cobalt are applied. The noble metal HC catalysts
are commonly used in very-low-sulfur or sulfur-free environments with the metal
loading typically below 1 wt%. The base metal content is substantially higher, with
2–8 wt% cobalt or nickel, and 10–25 wt% molybdenum or tungsten. The cracking
function is provided by one or a combination (Fig. 4.48) of zeolites (e.g., zeolite Y)
and amorphous silica–aluminas selected to suit the desired operating and product
objectives.

Hydrocracking catalyst
(dual function)

Cracking function Hydrogenation function

(acid support) (metals)

Amorphous Low zeolite/ High zeolite Nonnoble metals

(SiO2 , Al2O3, amorphous + binder Noble
MxSy from group VIA
Al2O3, Hal- (modified Y/ (modified Y+ metals
(Mo, W)
Al2O3) SiO2– Al2O3) Al2O3) (Pt, Pd)
+ Group VIIIA (Co, Ni)

Fig. 4.48: An overview of hydrocracking catalysts.

 4.4 Hydrocracking 445

Acidity of amorphous silica–alumina HC catalysts is moderate, requiring higher

temperature and longer residence time. At the same time, such catalyst can give a
higher yield of diesel fraction and prevent extensive HC. As illustrated in Fig. 4.49,
there is also a possibility to have a complex catalyst loading.

VGO Loading of catalysts Optimal reactor loading of a

zeolitic catalyst is 20-40%

NiMo/AI2O3
NiMo/ASA- 30%
NiMo/AI2O3
AI2O3
60% NiMo/ASA- NiMo/ASA-
60% NiMo/AI2O3
AI2O3 AI2O3
30% 100%
100%

NiMo/y-AI2O3 NiMo/y-AI2O3 NiMo/y-AI2O3

40% 40% 40%

Maximum yield High T

Products Diesel fraction yield ~ 50% of diesol fraction

Fig. 4.49: Options for using complex catalyst loading with different acidic catalysts (ASA stands for
aluminosilicates).

Unlike the amorphous-based catalysts, the zeolite-containing materials are usually

more acidic and thus selective to lighter products. Microporous character of zeolites
with narrow channels restricts accessibility of larger organic molecules to the active
sites. Y-type zeolites are thus modified by removing aluminum through hydrother-
mal treatment (steaming at temperatures above 540 °C) or first acid washing with
subsequent steam calcination. An alternative approach is to improve acidity of
amorphous aluminosilicates.
An HC catalyst support can therefore be tailor-made depending on how the cata-
lyst is to be operated and the product type needed. To obtain maximum conversion to
gasoline at the lowest operating temperature, up to 80% Y zeolite and about 20% of a
peptized alumina binder is used as support. High middle-distillate conversion is ob-
tained at higher operating temperatures with a support containing about 10% dealu-
minated Y-zeolite, 70% alumina or silica/alumina in varying proportions and 20% of
the binder. The appropriate Y-zeolite and the proportion to be used in the catalyst can
be selected for the products, as required.
For a long time, the concept of ideal HC was used as a guideline for development
of bifunctional HC catalysts, as well as skeletal isomerization of hydrocarbons. It im-
plies careful optimization of the cracking activity of the support and the hydrogena-
tion activity of the metal component, allowing maximum conversion with different
446 4 Engineering technology

operating conditions and feeds. Namely, an ideal HC catalyst was supposed to have a
close proximity between the metal and acid sites; thus, an optimal location of the
metal function in bifunctional catalysts was thought to be in the micropores of
the zeolite. More recently, it was shown that the metal particles can be located on the
binder or on the surface/in the mesopores of the zeolite [28].
HC catalysts during operation gradually lose activity, which is compensated by
an increase in temperature. HC catalysts typically operate for cycles of several years
between regenerations depending on the process conditions. Catalyst regeneration
involves combustion of coke in an oxygen environment (dilute air) either in the re-
actor or externally (Fig. 4.50). Until the mid-1970s, hydroprocessing catalysts were
regenerated in situ in the unit reactors, lately ex situ regeneration started to be
mainly used for many reasons, including corrosion, safety, time and better activity
recovery.

Fig. 4.50: Ex situ regeneration of spent hydrocracking catalyst by combustion in air.

Such regeneration procedure used for nickel-molybdate or nickel-tungstate cata-

lysts is followed by resulfidation. The situation with deactivated palladium sup-
ported on zeolite is more complicated because of metal sintering. Treatment with
an excess of ammonium hydroxide solution dissolves agglomerated palladium giv-
ing a tetramine complex, which is decomposed after calcination in air. In this way,
the original palladium distribution and activity are restored.
Activation of nickel catalysts by sulfidation is done either ex situ or in situ
using the same type of organic sulfur compounds and procedures as implemented
for hydrotreating catalysts. Palladium catalysts should be carefully reduced in hy-
drogen at approximately 350oC, avoiding overheating and thus sintering.
Development of commercial HC catalysts is aimed not only on optimizing the
chemical composition, but also on the catalyst shape. Some of the shapes are pre-
sented in Fig. 4.51. In Criterion’s Advance Trilobe Technology, a replacement of a
cylindrical catalyst to a trilobe can elevate the diesel yield by 1.5%, which for a hy-
drocracker capacity of 50,000 bbd capacity gives daily additionally 750 bbl.
 4.4 Hydrocracking 447

Cylinder Ring Twisted

Trilobe Quadrilobe
trilobe

Fig. 4.51: Some shapes of commercial hydrocracking catalysts.

Sulfided Ni–Mo catalysts supported on a suitable γ-alumina with the nickel mo-
lybdate content higher than in typical hydrotreating catalysts are used for first-
stage HC reactions. Sulfided Ni–W or palladium supported on zeolite or silica/alu-
mina are applied in the second HC stage. Palladium catalysts are active if the feed
contains residual sulfur compounds but are not active for the hydrogenation of ben-
zene rings.
A scheme for production of sulfided Ni–W catalyst is shown in Fig. 4.52.

H2WO4
H2O

Ni(NO3)2 6H2O Na2CO3 3 4 7

H2O H2O

air
8
H2O 10
1 2
5
9
6

air
18 air
16 11
H2

H2S 12 13 air
15 14

19 17

Fig. 4.52: A scheme for production of sulfided Ni–W catalyst. 1,2,7, tanks for aqueous solutions of
Ni(NO3)2, Na2CO3 and H2WO4; 3, 4, measuring tanks; 5, precipitation vessel; 6, filter press; 8,
mixer; 9, tabletting; 10, 11, drying ovens; 12, furnace; 13, heat exchanger; 14, sulfidation oven; 15,
milling; 16, bunker for milled catalyst; 17, mixing sliders; 18, tabletting; 19, packaging [29].

The catalyst is prepared in the following way. The aqueous solutions for nickel ni-
trate and sodium carbonate (pos.1) and (2) through the metering tanks (pos. 3, 4)
are introduced in vessel 5 (pos.5), where mixed nickel carbonate–nickel hydroxide
is precipitated according to
448 4 Engineering technology

5Ni ðNO3 Þ2 6H2 O + 5Na2 CO3 + ðn + 3ÞH2 O = 2NiCO3 3NiðOHÞ2 + nH2 O + 10NaNO3
+ 3CO2 + 30H2 O

The suspension is filtered (pos. 6), washed and milled (pos. 8) in the presence of
tungstic acid H2WO4. From the obtained suspension, the tablets are formed (pos. 9),
which are then dried for one day in pos. 10. In the downstream dryer (pos.11), the
temperature is 378–388 K. Sulfidation with a mixture of hydrogen and H2S is per-
formed in oven pos. 14. Milling (pos. 15) gives particles of the size 0.15 mm, which
after additional milling (pos. 17) are tablettized (pos. 18) to the size of 8–10 mm
with the bulk density of 2,000 kg/m3. Nitric acid is added during forming as a pep-
tizing agent.

4.4.3 Hydrocracking technology

The following factors affect product quality and quantity and the overall econom-
ics: catalyst type, process configuration (e.g., one or two stages) and operating con-
ditions (e.g., conversion, hydrogen pressure, liquid hourly space velocity, feed/
hydrogen recycle ratio).
The role of catalyst was described in Section 4.4.2. Here other parameters will be
analyzed. One of the parameters in HC which prolongs catalyst lifetime by diminish-
ing deactivation is hydrogen pressure. At the same time, an increase of pressure en-
hances hydrogenation of aromatic compounds diminishing the octane number.
Another important operation parameter is catalyst average temperature (CAT),
which is the volume average temperature of catalyst bed. Higher CAT allows lower
residence time and thus higher throughput for the same product quality or im-
proved quality at the same feed rate. Moreover, more difficult feedstock with, for
example, higher sulfur content can be processed. The downside is increased deacti-
vation and thus shorter cycle time.
Cracking of large multiaromatic compounds requires first saturation of rings.
Since hydrogenation is an exothermic reaction, from the viewpoint of equilibrium it
should be conducted at lower temperature. Thus, higher temperature makes it more
difficult to saturate and subsequently crack such aromatic compounds. An increase
of T also enhances production of naphtha and light gases.
Gas-to-oil (hydrogen-to-oil) ratio in most modern hydrocrackers is designed to
be 4 to 5 times larger than hydrogen consumption in various reactions. Increased
hydrogen recycle improves catalyst stability, minimizes overcracking and acts as a
heat sink, limiting the temperature rise in the catalyst beds. At the same time, too
high recycle rate is not economical.
Several flow schemes of HC are possible. The once-through arrangement is the
most cost-effective option and processes very heavy, high boiling feed. There is also
a possibility to have a separate hydrotreating unit, which is installed when there
 4.4 Hydrocracking 449

are special feed considerations, such as high content of product in the feed or high
nitrogen content.
Otherwise, either single- or two-state arrangements are commonly applied. The
first option is cost-effective for moderate capacity giving a moderate product qual-
ity, while the two-stage configuration is more effective for large capacity units af-
fording high product quality and can be applied for difficult feedstock.
Flow schemes for HC unit comprise such sections as a high-pressure reaction
loop, a low-pressure vapor–liquid separation section, a product fractionation sec-
tion and a make-up hydrogen compression section. As mentioned earlier, the gas-
to-oil ratio is higher than hydrogen consumption; thus, the excess of hydrogen is
recovered. A high-pressure loop can account for 70–85% of the installed cost of
the hydrocracker. The reactors are typically trickle-bed downflow reactors with
multiple beds.
Such multibed arrangement is needed as exothermic cracking and saturation
reactions result in a large heat release; therefore, a cold recycle gas is introduced
between the beds to quench the reacting fluids and to control the extent of tempera-
ture rise and the reaction rate. An alternative for heat management is to use heat
exchangers between the beds. Depending on the conversion level and the type of
feed processed, the number of beds can vary from two to eight.
Important for trickle-bed reactors is uniform reactant distribution across the
catalyst; otherwise, maldistribution leads to hot spots that deteriorate catalyst per-
formance and life.
A typical example of a two-stage hydrocracker is shown in Fig. 4.53. In the two-
stage flow scheme, feedstock is treated and partially converted in a first reactor sec-
tion. Products from this section are then separated by fractionation. The bottoms
from the fractionation step are sent to a second reactor stage for complete conver-
sion. This flow scheme is most widely used for large units as mentioned earlier.
The partially cooled reactor effluent is often flashed in a hot high-pressure separa-
tor (HHPS). This allows withdrawal of the heavy unconverted oil at a high temperature
minimizing heat input to the product recovery section. The vapor from HHPS is first
cooled in a heat exchanger. Addition of water is needed to prevent formation of solid
ammonium hydrogen sulfide (NH4HS) from ammonia and H2S, which can otherwise
deposit on the air-cooler tubes, reduce heat transfer and eventually plug the tubes.
After cooling the HHPS effluent vapor in an air cooler in the downstream, cold
high-pressure separator (CHPS) hydrogen-rich vapor is separated from water and
hydrocarbon liquid phases. The latter is sent to the cold low-pressure separator
(CLPS), while hydrogen-rich gas is either directly recycled to the reactor as feed or
quench gas with gas compressor or is first purified from H2S using absorption with
MDEA (methyl diethanol amine) or diethanol amine.
After expansion (pressure reduction), the liquid from the HHPS is sent to the
hot low-pressure separator (HLPS). Hydrocarbon liquid from the CHPS is combined
with the vapor from the HLPS and cooled in an additional air cooler before entering
450 4 Engineering technology

Make-up
hydrogen HHPS - Hot high-pressure separator
Compressor T
CHPS - Cold high-pressure separator
Recycle Lean
gas compressor amine HLPS - Hot low-pressure sepaarator
Amine
absorber CLPS - Cold low-pressure separator

Rich
Stage 2: amine C4 - light
hydrocracking Naphtha
naphtha
reactor To H4 to gas
Stage 1: to gas
hydrotreating/ recovery plant plant
hydrocracking Water CHPS
reactor
Fractionator

Sour Product
water stripper Jet fuel
Oil feed HHPS
Recycle
oil Steam
CLPS
Diesel
HLPS

Steam

Fig. 4.53: Two-stage hydrocracking [30].

the CLPS. Hydrogen still present in the vapor from CLPS is recovered, for example,
by pressure swing adsorption.
The fractionation section consists of a product stripper and an atmospheric col-
umn, separating reaction products into light ends, heavy naphtha, kerosene, light
diesel, heavy diesel and unconverted oil.
When a refiner wishes to convert all the feedstock to lighter products, the frac-
tionator bottoms stream can be recycled back to the reactor and coprocessed with
fresh feed.
The single-stage flow scheme with recycle is illustrated in Fig. 4.54. Typically,
there are two reactors with the first one for hydrotreating and the second for HC
catalyst. In most modern units, recycle oil is blended in with fresh feed before proc-
essing in the first reactor.
Hydroconversion of heavy oil feeds (with up 4% sulfur and more than 400 ppm
of metals) – gas oils, petroleum atmospheric and vacuum residues, coal liquids, as-
phalt, bitumen and shale oil is done using fixed, moving and ebullated bed reac-
tors. This process results in a wide spectrum of lighter products such as naphtha,
light and middle distillates, and atmospheric and vacuum gas oils. Such feedstock
 4.4 Hydrocracking 451

Make-up
hydrogen
T
Compressor
KO drum
Lean amine
H4S To gas plant
absorber

Rich amine
Water To H4recovery
injection Condenser
SW

CLPS
CHPS

SW
HCR
reacto
HHPS

Naphtha
HLPS

Fractionator

Jet fuel
Condenser

SW Diesel
Product
stripper Steam

Steam
Unconverted
High pressure oil to FCC
Low pressure

Fig. 4.54: One-stage hydrocracking [30].

contains high-molecular-weight polynuclear aromatic, coke-forming and organo-

metallic compounds (with nickel and vanadium).
Residium HC requires temperatures between 385 and 450 °C, hydrogen partial
pressures from 7.5 to 15 MPa and space velocities from 0.1 to 0.8 vol h−1. During HC
besides cracking of large molecules, heteroatoms (S, N, O) and metals (vanadium,
nickel) are removed. Removal of nitrogen is more difficult than sulfur removal, thus
some nitrogen compounds are merely reduced only to lower boiling range nitrogen
compounds rather than to ammonia.
Metal sulfides formed during HC are absorbed on the catalyst, plugging the
pores and resulting in catalyst deactivation. Minimization of coking can be done by
452 4 Engineering technology

elevation of hydrogen pressure, introduction of aromatic diluents and continuous

physical removal of coke precursors from the reactor loop.
As catalysts in residuum fixed-bed hydroprocessing, extrudates (8–16 mm) are
used with the active metals (Co, Ni, Mo, W, etc.) impregnated on alumina. The pores
should be large enough to allow transport of large asphaltene molecules.
Different catalysts can be used in different reactors, for example, a demetalliza-
tion catalyst in the first reactor, and highly active Ni–Mo desulfurization catalysts
in the subsequent two reactors.
Catalyst addition to optimize catalyst usage is done in a countercurrent fashion
with installation of the fresh catalyst into the third reactor. After a certain time, it is
withdrawn and reused in the second reactor. Finally, the catalyst from the second
reactor is reused in the first reactor. Accumulation of carbon can be as high as 40%
in the case of vacuum bottoms.
Few types of hydroprocessing reactors rely on moving or ebullating catalyst beds
with catalyst grain size less than 1 mm in size to facilitate suspension by the liquid
phase in the reactor. Ebullated bed reactors operating also with NiMo or CoMo cata-
lysts with the catalyst bed expansion of about 30–50% have up to 30 m height and
5 m diameter (Fig. 4.55). The feed and hydrogen are added from the bottom. In LC-

Catalyst
addition

Max liquiq
level

Level Expanded
detectors level

Settled
catalyst level

Gas/liquid
product
Distributor
to separators
grid plate
Catalyst
withdrawal

Make-up H2 Recycling
and feed oil oil

Fig. 4.55: Ebullated bed reactor for LC fining [31].

 4.4 Hydrocracking 453

fining technology, several reactors are arranged in a series with the final product
going to a separator. The vapor stream containing hydrogen is treated with amine in
an absorber, purified with pressure swing adsorption, recompressed and recycled
back to LC-fining reactors. A decrease of pressure before heat exchange, removal of
condensates and purification gives considerable savings in investment compared to a
conventional high-pressure recycle gas purification system. The liquid stream after a
separator is sent to the hydrotreated distillate fractionator.
In the ebullated bed reactor, a part of the liquid product is recycled through a
large pan at the reactor top and the central downcomer, by means of a pump mounted
in the bottom head of the reactor. Such flow is needed to keep the bed in an expanded
(ebullated) state and assure near-isothermal reactor temperature. Catalyst is added
and withdrawn from the reactor to maintain a stable activity without a need for unit
shutdown.
Fixed-bed hydrotreater/hydrocracker can be also put downstream of the ebul-
lated bed reactors operating at the same pressure level. The feed for such fixed-bed
reactor is the vapor stream from the ebullated bed reactors and the distillate recov-
ered from the heavy oil stripper overhead and the straight run atmospheric and vac-
uum gas oils. This design makes use of excess hydrogen in the effluent vapors to
hydrotreat the distillate fractions. Additional hydrogen, equivalent only to the
chemical hydrogen consumed in the fixed bed reactor, is introduced as quench to
the second and third catalyst beds.
In LC-fining technology with ebullated bed reactors, HC is performed at approx-
imately 410–440 °C, pressure 110–180 bar, giving HDS efficiency of 60–85% and
the following product distribution (% w/w): C4 2.35; C5 177 °C – 12.6; 177–371 °C –
30.6; 371–550 °C – 21.5; above 550 °C – 32.9 [32].
Eni slurry technology (Fig. 4.55) with the capacity of 2 million tons annually is
operating at Eni’s Sannazzaro refinery at 15–16 MPa and 410–450 °C with molybde-
nite MoS2-nanosized catalyst generated in situ from oil-soluble precursors. An at-
tractive feature of such technology is the absence of aging and plugging of pores,
avoiding, therefore, catalyst substitution with subsequent shutdown of HC plants
typical for other catalytic hydrotreating processes.
Small size of catalysts leads to the absence of mass transfer limitations, increas-
ing the overall catalyst activity and allowing to keep a very low (few thousand ppm)
catalyst concentration.
High degree of back-mixing in a bubble column reactor (height 58 m, diameter
5.4 m, weight 2,000 t) operating in slurry phase ensured almost flat axial and radial
temperature profiles making the reactor intrinsically safe against temperature
runaway.
The separation system in this technology is similar to other HC units with the
main difference that the unconverted bottom material is recycled back to the reactor
with the dispersed catalyst. A small purge (<3%) needed to limit the buildup of met-
als (Ni and V) fed when the heavy feed is processed.
454 4 Engineering technology

H2 Rec. Gas

Reaction
product

Refined
products

Slurry reactor
(SCO)

H2
Catalyst
precursor
Catalyst and residue Purge
Feed recycle
(a) (b)

Fig. 4.56: Eni slurry moving bed hydrocracking technology [33]: (a) scheme and (b) reactor.

4.5 Steam reforming of natural gas

4.5.1 General

Steam reforming of natural gas or methane is the most common method of producing
commercial hydrogen as well as the hydrogen used in the industrial synthesis of am-
monia and methanol (Fig. 4.57). In addition to natural gas, other hydrocarbons

Ammonia
synthesis

Natural gas
H2 Hydrotreating
or
hydrogenation
naphtha

Fuel cell
Steam reforming power
or
partial oxidation

FT synthesis of
liquid fuels
Biomass
or CO + H2
coal (syngas)
Methanol
synthesis

Fig. 4.57: Routes to hydrogen and syngas.

 4.5 Steam reforming of natural gas 455

containing streams such as associated gas, liquid petroleum gas, and naphtha boil-
ing up to 220 °C, can be fed to ammonia plants. The application of higher hydrocar-
bons, however, can lead to excessive coke formation on the catalysts.
At high temperatures (700–1,100 °C) steam yields carbon monoxide and hydro-
gen over a nickel-based catalyst in a reversible endothermic reaction with methane:

CH4 + H2 O $ CO + 3H2 ðΔH0298 = 206 kJ mol − 1 Þ (4:10)

Although natural gas consists of some other compounds this reaction is the main
one occurring during steam reforming and is accompanied by extensive coke forma-
tion, which leads to significant catalyst deactivation by methane decomposition:

CH4 $ C + 2H2 ðΔH0298 = 75 kJ mol − 1 Þ (4:11)

or by Boudouard reaction

2CO $ C + CO2 ðΔH0298 = − 173 kJ mol − 1 Þ (4:12)

The equilibrium gas composition is given in Fig. 4.58.

4.0

3.5
H2 (g)
3.0

2.5
H2O (g)
Mole

2.0

1.5
CH4 (g)
1.0
CO (g)
0.5
CO2 (g)
0.0
0 200 400 600 800 1,000 1,200
Temperature (°C)

Fig. 4.58: Equilibrium gas composition at 1 bar as a function of temperature at the steam to
methane molar ratio 2:1.

Obviously, the hydrogen and carbon monoxide content is increasing with tempera-
ture increase. At higher temperatures dry reforming of methane is also becoming
important, thus the concentration of CO2 passes through a maximum (Fig. 4.58):

CH4 + CO2 $ 2CO + 2H2 ðΔH0298 = 247 kJ mol − 1 Þ (4:13)

456 4 Engineering technology

The conversion of methane to CO and hydrogen is favored at high T, low pressure

and high steam to methane ratio [or steam to carbon (S/C)]. In practice, an increase
of temperature is limited by thermal and mechanical stability of tubes, as the pro-
cess requires the use of multitubular reactors. In particular, thick walls would
worsen heat transfer, while tubes with thin walls might be subjected to severe dam-
age at high temperatures, including bending, formation of cracks, etc.

While working in industry I once had a discussion about steam reforming catalysts with a plant
manager of an ammonia production plant, and I was suggesting that a novel catalyst will im-
prove the performance by having a higher approach to equilibrium. The reply was that the com-
pany should first replace 500 tubes, as they look like “macaroni”.

The operating steam to hydrocarbon ratio must be higher than the stoichiometric
level to avoid carbon formation on the catalyst by cracking reactions, and to pro-
vide enough steam to operate the water-gas shift reaction later in the process. Thus,
in steam reforming of natural gas the steam to carbon ratio is 1 to 3:3.8.
The presence of excess steam in the process gas to the reformer results in the
formation of carbon dioxide by the water-gas shift reaction. Thus, the gas leaving
the steam reformer also contains between 7 and 15% carbon dioxide.
In ammonia plants the methane reforming reaction from the tubular (primary)
reformer is continued in the secondary reformer via the introduction of air to the
reactor (Fig. 4.59). The combustion of the air produces temperatures around 1250 °
C, resulting in further reforming of methane.
In ammonia or hydrogen production units, additional hydrogen is obtained by a
standalone production step where water-gas-shift reaction is done with the carbon
monoxide produced upstream. This reaction, which is mildly exothermic, occurs at
a lower temperature than steam reforming:

CO + H2 O ! CO2 + H2 (4:14)

and is usually conducted in industry in two process steps with iron- and copper-
based catalysts called the high- and low-temperature shift reactions, respectively.
Thereafter carbon dioxide is removed by absorption using either KOH or amines
(monoethanolamine or activated MDEA) scrubbing systems, and traces of residual
carbon monoxide are converted to methane over a nickel-based methanation cata-
lyst. The final step is ammonia synthesis, which will be discussed in detail below.
The iron catalyst has not changed significantly since its first introduction by BASF
in 1913, although there are some recent developments to apply Ru/C instead of iron.
A variety of different converters has been developed and will be considered in the
subsequent section devoted to ammonia synthesis.
 4.5 Steam reforming of natural gas 457

Air compressor
Air
Shift
Water Steam converters
Recycle Natural gas and
gas Saturator purge fuel gas

Natural Steam Boiler

gas drum

Preheater
Steam Secondary
Desulfurizer Primary reformer reformer

Purge fuel gas

Hydrogen
recovery unit Synthesis
CO2 converters
Purge gas
Recycle
gas
Methanator

Separator
CO2 CO2
absorber stripper Synthesis gas
compressor
Refrigeration Liquid ammonia to
section refrigeration and storage

Fig. 4.59: Single train ammonia synthesis. (From [34]).

4.5.2 Process conditions for sulfur removal and primary reforming

Because the hydrocarbon feedstock for the production of hydrogen or synthesis gas
contain catalytic poisons, which are detrimental for nickel catalysts, feed purification
is performed in order to remove such poisons, namely sulfur and chlorine containing
compounds. They are removed first by the hydrogenation of organic sulfur, nitrogen
or chlorine compounds (for example, mercaptanes, etc.) to hydrogen sulfide using,
for example, cobalt/molybdenum hydrodesulfurization catalysts (eq. 4.7). These reac-
tions are exothermic and require hydrogen, which is generated in steam reforming.
The cobalt/molybdate component is sulfided during commissioning and operates
thereafter in a similar same way as in refinery hydrotreating.
At the end of its lifetime, the spent cobalt/molybdate catalyst in a typical am-
monia plant can contain 0.5–3.0 wt% sulfur and 2–10 wt% carbon depending on
the operation severity. As adsorbed hydrogen and carbon deposits could be present
in hydrodesulfurization catalysts, making them pyrophoric, after use catalyst dis-
charge is organized in such a way that prior to it the reactor must be flushed with
an inert gas until the catalyst temperature has decreased to ~25–30 °C.
458 4 Engineering technology

Because of very low levels of sulfur compounds in the feedstock, a temperature

rise in the fixed-bed reactors used for hydrodesulfurization is seldom observed.
This step is followed by scrubbing hydrogen sulfide with ZnO:

H2 S + Zn ≥ ZnS + H2 O (4:15)

This reaction is not a catalytic one and the front of the produced ZnS is slowly mov-
ing along the reactor length. An example of a hydrodesulfurization/steam reforming
unit in a plant producing 1,360 t of ammonia per day is given in Tabs 4.9 and 4.10.

Tab. 4.9: Conditions in hydrodesulfurization/steam reforming in an ammonia

plant producing 1,360 t of ammonia per day (AM76 design).

Parameter Hydrodesulfurization ZnO Primary reformer

Catalyst volume (m)   .

Bed height (m)  . .
Diameter (m) . . . (internal)
. (external)
Pressure (bar)   
T inlet (°C)   
T outlet (°C) 
CH, out (vol%) .
Number of tubes 
Lifetime (years) – –

Tab. 4.10: Conditions in hydrodesulfurization/steam reforming in an ammonia plant producing

1,360 t of ammonia per day (AM76 design).

Flow Amount T
 –
Hydrocarbons (vol%): C = ., C = ., C = ., C = ., , m h  °C
N = , CO = ., HS < ppm
H recycle (vol%): H = ., N = ., CH = ., Ar =  , m h–  °C
Steam to reformer , kg h–  °C

A new development in the area of hydrodesulfurization was made in 2010 by

Johnson Matthey Catalysis in collaboration with Prof. Avelino Corma (Valencia). The
new catalyst provides three-in-one functionality, combining full HDS conversion, H2S
absorption and ultrapurification. For a new plant design with an HDS vessel up-
stream of a ZnO based H2S removal system, the HDS vessel can be completely re-
moved from the design.
 4.5 Steam reforming of natural gas 459

4.5.3 Kinetics and mechanism

In the classical work conducted by Temkin and his co-workers [35], reforming kinet-
ics was studied on nickel foil at atmospheric pressure in a flow circulation system
in the temperature range of 470–900 °C. Nickel foil was selected because it is diffi-
cult to avoid diffusion restrictions of the reaction, even for fine catalyst grains of
commercially supported catalysts. It was concluded that the ratio of the partial
pressures in the exit mixture was approximately close to the equilibrium constant
of the water-gas shift reaction at 800 °C. A retardation of the reforming rate by hy-
drogen was observed at 470–530 °C. In a more extensive work by the same group
[36], the following mechanism was proposed with two routes, the second one being
the water-gas shift reaction.

Nð1Þ Nð2Þ
ð1Þ CH4 + z $ zCH2 + H2 1 0
ð2Þ zCH2 + H2 O $ zCHOH + H2 1 0
ð3Þ zCHOH $ zCO + H2 1 0
(4:16)
ð4Þ zCO $ z + CO 1 0
ð5Þ z + H2 O ≡ zO + H2 0 1
ð6Þ zO + CO ≡ z + CO2 0 1

Nð1Þ: CH4 + H2 O = CO + 3H2 ; Nð2Þ: CO + H2 O = CO2 + H2 .

In (4.16) z is the surface sites, steps 5 and 6 are quasi-equilibria, while other steps
are reversible.
The kinetic equation for steam reforming [route N(1)] has the form:

k1 PCH ð1 − xð1Þ Þ
r=   2  4  3    (4:17)
P P P
1 + I1 PH2 O + I2 P H2 + I3 P H2 1 + K5 PH2 O
H2 O H2 O H2

where k1, k2, k3, k4, k–1, k–2, k–3, k–4 are the rate constants of corresponding elemen-
tary steps; I1 = k–1/k2, I2 = k–1 × k–2/k1 × k2, I3 = k–1 × k–2 × k–3/k2k3k4, and K5 is the
equilibrium constant of step 5.
At sufficiently low hydrogen pressure the rate of the reforming reaction can be
expressed as:

r = kPCH4 ð1 − X1 Þ (4:18)

where Xð1Þ = ðPCO P3 H2 =K1 PCH4 PH2 O Þ reflecting the approach of reaction (1) in eq. 4.16
to equilibrium, K1 is the equilibrium constant of reaction (1).
460 4 Engineering technology

The first-order in methane, which follows for the equilibrium, corresponds to

the dissociative adsorption of methane as the rate-determining step.
The simplified mechanism was later on extended to address some important kinetic
(reaction orders at elevated pressures) and mechanistic aspects of steam reforming.
Thus the formation of carbon dioxide might proceed by the interaction between
adsorbed oxygen atom and adsorbed CO rather than as a Eley-Riedal type of mecha-
nism (step 6 in eq. 4.16). It is also probable that adsorbed H atoms participate in the
reaction steps.
The model of Xu and Froment [37] was based on the experimental data obtained
in a plug flow reactor between 500 and 550 °C and the total reaction mixture pres-
sure from 3 to 15 bars. The ratio of partial pressures of hydrogen to methane and
water to methane at the reactor inlet was 1.25, and 3 or 5, respectively. The catalyst
had grains of size 0.18–0.25 or 0.3–0.4 mm. The rate of methane consumption de-
creased with increasing hydrogen partial pressure and increased with increasing
steam partial pressure. The order of the methane consumption rate with respect to
methane was significantly lower than unity.
The mechanism of steam reforming proposed by Xu and Froment [37] consists
of three routes, each comprising a single rate-determining step, while the other
steps are at quasi-equilibria.

Nð1Þ Nð2Þ Nð3Þ

ð1Þ H2 O + z ≡ zO + H2 1 1 2
ð2Þ CH4 + z ≡ zCH4 1 0 1
ð3Þ zCH4 + z ≡ zCH3 + zH 1 0 1
ð4Þ zCH3 + z ≡ zCH2 + zH 1 0 1
ð5Þ zCH2 + zO ≡ zCH2 O + zH 1 0 1
ð6Þ zCH2 O + z ≡ zCHO + zH 1 0 1
ð7Þ zCHO + z $ zCO + zH 1 0 0
ð8Þ zCO + zO $ zCO2 + z 0 1 0
ð9Þ zCHO + zO $ zCO2 + zH 0 0 1
ð10Þ zCO ≡ CO + z 0 1 0
ð11Þ zCO2 ≡ CO2 + z 0 1 1
ð12Þ 2zH ≡ zH2 + z 2 0 2
ð13Þ zH2 ≡ H2 + z 2 0 2
Nð1Þ:CH4 + H2 O = CO + 3H2 ;Nð2Þ:CO + H2 O = CO2 + H2 ;Nð3Þ:CH4 + 2H2 O = CO2 + 4H2
(4:19)
 4.5 Steam reforming of natural gas 461

The rate of steam reforming is then given by the sum of two routes (the first and the
third):

k1 PCH4 PH2 O ½z2 k1 PCH4 P2 H2 O ½z2

rð1Þ = 5 ð1 − Xð1Þ Þrð3Þ = 5 ð1 − Xð3Þ Þ (4:20)
PH2 2 PH2 2

where:
1
½z =   (4:21)
PH O
1 + KH2 O PH
2
+ KCO PCO + KH2 PH2 + KCH4 PCH4
2

The kinetic model was able to describe experimental data, although there were
still some mechanistic deficiencies. The experimental dependence on the reaction
rate of the steam partial pressure cannot be explained using the theoretical value
of pre-exponential factor for steam adsorption coefficient. The model of Xu and
Froment [37] also suggests high coverage of non-dissociated methane, which is
unlikely.
A microkinetic model from Aparicio [38] comprised 13 elementary steps, such as:
adsorption and stepwise dissociation of methane to adsorbed CH3, CH2, CH, C and H
species; adsorption and dissociation of steam to OH and H; formation of CHO by reac-
tion of adsorbed carbon and hydroxyl; dissociation of CHO to adsorbed CO and H
atoms; molecular adsorption of CO2 with subsequent reaction with an adsorbed H
atom to form formate (COOH) intermediate; desorption of CO and H (with recombina-
tion). The assumption of the involvement of surface intermediates CHO and COOH
was based on low-temperature data, while these intermediates were not observed at
high temperature. This indicates a danger of direct application of a microkinetic ap-
proach relying on experimental data, which correspond to conditions (temperature,
pressure, catalyst) deviating substantially from conditions of real catalysis. In addi-
tion, the model is unnecessary complicated not allowing an explicit rate expression
to be obtained.
Further mechanistic elaborations by Avetisov et al. [39] resulted in the mecha-
nism that takes into account stepwise dissociation of methane, adsorption of water
with dissociation and a step of the interaction between adsorbed carbon monoxide
and adsorbed oxygen:
462 4 Engineering technology

Nð1Þ Nð2Þ
ð1Þ CH4 + 2z $ zCH3 + zH 1 1
ð2Þ zCH3 + z $ zCH2 + zH 1 1
ð3Þ zCH2 + z $ zCH + zH 1 1
ð4Þ zCH + z $ zC + zH 1 1
ð5Þ zC + zOH $ zCOH + z 1 1
ð6Þ zCOH + z $ zCO + zH 1 1 (4:22)
ð7Þ zCO ≡ z + C 1 0
ð8Þ zCO + zO $ 2z + CO2 0 1
ð9Þ H2 O + 2z ≡ zOH + zH 1 2
ð10Þ zOH + z ≡ zO + zH 0 1
ð11Þ 2zH ≡ z + H2 3 4
Nð1Þ:CH4 + H2 O = CO + 3H2 ;Nð2Þ:CH4 + 2H2 O = CO2 + 4H2

This mechanism leads to a very complicated rate equations (notation of the authors
of [39]):

rCH4 = k1 PððCHÞ4 Þ ð1 − ð1 =Kðð1ÞÞ ððPðH

3
2Þ
PCO Þ=PðH2 Þ PððCHÞ4 Þ ð1 + ðkð − 8Þ =kð − 7Þ ÞðPððCOÞ2 Þ =PCO Þ½z=

ð1 + ðkð − 8Þ =kð − 7Þ ÞKðð2ÞÞ ðPðH2 OÞ =PðH2 OÞ Þ½zÞÞÞ

(4:23)

with:
1
½z = 0 1 (4:24)
PCH
1 + KCH3 @ 1
4 A 1− ð1 − X
DEN + KCO PCO
P2
H2

5
1 3 PH2 2 PH3 2 ½z
DEN = 1 + A1 PH2 + A2 PH2 + A3 PH2 + A4
2 2
+ A5     (4:25)
PH2 k−8 PH O
PH2 O 1 + k−7 Kð2Þ PH
2
½z
2

k − 1 21 k − 1k − 2 k − 1 k − 2 k − 3 32
A1 = K − 11 , A2 = K − 11 , A3 = K − 11 (4:26)
k2 k2 k3 k2 k3 k4
 
k − 1k − 2k − 3k − 4 k − 5 52 k1
A4 = 1+ K − 11 , A5 = (4:27)
k2 k3 k4 k9 k6 Kð1Þ k − 7

where KCO = k − 7 =k7 .

 4.5 Steam reforming of natural gas 463

Simplifications, however, are possible, leading to eq. (4.18), which is recom-

mended for practical purposes, especially at sufficiently large steam to hydrogen
ratios.

4.5.4 Technology

In steam reforming for ammonia production, the conditions for most of plants are
designed to afford certain methane content in the outlet of the primary reformer to
be adequate for the downstream secondary reforming where the ratio between hy-
drogen and nitrogen of 3 mol/mol should be achieved. Some examples of operation
conditions for steam reforming in ammonia synthesis plant are given in Tab. 4.11.

Tab. 4.11: Plant data for steam reforming in ammonia synthesis of 1,000 metric
ton per day (mtpd) with 19 m3 of steam reforming catalyst.

Lifetime (months)   
 −
Feed (Nm h ) , , ,
S/C (mole mole−) . . .
Tube inlet (°C)   
Tube outlet (°C)   
Pressure inlet (bar) . . .
Pressure drop (bar) . . .
CH leakage (vol%) . . .

Inlet methane 85–88%, ethane 5–6%.

The amounts of CO and CO2 the conditions in Tab. 4.11 would be almost in the same
range as methane (9–11%), with a somewhat higher concentration of CO2 (11% CO2
vs 9% CO).
Increased heat duty (outlet temperature) in a primary reformer leads to lower
methane leakage, however, it increases the risk of excessive temperature rise in the
combustion zone of a secondary reformer.
Conventional reformers for ammonia synthesis production were designed with
a steam to carbon ratio of 3.3–3.6 mol/mol, while there is a trend nowadays to di-
minish this ratio as it reduces overall production costs.
The optimum conditions for hydrogen production, including the steam to car-
bon ratio, could be different from ammonia synthesis plants. They are dictated by,
for example, such considerations as the feedstock. Hydrogen plants can operate
based on natural gas or light hydrocarbons through to naphtha. As in refineries hy-
drogen production is integrated with other refinery units, feedstock to steam re-
former could contain a mixture of off-gases from the catalytic reformer, catalytic
cracker and other units.
464 4 Engineering technology

In a steam reformer (Fig. 4.60) the gas passes up through vertical reactor tubes
with a supported nickel catalyst, typically ca. 10 m long and ca. 10 cm in diameter.
The long catalyst tubes are suspended vertically in rows within the furnace. Heat is
supplied to the tubes using burners. For effective operation, heat must be trans-
ferred rapidly from the furnace itself to the surface of the catalyst, particularly at
the top of the tubes.

Inlet manifold

Burners Steam
reformer

Reformer
tubes

Cold outlet
manifold system
(A) (B)

Fig. 4.60: View of a steam reformer.

The hydrocarbon should react rapidly with the steam, which helps to avoid the
cracking reactions and thus catalyst deactivation. For inactive catalysts, overheat-
ing of the reactor tubes can occur. One of the ways to avoid overheating, besides
introducing an active catalyst, is also shaping the catalyst in a way that allows
rapid gas mixing, efficient heat transfer from the wall of the tube to the catalyst and
uniform temperature gradient throughout the tube. Careful loading of the catalyst
is also very important to achieve uniform gas flow and pressure drop through each
tube. When these conditions are not reached, flow maldistribution occurs, leading
to variable tube temperatures and hot spots.
A penalty of operating a reformer with hot tubes, for any reason, is that a tem-
perature only 10 °C above the design level can reduce the tube life by up to 50%.
The cost of tube failures resulting from the use of low activity catalysts or decreas-
ing the steam ratio is very high.
 4.5 Steam reforming of natural gas 465

Commercial tubular reformers operate at space velocities of 2,000–8,000 h–1

(STP) which are dictated by heat transfer and to lesser extent mass transfer. High
flow rates and the need to minimize pressure drop require the application of rather
large catalyst pellets. Therefore, a variety of shapes (rings or spheres with multiple
holes, wagon wheels, clover leafs, etc.) have been proposed to minimize pressure
drop on one hand and on another improve catalyst effectiveness factor, which is
rather low (0.03–0.1).
Catalyst performance in primary reforming reactors is evaluated by several cri-
teria, such as methane leakage (approach to equilibrium), pressure drop, tube wall
temperature (hot spots), resistance against failures and catalyst lifetime.
The tube wall temperature is influenced by the catalyst (Ni content; surface
area; shape; resistance to poisoning by S, Cl, As; degree of reduction) and process
conditions (steam/carbon ratio; average heat flux; content of higher hydrocarbons;
hydrogen content; pressure; inlet/outlet T; space velocity; operating failures).
While the process gas is fed downwards inside the tubes, the flue gas can go
downwards in top-fired reformer furnaces or upwards in side wall, terrace wall or bot-
tom-fired furnaces. The tube wall temperature is the highest on bottom-fired furnaces
compared to other types of furnaces for the same process gas outlet temperature.

4.5.5 Catalysts

The catalyst in steam reforming of hydrocarbons generally operates at severe condi-

tions, thus high thermal stability is important. In addition, as deactivation caused by
the formation of carbon deposits and coke could be detrimental for the catalyst per-
formance, the catalytic phase, promoters or supports should be carefully selected.
Carbon formation is, for example, favored at a low steam to carbon ratio and acidic
catalyst support, while the presence of hydrogen decreases the generation of carbon.
There are peculiar features in the carbon or coke formation during the steam reform-
ing of methane, making it different from coke in many other reactions related to
hydrocarbon transformations. In particular, carbon forms needles, growing as fila-
ments with the active catalytic metal (nickel) at the top. This can lead to structural
damages of the catalysts and should be avoided as it leads to catalyst replacement.
Although noble metals display higher intrinsic activity than nickel and are
more resistant to coke, in industrial formulations nickel is applied because of the
availability and costs. The choice of support is also crucial, as the proper carrier
material should ensure long lasting activity and mechanical stability. The surface
area of supports is rather low (5–80 m2 g–1).
Among thermally stable supports with low acidity, α-alumina, magnesium alu-
minate, MgO and CaO should be mentioned. CaO, although possessing the highest
thermal stability, has relatively low mechanical strength, therefore it is used as cal-
cium aluminate. Calcium aluminate is very hard and has high initial crushing
466 4 Engineering technology

strength. Under the high pressure of carbon oxides calcium can, however, react
with them somewhat, diminishing the crush strength. The silica content should be
kept minimal (< 0.2 wt%), as it is volatile under normal operation conditions.
MgO should be used with a precaution, because it can hydrate at below 425 °C,
giving magnesium hydroxide and dramatically weakening the catalyst. Reformer
startups and shut-downs with catalysts including MgO in the formulations should
therefore be done in a dry atmosphere (without steam) when temperature is below
the critical hydration temperatures.
Potassium compounds are sometimes used as promoters as they are effective in
lowering the acidity and allowing gasification of carbon with steam. The drawback
of utilizing potassium is in its volatility at high temperature. For example, one com-
mercial supplier offers the catalyst containing 25% NiO and 8.5% K2O on calcium
aluminate for primary reforming of naphtha, while for steam reforming of natural
gas the recommended catalyst is a 10-holed ring (19 × 16 mm) containing 14% NiO
on CaAl12O19 without potassium. For a mixed feed (natural gas/LPG) the catalyst
formulation from this supplier (Süd Chemie, now Clariant) is 18% NiO, 1.6% K2O on
CaK2Al22O34 in the form of a 10-holed ring (19 × 12 mm).
Another company (Johnson Matthey) manufactures several types of catalysts to
be used depending on the feedstock type and plant conditions. These catalysts are
nickel oxide on α alumina, calcium aluminate or a lightly alkalized nickel oxide cat-
alyst on a calcium aluminate support.
Even if the specification above refers to NiO content, catalyst suppliers can ship
primary reforming catalysts in the oxidized or pre-reduced, passivated form.
For example, Haldor Topsoe A/S uses magnesium aluminate as a support with
the nickel content above 12% and manufactures both nickel oxide and pre-reduced
catalyst. In order to initiate the reforming process immediately during the start-up,
it is advised to charge the top 10–15% in the tubes with the catalyst in the pre-re-
duced form.
Nickel can be incorporated onto the support by multiple impregnations,
which usually allows synthesizing stronger catalysts than by precipitation. After
impregnation and calcinations, nickel oxide can be reduced, preferably by hydro-
gen than by steam and hydrogen. The extent of reduction depends on the support
nature, the reduction temperature and time. Too high temperature and reduction
time might lead to sintering. Reduction is done at a temperature close to the outlet
temperature of the reformer, which at the end leads to a catalyst with low metal
dispersion.
In the old formulations for methane steam reforming the amount of nickel
could be rather high, reaching 25% with rather homogenous nickel distribution
along catalyst grains. As, previously mentioned however, the reaction is heavily
influenced by mass transfer with rather low catalyst effectiveness factor. In view of
this, it is apparently clear that with such nickel distribution not all metal is involved
in catalysis, thus nowadays nickel profiling is applied. For an average nickel
 4.5 Steam reforming of natural gas 467

concentration ~15% it is higher at the exterior of the catalyst and is diminishing

along the grain depth.
The catalyst shape is extremely important for steam reforming catalysts, be-
cause it influences activity and the heat transfer coefficient. The shape should be
designed to maximize the heat transfer rate, and subsequently the reaction rate,
and minimize the tube wall temperature. An optimum shape also decreases pres-
sure drop by increasing the voidage and size. In addition, by increasing geometrical
surface area the activity could also be increased with a decrease in the tube wall
temperature.
Although use of Raschig ring-shaped catalysts with different sizes was a common
practice for several decades, modern steam reformers with increased heat transfer
duty require more complex shapes (some of them are presented in Fig. 4.61). Such
complexity is created by using multiple holes or adding flutes to the outside structure.

Fig. 4.61: Complex shapes for steam reforming catalysts.

It should be noted, however, that each hole diminishes mechanical strength, which
decreases with an increase of the size of holes. At the same time, geometrical surface
area is improved, pressure drop is lowered, affording also better heat transfer from
the tube wall to the reacting gases and thus lower tube wall temperature (Fig. 4.62).
A comparison between the plant operation of a spherical catalyst with seven
holes, which has been in operation in one of the ammonia synthesis plants since
2010, with another shaped catalyst is given in Tab. 4.12. In this table the tempera-
ture approach to equilibrium is defined as the difference between the actual and
equilibrium temperature. Obviously, lower numbers indicate a close approach to
equilibrium and thus better catalyst performance.
Careful loading is important in every steam reformer tube in order to have even
distribution and to avoid side reactions, such as carbon formation. Such even load-
ing is done using a special type of socks as discussed in Chapter 3. After filling each
sock, vibration of the tube could be done to ensure uniform packing and minimiza-
tion of the shrinkage during operation. Thereafter pressure drop is measured for all
tubes. Such pressure drop measurements could be done after half filling and com-
plete filling of the tubes. As already mentioned in Chapter 3, tubes displaying a
pressure drop outside ±5% from the average should be refilled or additionally vi-
brated if the pressure drop is too low.
468 4 Engineering technology

900 1,652
1,602
850 1,552
Maximum TWT °C

Maximum TWT °F
1,502
800
1,452
1,402
750
1,352

700 1,302
1,252
650 1,202
0 5 10 15 20 25 30 35 40 45 50
Distance downtube (%)

Fig. 4.62: Maximum wall temperature for different shapes. (From [40] copyright
© Uhde GmbH 2005.).

Tab. 4.12: Comparison between spherical and cylindrical catalysts. Plant data. (From [41].).

Catalyst Run Volume S/G Inlet pressure ΔP Tin(°C) Tout(°C) Approach to

type time (m) (bar) (bar) equilibrium
(°C)

 year  .  .   

 year  . . .   –

4.6 Ammonia synthesis

4.6.1 General

Large-scale production of ammonia is one of the most important catalytic processes.

The process was discovered in 1909 by Fritz Haber and his co-workers (Fig. 4.63),
who demonstrated that it was possible to produce ammonia at a scale of 2 kg day–1
using osmium as a catalyst. The technology was further developed by BASF
under the leadership of Carl Bosch. Such development required the construction
of new high-pressure equipment and the application of a less exotic and ex-
pensive catalyst. Efforts of Alvin Mittash and co-workers, who tested over 2,500
formulations, resulted in Fe catalyst promoted with alumina, calcium oxide and
potassium. The very same catalyst is essentially used in hundreds of industrial
 4.6 Ammonia synthesis 469

1 2

Fig. 4.63: The equipment of Haber and LeRossignol (1909).

installations with a capacity that grew from several tons a day in the original col-
umn erected in Oppau (now part of BASF site in Ludwigshafen) to single trains
with 1,500–1,800 tons per day. Haber and Bosch were later awarded Nobel prizes,
in 1918 and 1931, respectively, for their work.
Development of much more active Ru/C catalysts led to commercial application
of Ru catalysts in few plants in the 1990s as will be discussed later.
Although the chemistry of ammonia synthesis is superficially very simple, this
reaction, which operates at high pressures and temperatures, gave rise to the devel-
opment of many fundamental concepts, such as surface heterogeneity, structure
sensitivity, virtual pressure, rate-determining steps, microkinetics, etc. Reactor de-
sign is also very rich and sophisticated, with different types of reactors employed in
industry.

4.6.2 Thermodynamics

The ammonia synthesis reaction:

N2 ðgÞ + 3H2 ðgÞ ! 2NH3 ðgÞðΔH = − 92.22 kJ mol − 1 Þ (4:28)

is exothermic, with a volume decrease ( ΔV < 0) and is limited by thermodynamics.

Therefore, high P and low T should be preferred (Tab. 4.13).
470 4 Engineering technology

Tab. 4.13: Variation in equilibrium constant Keq for the

ammonia synthesis as a function of temperature.

Temperature (°C) p2NH3

Keq =
pN2 · p3H2

 . × –

 . × –
 . × –
 . × –
 . × –
 . × –

Equilibrium concentrations of ammonia as a function of temperature and pres-

sure are given in Fig. 4.64.

70

60 350 °C
Equilibrium (mol% NH3)

50 400 °C

40
450 °C
30
500 °C
20
550 °C

10

0
0 50 100 150 200 250 300 350 400
Pressure (bar)

Fig. 4.64: Effect of pressure and temperature on equilibrium ammonia concentration at an inlet
hydrogen to nitrogen ratio 3:1. (From [11]).

A catalyst of high activity is thus needed that should be able to operate at low tem-
perature. Unfortunately, high activity could be achieved for rather inexpensive iron
catalysts at 400–450 °C, leading to severe equilibrium limitations. The process is
thus conducted industrially at 15–30 MPa between 300 and 550 °C. High tempera-
ture and pressure are beneficial from the point of view of kinetics, although at the
expense of conversion limitations because of thermodynamic restrictions and costs,
associated with high-pressure technology.
 4.6 Ammonia synthesis 471

Ammonia synthesis reaction is exothermic, thus in order to operate efficiently

equilibrium should be continuously shifted and heat should be removed. This is
done by organizing the process in several stages. After one bed of catalyst, concen-
tration of ammonia is becoming rather substantial (Fig. 4.65), coming closer to the
equilibrium line. Thereafter either intermediate cooling is done by heat exchangers
or ammonia is removed through quenching (adding cold reactants). In this case,
the concentration line (dotted) is not going parallel to x-axis in Fig. 4.65. In should
be noted that at the inlet of the catalyst bed, the reaction is far from equilibrium,
while at the outlet because of high ammonia concentration the reverse reaction is
becoming more prominent influencing in a negative way the forward reaction.

Equilibrium

10
NH3 (vol %)

2
400 450 500
Temperature (°C)

Fig. 4.65: Concentration – temperature diagram in ammonia synthesis.

The single pass through a reactor yield is then ~15%, calling for the reactants recy-
cling. This is done by cooling gases in order to liquefy ammonia, removing it as liq-
uid, and returning hydrogen and nitrogen in the reactor.

4.6.3 Kinetics

Kinetics of ammonia synthesis has received a lot of attention through the years.
Schematically, in the simplest treatment, the ammonia synthesis mechanism near
equilibrium was pictured in the following way:
472 4 Engineering technology

1. N2 + * , *N2
2. *N2 + 3H2 ≡ * + 2NH3 (4:29)
N2 + 3H2 = 2NH3

Here *N2 is an adsorbed intermediate in the form of dinitrogen, step 1 is reversible

and step 2 is in equilibrium.
Based on the experimental data, which covered the pressure range from below
1 atm up to 500 atm, it was proposed that near equilibrium the reaction rate is de-
scribed by the following equation, which is often referred to in the literature as the
Temkin-Pyzhev equation:
!m !1 − m
PH3 2 2
PNH
r = k′PN2 2
− k′′ 3
(4:30)
PNH 3
PH3 2

where m is a constant (0 < m < 1). Under equilibrium, the reaction rate equals 0;
therefore, k+/k– = K, where K is the equilibrium constant. Hence, only one of the
constants, either k+ or k, together with m should be determined from the experimen-
tal data. Equation (4.30) was based on the assumption that nitrogen chemisorption
on an energetically non-uniform surface determines the rate of the overall reaction.
The second step is considered to be an equilibrium involving the adsorbed nitrogen
and the gas-phase hydrogen and ammonia.
Although in the original derivation it was supposed that nitrogen is adsorbed
in the molecular form, an assumption on the dissociation nitrogen adsorption also
leads to the same equation.
At high pressures, eq. (4.30) should be modified to include the deviations from
the laws of ideal gases and to incorporate the effect of pressure on the reaction rate
depending on the volume change of activation. Therefore, eq. (4.30) at high pres-
sures contains not partial pressures, but fugacities.
A concept of fugacity was used also by Nielsen, Kjaer and Hansen [42] who pro-
posed the following equation:

kA pN2 − kB p2NH3
p3H2
r= (4:31)
1 + Ko pNH2
3=2 2m
ðpH2 Þ

where p is fugacity (or effective pressure).

At low partial pressures of ammonia eq. (4.30) predicts infinite reaction rates,
thus for such conditions a modified Temkin, Morozov and Shapatina equation [43]
could be used:
 4.6 Ammonia synthesis 473

2
!
PNH
k1±−PmN 1− 3
2 KPH3 2 PN2 2
r= 2
! m 1 − m (4:32)
l PNH l
+ 3
+ 1
PH2 KPH3 PN2 PH2
2 2

This equation, which provides a more statistically significant fit to the experimental
data, generated by ICI, is based on the assumption that the reaction rate is deter-
mined by two slow irreversible steps. The first step is nitrogen chemisorption, and
the second step is the addition of hydrogen to molecular adsorbed nitrogen:

1. N2 + * , *N2
2. *N2 + H2 , *N2 H2
3. *N2 H2 + 2H2 ≡ * + 2NH3 (4:33)
N2 + 3H2 = 2NH3

In eq. (4.32), which can also be derived with the supposition of dissociative nitrogen
adsorption and which could be transformed to a Temkin-Pyzhev equation closer to
equilibrium, l = k–1/k+2.
Equations of Ozaki-Taylor-Boudart:

kA PN2 − kB ðPNH3 Þ2
ðPH2 Þ3
r= " #2α (4:34)
1 + Kc ðPNH3 Þ
ðPH2 Þ3=2

and Nielsen:
1 1
r = kPH2 2 PN2 2 (4:35)

were also proposed to be applicable at low ammonia concentrations [11]. Equation

(4.34) could be also used at conditions far from equilibrium, when not only the first
step in (4.29) is kinetically significant.
Recent microkinetic analysis by several groups (Dumesic, Stolze, Norskov) led
to the following reaction mechanism, which is thought to operate on both Fe and
Ru catalysts:
474 4 Engineering technology

1. N2 ðgÞ + * ≡ N2 *
2. N2 * + * ) 2N*
3. H2 ðgÞ + 2* ≡ 2H*
4. N* + H* ≡ NH* + *
(4:36)
5. NH* + H* ≡ NH2 * + *
6. NH2 * + H* ≡ NH3 * + *
7. NH3 * ≡ NH3 ðgÞ + *

N2 + 3H2 = 2NH3

This mechanism is based on rapid adsorption of molecular nitrogen and slow dissocia-
tion to atomic nitrogen while other steps are quasi-equilibrated. This in essence means
that only the second step is rate determining leading to the following rate equation:
" #2
PNH3 *2
r = k + 2 θN2 θ* − k − 2 θ2N = k + 2 K1 PN2 θ2* − k−2 pffiffiffiffiffiffiffiffiffiffiffiffi 3 θ (4:37)
K3 K4 K5 K6 ð K7 PH2 Þ

where the fraction of vacant sites is expressed by:

1
θ* = !!
pffiffiffiffiffiffiffiffiffiffiffiffi PNH3 1 1
1 + K1 PN2 + K7 PH2 + 1
K6 PNH3 + pffiffiffiffiffiffiffiffiffiffiffiffi 1 + pffiffiffiffiffiffiffiffiffiffiffiffi 1 + pffiffiffiffiffiffiffiffiffiffiffiffi
K5 K6 K7 PH2 K4 K7 PH2 K3 K7 PH2
(4:38)

Simulations of ammonia synthesis on potassium promoted iron for mechanism

(4.36) using rate and equilibrium constants from surface science studies [44] dem-
onstrated the ability of microkinetic analysis to rather accurately describe experi-
mental data. The agreement between the models (not requiring any data fitting)
and experiments obtained with a commercial catalyst is rather good, considering
that modeling involves an extrapolation of nine or more orders of magnitude in
pressure. Certain discrepancies could be associated with incorrect determination of
the number of sites. Thus, the prediction of ammonia conversion can be improved
considerably if the number of assumed sites is increased by a factor of only 3.

4.6.4 Catalysts

4.6.4.1 Iron
Typical commercial unreduced iron-based catalysts (Fig. 4.66) contain ~0.5–1% K2O,
2–4% Al2O3, 2–4% CaO, while the rest is magnetite Fe3O4 (and some wustite FeO).
 4.6 Ammonia synthesis 475

Fig. 4.66: Commercial iron catalyst for ammonia synthesis.

The formulations are essentially the same as developed by Alwin Mittash in

1909. Alumina functions as a structural promoter, enabling the production and
preparation of an open and porous structure. During the initial catalyst activation,
reduction starts on the outside of the granule, and dissolved alumina separates out
of solid solution, forming in the pores between the iron crystallites. This minimizes
further growth during completion of reduction and later operation, giving small
20–49 nm Fe crystallites.
Calcium oxide, and other basic promoters (for example, MgO, < 0.5%) react
with silica impurities in the raw materials to form glassy silicates, which themselves
can enhance the thermal stability of the reduced iron. The main benefit is to mini-
mize any neutralization of the K2O promoter, which would diminish its effective-
ness. The role of K is more complicated. This component is considered to be an
electronic promoter, functioning by considerably increasing the intrinsic activity of
the high surface area iron particles produced on reduction. Such an enhancement
of activity caused by formation of potassium ferrites is seen up to about 0.8% pot-
ash, while above this level the activity falls. The chemical action of potassium is
associated with a decrease of the adsorption energy on iron, lowering concentration
of adsorbed ammonia and thus increasing the number of sites available for nitrogen
adsorption.
Natural magnetite with low impurity levels is nowadays used to produce ammo-
nia synthesis catalysts.
Preparation of the catalyst by fusion of a mixture of magnetite and the pro-
moters (added as carbonates, oxides or hydroxides, or as nitrate in the case of po-
tassium) is described in Chapter 2. The resulting catalyst is a solid solution of
wustite and alumina in the magnetite crystal lattice. The surface area of the final
catalyst is rather small, in the range 1–2 m2 g–1.
The concentration of alumina should be less than the solubility of alumina in
magnetite, which corresponds to a maximum content of about 3% alumina.
Ammonia synthesis catalysts are prepared in the form of magnetite, which
must be activated by reduction to metallic iron before use. The reduction process is
476 4 Engineering technology

done when the unreduced catalyst is loaded in an ammonia converter prior to oper-
ation. On reduction, the catalyst becomes porous and develops a surface area
around 20 m2 g–1.

In plants using large volumes of catalyst, the reduction process can take several days. This is
because it should be carried out slowly to avoid high levels of water, which then lead to the
growth of the small crystallites and decrease catalyst activity. I once took part in reduction of
ammonia synthesis and had to stay ~30 h in the control room without leaving it to ensure that
no damage to the catalyst would occur.

Assistance in catalyst reduction is usually provided by catalyst suppliers and engi-

neering companies, because plant operators rarely face this operation (the lifetime of
ammonia synthesis catalysts can be 14–15 years and the catalyst is replaced when
the vessel operating at a high pressure requires inspection because of regulations). In
axial reactors with a large diameter the linear gas velocity is relatively low and the
reduced catalyst can suffer some re-oxidation caused by back-mixing of the reducing
gas. Gas flow should therefore be as high as possible. The same is applied in radial
converters, where high flow rates are required to provide uniform reduction profiles.
As the water content of the gas exiting the converter should be limited (max.
3,000 ppm) it is carefully controlled. During reduction more catalyst becomes re-
duced and the heat from exothermic ammonia synthesis supplements the heat from
the start-up heater. This allows an increase in the gas rate while keeping temperature
rise of a bed inlet temperature to about 5 °C h–1 until a maximum outlet temperature
of 500 °C is reached. The gas rates can be increased as more beds are reduced.
In order to diminish reduction time ammonia synthesis catalysts are manufac-
tured in the pre-reduced form. After such reduction done by a catalyst manufacture,
the catalyst is carefully stabilized by a controlled flow of a mixture of nitrogen and
air, resulting in less than 10% of iron re-oxidation to thin oxide film around iron par-
ticles. This procedure allows a starting reduction at a lower temperature (330 °C vs
400 °C for non-reduced catalysts) giving a shorter reduction time at a plant site. Such
shorter procedure brings substantial savings as during catalyst reduction the ammo-
nia synthesis train, including steam reforming of natural gas, is in full operation, con-
suming several dozens of thousands m3 per hour of natural gas. It is thus desirable to
reach the full production capacity as soon as possible, without compromising, how-
ever, long-term catalyst activity.
The pre-reduced catalysts are more expensive than the unreduced one, thus op-
erating costs are balanced with the higher catalyst price. Typically, pre-reduced cat-
alysts are used in only the top reactor sections, simplifying the beginning of the
reduction process.
An example of a start-up procedure is given by Shepetovsky et al. [45]. The ammo-
nia converter located in Berezniki, Russia with a capacity 1,360 t day–1 was revamped
from axial flow to axial-radial flow by Ammonia Casale. It had three beds with axial-
 4.6 Ammonia synthesis 477

radial flow in the top bed (Fig. 4.67) and radial flow of gases through the second and
third beds. For the first bed with axial-radial flow, 13.5 t of pre-reduced catalyst was
used, while 22 and 90 t were used in the second and third beds respectively.

Fig. 4.67: Axial-radial flow distribution. Most of the gases pass in

the radial direction. The balance is passed through the top in axial
direction to eliminate the need for a top cover of the catalyst bed.

It is a typical procedure to load catalysts from a small hopper or a large bag fitted
with a suitably flexible chute. Moving around the chute provides an even level of
loading, keeping the desired catalyst bulk density in all parts of the catalyst bed. In
the example above, during the catalyst loading after filling each 40 cm of bed
height the layer was levelled using first shoveling and subsequently a vibrator.
The temperature, gas-flow rate and pressure profiles are presented in Fig. 4.68.
The following procedure was applied. Heating of catalyst in the first bed, ini-
tially to 300 °C and then to 450 °C, took 48 h. The amount of water released during
this period was 0.15–0.22 vol%. Thereafter, for another 6 h the catalyst was reduced
in the first bed and simultaneously reduction of catalyst in the second bed was initi-
ated. The catalyst in the first bed was completely reduced at 500 °C and in the second
bed the reduction continued at 450 °C for 6 h and then for another 6 h at 450–475 °C.
Simultaneously reduction in the third bed started at 450 °C. During this 12 h-period
the water content at the outlet was 0.1–0.9 vol%. Reduction of the catalyst in the bed
2 was finalized at 500 °C while reduction for the third bed continued for 14 h. Water
content at the outlet was 0.08–0.7 vol%. At the end of this period ammonia was pro-
duced with 99.82% purity. The total duration of the process was 91 h. Plant perfor-
mance data a few weeks after the catalyst reduction are given in Tab. 4.14.
After another 2 weeks the ammonia output was 60 t h–1 (1,450 t day−1) with an am-
monia content at the reactor outlet of 16–18 vol%.
Ammonia synthesis is one of several catalytic processes where catalyst manu-
facturers guarantee catalysts performance. Comparison between guaranteed param-
eters, mean values during warrant tests and performance at nominal capacity is
given in Tab. 4.15.
478 4 Engineering technology

600 Gas load (103 nm3/h)

Pressure (bar)
t (°C), bed 1
500 t (°C), bed 2
Temperature (load, pressure)

t (°C), bed 3

400

300

200

100

0
0 10 20 30 40 50 60 70 80 90 100
Time (h)

Fig. 4.68: An example for ammonia synthesis catalyst reduction procedure. (Redrawn from [45]).

Tab. 4.14: Plant performance data few weeks after start-up. (Modified from [45]).

Date Pressure Gas flow  Inlet/outlet (°C) Outlet/inlet ΔP, NH output
(MPa) (m h–) (MPa) (t h–)
Bed Bed Bed Reactor %
I II III (°C) NH

.. .      . . 

    .

.. .      . . .
    .

.. .      . . .
    .

.. .      . . .
    .

4.6.4.2 Ruthenium
Part of the industrial production now takes place with a more expensive ruthe-
nium-based catalyst (the KAAP process) rather than with cheaper iron because this
more active catalyst allows reduced operating pressures. Little information is
 4.6 Ammonia synthesis 479

Tab. 4.15: Warranty testing data. (Modified from [45]).

Parameter Units Guarantee Mean values Values at nominal capacity

  –
Inlet gas flow  m h   
Inlet reactor T, not more °C   
Outlet reactor T, not more °C   
ΔP MPa . . .
Outlet pressure MPa . . .
Ammonia output t day– , . ,
Inerts % –  

available, however, on the commercial production of the ruthenium-based ammo-

nia synthesis catalyst.
Ruthenium, promoted by alkali metals and barium and then supported on a
cheap carbon support, has provided a significant increase in activity compared
with iron catalysts. K and Cs apparently facilitate nitrogen adsorption, allowing
electron charge transfer to Ru, similar to iron-based catalysts. A magnesium oxide
support has been claimed to give a longer life. Ru/C demonstrated much higher ac-
tivity than iron at moderate hydrogen to ammonia ratios and ammonia contents,
while the rate of ammonia production is strongly inhibited by ammonia. Operation
at lower temperatures and pressures is thus favorable over Ru, and this catalyst
should not be used at higher ammonia concentrations.
The history of Ru catalyst development and scale-up at BP is documented in
the literature [46].
In 1992, low-pressure ammonia synthesis was implemented using a KBR
Advanced Ammonia Process (KAAP) technology at the Methanex (formerly Pacific
Ammonia) plant in Kitimat, British Columbia, Canada. This was the result of a retrofit
of an existing plant with a new radial flow ammonia converter filled with Ru on a
stable graphite carbon support. Ru catalyst is 20 times as active as the iron catalyst.
During the scaling up of the process, a process demonstration unit was erected
capable of producing about 2 tons of ammonia per day. The process demonstration
unit ran in various tests for over 8 years. A major problem during the start-up was
deactivation caused by lower levels of contaminants in the process gas compared to
the cylinder gas used in the laboratory reactors. Such low levels enhanced reduc-
tion of the promoters, thereby deteriorating catalyst performance at the demonstra-
tion plant. Diminishing the promoter loadings mitigated catalyst deactivation.
A wood-based carbon was selected as a catalyst support. The challenge is the
carbon support scale-up involved in the development of furnace facilities for con-
tinuous treatment of the active carbon precursor at temperatures up to 2,500 K and
controlled air oxidation.
In the first KAAP plant the effluent gas contained about 20% ammonia. The cat-
alyst was reduced at 300 °C, which took less than one day, with the evolution of
480 4 Engineering technology

only small amounts of water. The catalyst was operated at lower pressures and tem-
peratures than the iron catalyst, and contrary to iron could also tolerate the pres-
ence of 4,000 ppm of carbon monoxide. Because of high costs of the natural gas
this plant was first closed temporary in 2000–2002 and then finally in 2006.
In 1998 two grassroots KAAP plants producing ammonia at 1,850 t day–1 began
operating in Trinidad and Tobago using one bed of iron catalyst and three beds of
ruthenium catalyst. A third KAAP plant in Trinidad started up in 2002 and a fourth
began operating in 2004.
Start-up of other KAAP plants in Trinidad, Egypt and Venezuela was in 2008–
2010. In new designs, lower pressure (90 bar) and temperature, less gas compression
and less high-pressure equipment lead to lower operating and capital costs, less down-
time while still being capable of maintaining the required ammonia yield per pass.

4.6.5 Reactors and process design

The ammonia synthesis reaction occurs in a loop where hydrogen and nitrogen syn-
thesis gas is circulated continuously through a converter containing the catalyst
(Fig. 4.69).

Steam
(super-
heated)

Economizer

Feed
water
Chiller #2 Chiller #1 Air
cooler
Liquid Ammonia
ammonia tank
Ammonia
synthesis
Feed gas converter
Syn-gas preheater
Make-up gas circulation

Fig. 4.69: Ammonia synthesis loop. (From [47]). Reproduced with permission.

Ammonia is condensed (condensation temperature in industry is typically −6 °C) and

removed from the loop. In order to avoid accumulation of inert gases in the loop,
such as methane and argon, they are continuously removed in a purge gas stream.
In the past, axial flow reactors were used, operating at high pressure (300 bar).
Larger catalyst particles (6–10 mm) were required in order to limit the pressure
 4.6 Ammonia synthesis 481

drop through the catalyst. As the catalyst effectiveness factor (at large values of
Thiele modulus) is inversely proportional to particle size, strong diffusional limita-
tions were present. By designing converters in such a way that gas flows radially
through the catalyst bed, it is possible to decrease the overall pressure drop, using
smaller catalyst particles (1.5–3 mm) exhibiting higher activity per unit volume.
Revamp of axial flow plants to radial flow operation mode does not lead, however,
to higher ammonia production. Instead, pressure in the converter is decreased to
20–22 MPa, lowering production costs.
As explained above, because of equilibrium limitations, reactors typically in-
clude either interstage cooling or the addition of cold synthesis gas, an operation
known as quenching.
Several reactor options are installed in commercial ammonia plants. In the
Haldor Topsoe S-200 radial converter, two catalysts beds are applied with an inter-
mediate cooling (Fig. 4.70A). The S-300 reactor from the same company has three
beds and two interbed exchangers. This converter enables a higher conversion for
the same catalyst volume or smaller catalysts volumes and lower investment costs
for new plants. Similar designs (with three beds and radial flow from outside to in-
side) are available from other companies (for example, Uhde GmbH, Fig. 4.70B).
Kellogg axial converters (Fig. 4.70C) have several beds with quenching.

Pressure shell Start-up Gas inlet Outlet

gas Bypass
control
Indirect Main gas
inlet Heat
quench Quench
exchanger
gas inlet gas

Inter-bed First bed

First
heat
catalyst
exchanger
bed

Second bed
Annulus
Second around Cartridge
catalyst catalyst
bed bed
Third bed
Lower heat
exchanger Catalyst

Ga outlet
Gas Inlet

(A) Cold bypass (B) Gas outlet (C)

Fig. 4.70: Ammonia synthesis converters with: (A) reactor with two beds, from [48] copyright UNIDO
and IFDC, (B) reactor with 3 beds (from [49]), (C) reactor with 3 beds and quenching.

In Fig. 4.71 a horizontal ammonia synthesis reactor is shown that allows removal of
the catalyst basket from the converter without aid of heavy cranes usually needed
to lift the catalyst basket out of vertical converters.
482 4 Engineering technology

Out
Thermowell

Manhole Manhole

Bed 2B Bed 2A Bed 1

Thermowell
Quench

In

Fig. 4.71: KBR horizontal ammonia synthesis reactor. (From [48]).

In the reactors discussed above, the catalysts were distributed in several beds op-
erated adiabatically in an attempt to follow the optimal equilibrium curve (Fig. 4.72).

Equilibrium line
1
Ammonia conversion

Maximum
reaction curve

(A) Temperature (B)

Fig. 4.72: Isothermal ammonia converter with (A) temperature profiles with (1) multitubular reactor,
(2) isothermal design; (B) cooling plates.

In a departure from this concept Ammonia Casale developed a pseudo-isother-

mal ammonia synthesis converter. The new design is based on the use of cooling
plates immersed in the axial-radial catalytic bed to remove the reaction heat
while it is formed. This is a very efficient design and can help to build mega ca-
pacity ammonia plants. So far, Ammonia Casale revamped a plant in Thailand
running at capacity of 753 mtpd to boost it to 953 mtpd with energy savings of
0.96 Gcal MT–1. Significant reductions in the circulation rates and specific duties
of heat exchangers were obtained. Lately, reactor internals were installed in an-
other plant in Canada [50].
In ammonia dual-pressure process by Uhde (Fig. 4.73), the synthesis section
comprises a once-through section followed by a synthesis loop. In this way, large
capacities can be achieved (4,000–5,000 metric tons per day).
 4.7 Oxidation 483

Conventional process NH3

once-through Synthesis
+ Once-through synthesis reactor
S: Separator
ammonia synthesis WC: Water cooler
= Dual-pressure process CH: Chiller with
HP BFW / steam NH3 refrigerant

Purge gas
1 2 3 R

NH3 synthesis
loop
S S
Synthesis gas NH3 product
WC H2O WC CH CH CH ≈2/3
(make up gas)
≈1/3

Fig. 4.73: Ammonia dual-pressure process by Uhde [51].

The ammonia synthesis loop consists of two stages. An indirectly cooled once-
through converter operates at approximately 110 bar. The major part of ammonia
produced in this converter after cooling is separated from the gas products. Overall,
the once-through converter generates one-third of the total ammonia. The ammonia
synthesis loop per se operates at approximately 210 bar.
Although many ammonia plants were revamped to improve performance effi-
ciency there are still some barriers to their implementation of new technologies.
Among these barriers, limited knowledge of process technologies outside the plant
should be mentioned, as small operators are not aware of all the best technologies.
Plant operators also want more data on new technologies to prove its reliability and
are not eager to be the first to make risky investments. In additional, key barriers
such as low energy costs, lack of corporate R&D, capital availability and corporate
priorities could hinder more widespread implementation of new technologies.

4.7 Oxidation

4.7.1 General

Oxidation reactions are very important for production of intermediates, different

chemicals and monomers for the polymer industry (Tab. 1.3). A key issue in many
catalytic oxidation reactions is selectivity, as even a superficially minor increase
in selectivity brings substantial economic returns because of large production
484 4 Engineering technology

capacity. The challenges in improving selectivity are related to (a) potentially

higher reactivity of the desired product to further oxidation, (b) intermediates in
complex oxidation reactions can lead to by-products rather than the desired ones,
(c) presence of different functional groups in the reactant also prone to oxidation
in addition to the desired functionalization and (d) homogeneous radical reac-
tions can contribute to formation of by-products.
Typically, during oxidation reactions complete oxidation can also take place in
addition to partial oxidation, which is often the desired one. Selectivity control by
variation of reactant concentration is difficult to achieve, thus temperature remains
as the main parameter to regulate selectivity. Typically, the activation energy of
complete oxidation is much higher (by 20–40 kJ/mol) than for the selective oxida-
tion, thus with temperature increase selectivity is diminished. This points out on a
need of efficient temperature control. Overheating in such exothermal processes
should be avoided as it worsens the overall performance. For consecutive reactions
when the desired intermediate product is further oxidized, selectivity toward this
intermediate product is decreasing with conversion. Consequently, an optimum
conversion should be typically achieved, which is done by regulating the contact
time (in continuous processes) and applying an excess of the organic compound,
while oxygen is in deficiency.
The latter operation strategy is also related to safety aspects, which are ex-
tremely important, as many mixtures of organic compounds are explosive. Thus,
safe processing requires operation below or above explosion limits. Low concen-
tration of the organic substrate implies low rates and need to recycle substantial
amounts of the oxidant. The reaction mixture can be also diluted with steam.
Introduction of oxygen at different points in the reactor is another operation
strategy.
From the engineering viewpoint, an important characteristic of catalytic oxida-
tion reactions, besides explosivity, is their high exothermicity. Apart from few ex-
ceptions (e.g. oxidation of SO2 to SO3) adiabatic reactors are not suitable, as
overheating of the catalyst bed leads to deactivation or thermal runaway. In order
to facilitate the heat removal catalytic oxidation is thus often conducted in multi-
tubular or fluidized bed reactors. Multitubular reactors do not have a uniform tem-
perature profile, and hot spots exist moving along the reactor length as the catalyst
deactivates with time-on-stream. From the point of view of heat removal, fluidized
bed reactors can be considered as isothermal. On the other hand, such arrangement
results in back-mixing diminishing selectivity. This could be partially avoided by
having several segments (Fig. 4.74a) or using a riser (similar to FCC), where the cat-
alyst is separated in a cyclone and send back to the reactor. In Fig. 4.74a, a possibil-
ity to use cooling agents at different stages of the multistage fluidized bed column
is illustrated. Temperature regulation can be also done by quenching with the cold
reactant (Fig. 4.74b).
 4.7 Oxidation 485

Products Products

Steam
Steam

Reactants
Reactants

Water
Water

Fig. 4.74: Partial segmentation in the fluidized bed reactor (a) without and (b) with quenching.

Selectivity could be also controlled by a proper catalyst design and addition of

inhibitors, which diminish the overall catalyst activity, improving selectivity. For
example, some known catalyst poisons (halogens, selenium) are added to silver-
based catalyst for oxidation of ethylene to ethylene oxide. In the same process, di-
chloroethane is introduced to the feed diminishing many-fold activity, increasing,
however, selectivity toward the desired product of partial oxidation. Selectivity
could be also regulated by selecting a proper type of the support, its porosity or size
of the catalyst grains.
The catalyst composition and morphology play an important role. Several
groups of catalytic materials are used for oxidation, such as metals (Cu, Ag, noble
metals), oxides of transition metals and mixed oxides of, for example, V, Bi, Mo, W.
The mechanisms of oxidation using heterogeneous catalysts were discussed in
Section 1.2.4.1.

4.7.2 Epoxidation of olefins

4.7.2.1 Ethylene oxide

One of the most versatile organic intermediates is ethylene oxide with an epoxide
ring, which can be readily opened with, for example, water giving ethylene glycol.
Opening of the ring in ethylene oxide by reacting with alcohols, fatty alcohols and
ammonia (amines) give, respectively, polyols, surfactants and ethanolamines. EO is
also an important monomer in production of polyesters and polyurethanes. Annual
worldwide production of EO is expected to exceed 32 Mt by 2023 [52].
486 4 Engineering technology

Oxidation of ethylene is done at elevated pressures (between 1 and, respec-

tively, 2.2 or 3 MPa for oxygen or air-based processes) using silver catalysts with
metal loading 7–20 wt% on an inert support with no acid centers to avoid side reac-
tions. The surface area of the support is rather low (<2 m2/g) and only high-tempera-
ture stable supports (such as α-alumina or SiC) are used. One of the reasons to use
a low surface area support is to avoid oxidation of ethylene oxide in the pores when
there are internal mass transfer limitations.

This is an interesting example how knowledge in reaction kinetics and catalytic reaction engi-
neering, namely dependence of selectivity to the intermediate product in consecutive reactions
on Thiele modulus (Section 3.5.6), can guide design of catalytic materials. Many textbooks on
chemical reaction engineering do not emphasize clearly such links between theory and practi-
cal chemical applications. As a result, sometimes “classical” chemists do not appreciate the
impact that kinetic analysis and reaction engineering can bring to their everyday work on for
example organic reactions or materials development. I remember a colleague, an organic chem-
ist, who attended a conference on chemical engineering and was complaining that “this dc/dt
is not chemistry at all”. Apparently, the driving force for attending that conference was its loca-
tion (marvelous city of Florence in Italy).

Catalyst manufacturers keep information about the composition of catalysts confi-

dential; therefore, details of formulations are not readily available. Several pro-
moters, for example, salts or other compounds of alkali (Li, K, Na) and alkaline
earth metals (Ba, Ca) are added to the catalyst significantly improving selectivity.
Application of Rb or Cs oxide in combination with Re and its copromoter (e.g., S, P,
Mo, Mn or Cr) was also reported. Light alkali (e.g., LiCl) can be used in combination
with the heavy ones (e.g., CsCl). Moreover, chlorine compounds (e.g., 1,2-dichloro-
ethane) are introduced to the reaction gases, having a negative impact on activity,
improving, however, selectivity by suppressing total oxidation of ethylene to car-
bon dioxide and water.
These measures allow to reach 80–90% selectivity with oxygen as an oxidation
agent. Selectivity is declining with time-on-stream due to several reasons (poison-
ing, sintering, dust formation, etc.). The range of temperatures is 220–250 °C. Too
high temperature obviously leads to overoxidation and formation of carbon diox-
ide, while catalysts might not be active enough at too low T. Oxidation of ethylene
to epoxide is mildly exothermic (ΔH = −105 kJ/mol) while complete oxidation of eth-
ylene and ethylene oxide to carbon dioxide and H2O is much more exothermic
(ΔH = −1324 and −1220 kJ/mol, respectively).
Besides CO2, water and ethylene oxide traces of acetaldehyde and formaldehyde
are formed and must be removed or separated from the recycle gas stream. Catalyst
life is typically 2–5 years. Thereafter, silver is recovered from the spent catalyst.
Either air or oxygen is used for oxidation of ethylene oxide synthesis in multi-
tubular reactors (Fig. 4.75) containing several (up to 20) thousands of 6–13 m long
tubes with internal diameter of 20–50 mm filled with the catalyst in the form of
 4.7 Oxidation 487

Fig. 4.75: Reactor for ethylene oxide synthesis: (a) outside view and (b) schematics [53].

spheres or rings with a diameter of 3–10 mm. The contact time is approximately 1 s
at 230–290 °C and 10–30 bar.
As in other selective oxidation processes, a special care should be taken not
only about conversion to avoid unwanted complete oxidation, but also about con-
centration range of ethylene in the mixtures with oxygen, since such mixtures are
explosive. As a result, ethylene conversion per pass is limited being 20–65% in the
air-based compared with 7–15% for the oxygen-based process. The catalyst produc-
tivity is approximately twofold higher in the oxygen-based process, which also dis-
plays better selectivity to ethylene oxide. The oxygen-based process is organized in
the concentration domain above the explosion limit. In the oxygen-based process,
methane and argon are applied (Tab. 4.16) to improve the overall safety.
Incomplete conversion calls for utilization of gas recycling which in the case of
air-based process means, on one hand, a recycle of a large amount of nitrogen in
the recycle gas and, on the other hand, much larger amounts of purge gas to pre-
vent build-up of nitrogen concentration in the recycle stream. Subsequently, the
off-gas leaving the primary reactor still contains substantial amount of ethylene
which should be processed before venting into the atmosphere. Figure 4.76 illus-
trates the flow scheme of an oxygen-based process.
When oxygen being more expensive than air is used, selectivity is higher, losses
of ethylene are lower as well as the equipment size is smaller. Pressure is typically
1–2 MPa, which does not influence selectivity in the oxidation reaction but improves
downstream absorption. Ethylene is taken in excess compared to oxygen. Due to for-
mation of CO2 it should be efficiently removed. Recycled gas along with fresh oxygen
and ethylene (and inhibitor, not shown) is fed into the reactor 1. Steam is generated
in steam generator 3. Outlet gases are cooled in the heat exchanger 2, cooler 4 and
thereafter send to absorber 5, where ethylene oxide is dissolved in water due to its
488 4 Engineering technology

Tab. 4.16: Process parameters in ethylene oxide production processes.

Parameter Air based Oxygen based

Ethylene concentration, vol% – –

Oxygen concentration, vol% – –

CO concentration, vol% – –

Ethane concentration, vol% – –

Argon concentration, vol% –

Methane concentration, vol% –

Temperature, Co
– –

Pressure, MPa – –

Ethylene conversion, % – –

Overall ethylene oxide yield – –

Fig. 4.76: Oxygen-based synthesis of ethylene oxide. 1, reactor; 2, 8, 10, heat exchanges, 3, steam
generator; 4, cooler, 5, 6, 13, absorbers; 7, compressor; 9, stripper (desorber); 11, column; 12, 14,
distillation columns; 15, pump; 16, valve; 17, condenser; 18, separator; 19, reboiler. Modified from [54].
 4.7 Oxidation 489

miscibility. In addition, a part of CO2 is also adsorbed. A part of the gas (unreacted
ethylene and CO2, along with some impurities) is recycled while the rest is routed to
absorber 6, where CO2 is removed using, for example, hot pot scrubbing or water so-
lution of alkanol amines. Subsequently, in a stripper 9 higher temperatures than for
absorption are applied and CO2 is released. The lean solvent is sent back to the ab-
sorber. After a pressure release to approximately 0.5 MPa (expander 16), water solu-
tion of ethylene oxide and CO2 from the bottom of absorber 5 passes through a heat
exchanger 10 to column 11 where ethylene oxide and CO2 is distilled from water. The
latter after passing a heat exchanger 10 is sent back to absorbers 5 and 13. After distil-
lation (column 12) ethylene oxide (bottom) is sent to the final distillation (column 14),
while a part of ethylene oxide (top of column 12) is captured in the absorber 13 with
water and rerouted to column 11.

“Ethylene oxide is one of the most dangerous large-scale chemicals present in petrochemical
industry and an ability of its safe production is a benchmark, reflecting a level of technological
advancement and maturity of chemical industry in a particular country” told a seemingly old
lecturer to a group of students, including the author, when back in 1984 we took the course on
chemistry and technology of petrochemicals and basic organic chemicals at Mendeleev
University. In fact, the lecturer, probably at that time younger than the author nowadays, did
not elaborate too much on the technological aspects of this technology.
The author lately was involved in several research project on oxidation of ethylene to ethyl-
ene oxide, resulted in few scientific publications, and have visited few years ago production
facilities in Dzerzhinsk, where 4 multitubular reactors with capacity 50,000 t/a each were put
on stream in yearly 1980s. It is instructive to review the historical development at that particular
company [55]. The original selectivity of a silver containing catalyst supported on corundum
was 70–72%, which was quite high in those days. The only by-product was carbon dioxide
apart from almost negligible amounts of aldehydes. The reactors had 13,500 tubes with internal
diameter of 21 mm and the catalyst bed length of 7 m. An organic compound was used as a heat
carrier. The operation pressure was and still is 2 MPa. The inlet contained (vol%) %: 15–20
C2Н4, 6.5–6.8 О2, 7–9 СО2, and nitrogen as the rest. After the revamp in 2005 nitrogen was
switched to methane. The reactor capacity was improved by 20%. The plant operates with a
more modern catalyst affording 84% selectivity upon introduction of less than 5 ppm of a
chlorocontaining compound. A more selective catalyst allows to decrease the generated heat as
less products of total oxidation are formed and to diminish the consumption of ethylene and
oxygen per ton of product. With not that selective catalyst the company experienced a lot of
problems in the initial phase of running the plant. Just during the first year there were 333 unex-
pected shutdowns, sometimes up to 10 per day. Such frequent stops of the process induced by
sudden, without visual causes, sharp jumps of temperature and pressure in the outside-catalyst
space of the reactor, continued for the next few years leading to a considerable decrease in
capacity.
A large heat of the main reaction, substantially increasing when the selectivity is lost and com-
plete oxidation prevails, operation close to an explosion limit and high sensitivity of the reactor to
operational parameters and temperature fluctuations were obvious reasons for such frequent
stops.
In practice after a sudden temperature increase in the bed there was an increase in the pres-
sure drop and a subsequent decrease in the overall gas flow through the reactor, increasing
490 4 Engineering technology

obviously the contact time. Then after 15–30 s the exit gas temperature immediately reached
400–600 °C.
Thermal runaways happened when as a method to mitigate catalyst aging and to maintain pro-
duction capacity, changes of the operation parameters (e.g. elevation of temperature and promoter
concentration) were carried out. Thereafter, such increases in temperature occurred at certain time
points.
Analysis of the reactor performance demonstrated that there was a distribution of catalyst ac-
tivity in different tubes. The same holds for heat exchange efficiency. Tubes with anomalously
high activity and low heat exchange efficiency had increased parametric sensitivity, leading to
thermal runaway in those tubes in the case of small disturbances of process parameters. Ignition
of the reaction mixture happened downstream the tubes when the volume was expanded.
Somewhat similar behavior will be discussed in Section 4.7.3.2 in connection to thermal run-
aways in o-xylene oxidation plants.
There were several other reasons for thermal runaways such as mechanical damage of shaped
catalysts, attrition and formation of dust. Corrosion products coming from pipelines could also
influence thermal runaways. Finally, catalyst exposure to the liquid sorbent led to leaching of pro-
moters across the reactor diameter and the reactor length, especially in the tubes at the reactor
periphery. Subsequently selectivity was decreased and more heat was released.

The first generation of high-selectivity catalysts introduced in the late 1980s allowed
an increase of initial selectivity values by more than 6 percentage points (e.g., from
80%) to give start-of-cycle selectivity values of 86% or larger bringing enormous sav-
ing in ethylene feedstock costs. Activity and stability of these catalysts were lower
compared with the traditional ones, thus selectivity declined more rapidly and the life-
time of high-selectivity catalysts was 1.5–2 years being approximately half of the life-
time of traditional EO catalysts. More recently developed so-called high-performance
catalysts [56] by CRI Catalyst Company (Fig. 4.77) allowed approximately the same ini-
tial selectivity and a significantly slower performance decline. According to real plant
data, the average selectivity over the lifetime of a catalyst is approximately 87%.

91
90
HP catalyst
89
88
Selectivity (%)

87
86
HS catalyst
85
84

82
Catalyst cumulative production (kton/m3)

Fig. 4.77: Selectivity curves comparing high selectivity (HS) and high-performance (HP) catalysts
over their live time at the same operating conditions [56].
 4.7 Oxidation 491

4.7.2.2 Propylene oxide

A similar process for propylene oxide has not been developed, since there are no
catalysts available up to now affording selective oxidation of propylene. Thus,
other routes have been practiced along the years, including the chlorohydrin routes
resulting in large amount of wastes (2.2 tons of salt per ton of product)

An alternative route is to use oxidation of ethylbenzene giving a peroxy compound,

which subsequently reacts with propylene resulting in propylene oxide and a sec-
ondary alcohol.

The later after dehydration gives styrene (2.5 tons per ton of propylene oxide).
Recently, a more direct method for propylene oxide synthesis was developed
based on hydrogen peroxide giving water as the only by-product.

The process flow scheme is presented in Fig. 4.78. Oxidation is done in methanol
solution with titanium silicate TS-1 as the catalyst. Water is formed as a product
from H2O2 along with glycols, which are generated in small amounts as by-
products.

4.7.3 Oxidation of alkanes to anhydrides

4.7.3.1 Maleic anhydride

Maleic anhydride has numerous industrial uses and is of significant commercial in-
terest worldwide. The primary use of maleic anhydride is in the manufacture of
polyester and alkyd resins. Benzene was used as the dominating feedstock
492 4 Engineering technology

Lowboilers

C3″ C3″ Pure PO

H2O2 H2O, glycols
MeOH

Offgas

Water glycols
Main reactor

MeOH purification
PO separation

Finishing reactor

PO purification
Crude PO
O2 removal

separation
Fig. 4.78: Oxidation of propylene to propylene epoxide with H2O2 [57].

being in the recent years continuously replaced by oxidation of n-butane or n-bu-

tane–n-butene mixtures with a high paraffin content (Fig. 4.20).
The traditional process begins by mixing benzene with an excess of air to give
concentrations from 1 to 1.4 mol%. A low benzene concentration must be utilized in
order not to exceed the flammability limit of the mixture. The reaction gas mixture
then passes over the catalyst in a multitubular (diameter ca. 25 mm) fixed-bed reac-
tor at an optimum pressure range of 0.15–0.25 MPa. The desired reaction is highly
exothermic, causing hot spots of 340–500 oC. As a rule, commercial catalysts for
benzene oxidation are supported by an alumina or silica carrier and have a surface
area of 1–2 m2/g. A typical catalyst is comprised of V2O5, MoO3 and some other pro-
moters such as phosphorus, alkali and alkaline earth metals, tin, boron and silver.
Benzene is passed through the catalyst at the concentration of 60–130 g benzene
per liter of catalyst per hour. Conversion 97–98% is achieved with an initial selec-
tivity over 74%. The lifetime of the catalyst can be as long as 4 years depending
mainly on the reactor operating temperature and the purity of the starting
materials.
Recovery of the significant part of the product (40–60%) is done by cooling the
reactor outlet stream to 55 °C. Condensed maleic anhydride must be removed as
 4.7 Oxidation 493

soon as possible to avoid prolonged contact with water in the reaction gas, other-
wise resulting in formation of maleic acid. The rest of maleic anhydride that cannot
be recovered is washed out with water as maleic acid. Water scrubbing and subse-
quent dehydration of maleic acid at approximately 130 °C is required to purify and
reform the remaining maleic anhydride. A side reaction is isomerization of maleic
acid to fumaric acid.
The processes developed or in the developmental stage for the oxidation of C4
hydrocarbons to maleic anhydride imply fixed-bed (Fig. 4.79), fluidized bed and
transport bed reactors.

Tail gas to
incineration Crude MA
n-Butane

6
4 5 Light Pure
1
2 ends MA

3
Air

Heavy Solvent
waste recycle

Fig. 4.79: Synthesis of maleic anhydride by butane oxidation in a fixed-bed reactor.

The reaction is the complex one, requiring extraction of eight hydrogen atoms and
insertion of three oxygen atoms to butane along with a ring closure. High concen-
tration of butane is advantageous, but dangerous since butane and oxygen could
be within explosion limits, thus the concentration of butane in air is approximately
1.8 mol%. A major advantage of C4-based process compared to benzene oxidation
is that no carbon is lost in the reaction to form maleic anhydride. Moreover, the
feedstock is cheaper, not carcinogenic contrary to benzene, and the flammability
limits for C4 hydrocarbons are lower. Oxidation of C4 hydrocarbons at 80% conver-
sion and 70% selectivity gives comparable weight yields as for benzene oxidation.
In the fixed-bed process (Fig. 4.79) oxidation is done in tubular reactors similar to
benzene oxidation with vanadium and phosphorus oxide (V–P–O) unsupported cat-
alysts. In the catalyst formulations having phosphorus–vanadium ratio of 1.2, sev-
eral promoters (Li, Zn, Mo, or Mg, Ca, and Ba) are applied improving conversion
and selectivity compared to unpromoted catalysts. Higher phosphorus–vanadium
ratios increase conversion but lead to shorter catalyst lifetime. An oxidation reactor
494 4 Engineering technology

operates at 400–480 °C and 0.3–0.4 MPa. Formation of water is higher in butane

oxidation processes compared to benzene oxidation, therefore only a small amount
of maleic anhydride can be condensed from the reactor effluent. The remaining
product is washed out as maleic acid in a scrubber and then dehydrated similar to
the case of benzene oxidation. Tail gases after the absorption column are inciner-
ated, as recovery of unreacted C4 hydrocarbons is substantially more difficult than
in the case of benzene. An alternative recovery method utilizes absorption of the
product from the reactor outlet by organic solvents and will be explained below for
the case of the fluidized bed reactor.
Application of such fluidized bed reactors allows increasing the concentra-
tion of butane in the feedstock compared to fixed-bed operation by 50–100%.
The fluidized bed process has the advantage of a particularly uniform tempera-
ture profile without hot spots improving the process selectivity. A particular
problem of this process is mechanical stress on the catalyst (V–P–O), its abra-
sion, and erosion at the heat-dissipating surfaces. On the other hand, the fluid-
ized bed constitutes an extremely effective flame barrier, with the result that the
process can operate at higher C4 concentrations (i.e., within the explosion range)
than the fixed-bed process. The process flow scheme is given in Fig. 4.80. In the
recovery system, crude maleic anhydride is distilled from the high boiling solvent
in a fractional distillation column (stripper in Fig. 4.80) while the solvent is re-
cycled. Further refining of the product is done by removing the light ends and
recycling the heavier ones from the final distillation column to the stripper. The
tail gases are incinerated.
Butane oxidation can be also done utilizing a transport reactor (Fig. 4.21),
where oxidation of the reactants by the lattice oxygen and subsequent reduction
of V+5 to V+4 is separated in space from oxidation of V+4 to V+5. Comparison be-
tween oxidation of butane to maleic anhydride in the fluidized bed and a transport
reactor-regenerator tandem is given in Tab. 4.3. Despite clear advantages of the latter
process with higher throughput and inherent safety, its industrial implementation
was not very successful, and the production was stopped after some time in
operation.

4.7.3.2 Phthalic anhydride synthesis

4.7.3.2.1 Feedstock
Commercialization of phthalic anhydride (PA) production goes back to 1872
when BASF developed the naphthalene oxidation process (Fig. 4.81). Details of
physical and chemical properties of phthalic anhydride are presented in Lorz
et al. [59].
The most important derivatives of phthalic anhydride, which is produced glob-
ally with a capacity ~4.5 million tons per year, are phthalic esters which can be
 4.7 Oxidation 495

Vent
Dry
filter

Maleic
absorption column anhydride

Light ends column

Purification column
Maleic anhydride

stripping

Fresh solvent Purification

of the solvent
Reactor

Steam Catalyst
n-Butane make-up

Air

Fig. 4.80: Fluidized bed process for oxidation of butane in maleic anhydride with solvent recovery
(modified after 58).

+ 4.5 O2 O + 2 CO2 + H2O

+ 12 O2 10 CO2 + 4 H2O

Fig. 4.81: Oxidation of naphthalene to PA.

used as plasticizers, polyester resins and dyes. In particular the diester, dioctyl
phthalate (DOP) has a widespread use, improving flexibility of polyvinyl chloride,
which otherwise is a rather rigid material. Until the early 1960s coal tar naphtha-
lene was the dominant feedstock, while later on gas-phase oxidation of o-xylene
496 4 Engineering technology

over vanadium-based catalysts gained industrial importance and almost completely

replaced oxidation of naphthalene.
In industry, oxidation of o-xylene is done in fixed-bed tubular reactors
(15,000–30,000 tubes) at 300–400 °C and 0.5–1.8 vol% of xylene in air.
One of the reasons for the feedstock switch was o-xylene availability from pe-
troleum, while quantities of naphthalene derived from coal tar were dependent on
the production of coke and were not sufficient for the growing PA and plasticizers
markets. The other reason is atom efficiency. As can be seen from Fig. 4.81, naph-
thalene oxidation inevitably results in the consumption of two carbons and the gen-
eration of two moles of carbon dioxide per each mole of product formed. Oxidation
of o-xylene does not lead to CO2 release in the main reaction, while it could be cer-
tainly formed as a side product (Fig. 4.82).

CH3 O
CH3
+ 3 O2 O + 3 H2O

O
CH3
CH3
+ 10.5 O2 8 CO2 + 5 H2O

Fig. 4.82: Oxidation of o-xylene to phthalic anhydride.

In addition to feedstock availability and unnecessary CO2 emissions, in some areas

and in particular in cold climates, handling of naphthalene, which is solid at room
temperature, was challenging.
Although more than 90% of PA is nowadays produced from o-xylene there are
still a few plants operating with naphthalene that are mainly located close to major
steel mills. Naphthalene is much cheaper than o-xylene as it is a by-product of coke
production. Few installations utilize mixed o-xylene-naphthalene feeds.

4.7.3.2.2 Thermodynamics
The oxidation reactions are not limited by equilibrium and are highly exothermic.
High activation energy means that even with a small temperature increase, heat pro-
duction increases substantially, leading to prominent hot spots in fixed-bed reactors.
The reaction enthalpy of the o-xylene oxidation is −1,108.7 kJ mol−1 at 298 °C while it
is −4,380 kJ mol−1 for complete oxidation. As overall selectivity in o-xylene oxidation
 4.7 Oxidation 497

is ~80% at almost complete conversion, heat generation in this reaction is significant,

reaching 1,300–1,800 kJ per mole of o-xylene. Heat removal is thus essential. As men-
tioned above, oxidation is performed in fixed-bed tubular reactors. Another option
would be to use fluidized bed reactors, which are very efficient in controlling temper-
ature, maintaining a uniform temperature throughout the bed. Such option was in
fact implemented in only few places for naphthalene oxidation, where V2O5–K2SO4
on silica was applied, allowing good fluidization properties. As such support does
not allow high yields of phthalic anhydride in o-xylene oxidation there are appar-
ently no industrial o-xylene- based fluidized bed reactors currently in operation, al-
though development work in companies and academia might still continue.
It should be noted that the feedstock as well the product can be explosive in
mixtures with air. Thus, a special care should be taken to ensure that the feedstock
and product are not within explosion limits with air. It can be seen from Fig. 4.83
illustrating general principles of reactions which can lead to explosions that at a
certain temperature, say 500 °C, at low pressures the whole system is below the
first explosion limit. When the concentration of hydrocarbons increases (partial
pressure increases) explosions can occur, while with a further partial pressure in-
crease safe operation is once again possible.

3rd
lim it of
exp
losi
on Explosion

ion
plos
Pressure

f ex
it o
Stationary lim
d
reaction 2n
Explosion

1st limit
o f explos
ion

460 °C 600 °C
Temperature

Fig. 4.83: Examples of reactions with explosions.

At one of the PA plants that I visited very often while working in industry, there were ignitions
observed with the reaction mixture when the catalyst was not able to convert o-xylene efficiently
during reactor start-up. At the outlet of the reactor there is a gas cooler, with surface-to-volume
ratio different from that of the pipe connecting the reactor with the gas cooler. The length of the
gas cooler is ~6 m and I could see from the outside that the color of the gas cooler at least for 1m
498 4 Engineering technology

from the inlet was somewhat reddish because of ignition of the unconverted reactant. These ob-
servations strongly indicated the presence of branched chain reactions. Once the catalyst was
back to conventional operation, such ignition terminated.
In another PA plant with naphthalene as a feedstock I was shown a production site with five
reactors, where at some point there were six reactors. The reactor had completely disintegrated
and the reactor upper hood was never found as the reaction mixture ignited, most probably
caused by heat transfer salts entering the reactor as the reactor tubes were broken.

The lower explosion limit for o-xylene or naphthalene in air is 47 g m−3 (STP).
Compression and preheating of air represent additional costs, thus the increase of o-
xylene loadings above this limit was addressed throughout the years. Many plants
operate now at 60, 80 or even 100 g m−3, as it was shown that an explosion starting
in the catalyst bed is quenched by the catalyst heat capacity. Extra thickness in the
upper part of the reactor combined with adequate rupture disks is implemented to
avoid explosions in the free space above the catalyst bed. The probability of phthalic
anhydride auto-ignition is minimal because the PA auto-ignition temperature (580 °C)
is well above typical operation temperatures (salt bath temperature 345–350 °C).

4.7.3.2.3 Mechanism
Fig. 4.84 summarizes the important by-products formed in the oxidation of o-xy-
lene. The reaction requires breaking of two and formation of 12 bonds, thus a de-
tailed analysis of all reaction pathways is challenging.

O
O
10 HO
O 9 HO
COx PAId
O
CO2, CO 6 O PAc
O

2 1 5 O 7 PH O
TA
HO

13 11
TAc O
O O
oX 4 8
14

12 CA

BAc O
OH
O
3 O O
O
15
MA PA O

Fig. 4.84: Oxidation of o-xylene. (From [60], reproduced with permission).

 4.7 Oxidation 499

The major intermediates in the selective oxidation path are tolualdehyde (TA)
and phthalide (PH), while such intermediates as toluic (TAc), phthalic (PA) acids,
and phthalaldehyde (PAld) were also proposed. A non-selective route, i.e., a route
that does not result in PA, includes the formation of benzoic (BA) acid, citraconic
aldehyde and maleic anhydride (MA) and COx (CO and CO2).
The composition of the product gas is presented in Tab. 4.17.

Tab. 4.17: Main by-products in the reaction gas for the latest generation catalysts (PA yield
94.5–95.6%).

MA BA CA PH TA

.–. wt% .–. wt% .–. wt% .–. wt% .–. wt%

The by-product formation is crucial not only in terms of product selectivity. In some
instances excessive presence of by-products deteriorates the product quality, in par-
ticular the Hazen number, which reflects the PA color, should be rather low in
some applications, between 5 and 10.
In addition to the main by-products, given in Tab. 4.17, other carbon-containing
by-products are carbon monoxide and carbon dioxide. In general, the reactor yields
do not exceed 112–115 kg PA/100 kg o-xylene, which corresponds to ~82% of stoi-
chiometric yield. Taking into account losses during condensation and distillation,
the yields of pure PA are about 112 kg PA/100 kg o-xylene.
For the sake of comparison a few key numbers are also given here for oxidation
of naphthalene. The heat of reaction is 1,788 kJ mol−1, while in complete combus-
tion 5,050 kJ mol−1 is evolved. The main by-products are maleic anhydride and
naphthoquinone (0.05–1.0 wt%) and the PA yield at the reactor outlet for the latest
generation catalysts reaches 105–106 kg kg−1 naphthalene.

4.7.3.2.4 Catalyst
Oxidation reactions with either of the feedstock (o-xylene or naphthalene) are car-
ried out in tubular reactors cooled by a molten salt. Ring types of catalysts are ap-
plied, which replaced spherical catalysts used in the old plants.
For naphthalene partial oxidation V2O5 promoted by K2SO4 was applied on a
moderately porous silica carrier. For o-xylene oxidation porous supports were
found to be detrimental and largely non-porous carriers, such as silicon carbide,
quartz, steatite ceramics or porcelain with dimensions such as 8 × 6 × 5 mm or
7 × 7 × 4 mm and filling density 0.88–0.93 kg l−1 are used. Low area of inert ceramic
supports and coating of them with vanadia/titania in the form of a thin layer (“egg-
shell” catalyst) with a thickness of up to 100 μm is thus needed to avoid internal
500 4 Engineering technology

mass transfer limitations, which would otherwise diminish catalyst activity and se-
lectivity to phthalic anhydride.
Finely divided titanium dioxide (anatase) and vanadium pentoxide oxide layer
together with promoters (antimony oxide, rubidium oxide, cesium oxide, niobium
and phosphorous) are introduced on the support by, for example, spray-drying. In
the monolayer formed on the surface there are chemical bonds between V ion and
Ti surface ions via oxo bridges. The layer is a two-dimensional sheet of the formula
VO2.5, which is believed to contain both strongly bound tetravalent and pentavalent
vanadium. Both of the valence states take part in the oxidation of o-xylene by a typ-
ical redox mechanism. Titania is considered to inhibit desorption of the many inter-
mediates involved in the overall reaction.
Formation of bulk (crystalline or amorphous) vanadia, if it is introduced in ex-
cess, might have a negative impact on catalysts, thus up to 8–10% of vanadium
pentoxide in combination with titanium dioxide is used. Details of the catalysts
properties and synthesis were discussed by Cavani et al. [61].
In the past some of the catalysts, such as a silica gel supported naphthalene oxi-
dation catalyst containing V2O5 and K2SO4, required the addition of SO2 for activation
and longer service life, an option that is not required with the modern catalysts.
In industry for low o-xylene loadings, two-zone catalysts (i.e., combination of
catalysts with low and high activity) were used, while higher o-xylene loadings re-
quire three- and four-zone catalysts (Fig. 4.85).

CL1
CL1 CL2 CL3 CL4
Selective
layers

CL2

CL3
Active
layers
CL4

(A) (B)

Fig. 4.85: Four-zone BASF O4-68 catalyst for o-xylene load of 80 g Nm–3. (From [62].

The optimal amount of the active phase varies depending in which catalyst zone of
the multitubular reactor the catalyst is located, as different reaction stages require
different catalyst compositions. It could be easily anticipated that because of high
 4.7 Oxidation 501

reaction exothermicity, the top catalyst layer should be able to minimize the tem-
perature increase in the hot spot and at the same time ensure fast start-up of cold
reaction gas. In the second layer, most of o-xylene conversion takes place and the
catalyst should have high selectivity with the lowest activity. Two final layers in the
four-zone catalyst system should be more active and less selective ensuring good
conversion of remaining underoxidation products to phthalic anhydride and com-
plete combustion of overoxidation products to CO2. The amount of cesium in differ-
ent layers is decreasing from top to bottom. Besides helping to avoid excessive hot
spots, dopants such as Cs are also used to minimize formation of maleic anhydride
and carbon oxides.
Such operation with several layers allows for minimizing the negative influence
of the hot spot, which is typically moving along the reactor length with time-on-
stream (Fig. 4.86).

2 layer catalyst 300 350 400 450

169 325
Fresh
269 360
369 387
469 407
Aged 569 419
669 429
769
869 430
n layer catalyst 969 426
Fresh 1,069 418
1,169 411
Aged 1,269 402
1,369 395.5
1,469 388.5
1,569 386.2
(A) (B)

Fig. 4.86: Temperature profile along the reactor length: (A) typical behavior, (B) experimentally
observed in the first zone of a two-zone catalyst.

BASF, one of the commercial catalyst suppliers, introduced in 2011 a five-layer cata-
lyst O4-88 (Fig. 4.87) with first three selective layers affording phthalic anhydride
yield up to 116.5 wt%. The hot spot is located in the CL0 layer. Details on the com-
position are not available, while the patent literature [63] disclosed, for example,
the pre-catalyst layer of Ag/V(O) bronze (mixed oxides of silver and vanadium with
atomic ratio Ag/V < 1) dispersed on Mg silicate rings. This formulation should not
be, however, necessarily the one used commercially.
OxyMax® phthalic anhydride catalysts from Clariant for selective oxidation of
o-xylene, naphthalene and mixed feedstocks, produced by fluid-bed coating pro-
cess, also consist of up to five catalyst layers.
502 4 Engineering technology

Fig. 4.87: Five -zone BASF O4-88 catalyst for o-xylene load up to 100 g Nm–3. (From [62]).

4.7.3.2.5 Kinetics
The complexity of the reaction prevents detailed analysis, and mostly some simpli-
fied schemes were discussed in the literature. One of such approaches is to lump all
intermediates into one pseudo-compound. The kinetic scheme applied most often
was suggested by Calderbank et al. [64]:

1 2 5
A B C D
3 6
COx

where A, B, C, D are o-xylene, o-tolualdehyde, PH and PA, respectively. Oxidation

of PA (step 6) is considered insignificant.
The following rate equations were derived:

r1 = k1 aPA, r2 = k2 aPB, r3 = k3 aPA, r4 = k4 aPA, r5 = k5 aPC, r6 = k6 aPD (4:39)

where

kc PO2
α= (4:40)
kc PO2 + PA ðm1 k1 + m3 k3 + m4 k4 Þ + PB m2 k2 + PC m5 k5 + PD m6 k6

and m is the moles of oxygen that react per mole of xylene consumed. The values of
m are 1, 6.5, 3, 1, 1 and 3.5 for steps 1–6, respectively.

4.7.3.2.6 Technology, reaction and separation

Oxidation of o-xylene is done industrially in multitubular reactors containing up to
25,000 tubes with ~2.5 cm internal diameter. In the initial design, reactors contained a
salt bath circulation pump and a fitted internal cooler. This allowed temperature
 4.7 Oxidation 503

control by a molten salt, which is eutectic of potassium and sodium nitrates, with a
melting point of 141 °C. At a later date external cooling was introduced. The molten
salts circulate through a waste heat boiler, which is used to generate saturated steam.
In case of industrially high o-xylene loadings used nowadays, reactors use radial salt
flow (contrary to axial flow in previous designs) for improved heat transfer (Fig. 4.88).

Steam

Water

Fig. 4.88: Reactor with radial salt flow (DWE type). (From [65], copyright Elsevier.).

The oxidation air is pre-heated to about 180 °C and charged with feedstock in con-
centrations up to 100 g Nm–3. The air rate is typically 4 Nm3 per reactor tube and
per hour.

Sometimes cleaning of air can be very important. Once while working in industry I visited a
plant, which was build using a licensed technology and was a carbon copy of another plant
also located in Europe. Unfortunately, after the plant start-up, the values of selectivity
guaranteed by the technology licenser could not be reached and the catalyst deactivated. It
turned out that troublesome plant was located at a riverbank (literary) and the air was con-
taminated with aerosol containing potassium. The plant manager proudly showed an installa-
tion with high-efficient air purification from aerosols using fibrous filters developed by
Petryanov in 1939 and used shortly after the Second World War to protect personnel of ura-
nium mining facilities from exposure to radioactivity. The self-made addition to the existing
technology of phthalic anhydride production mitigated the catalyst contamination problem
completely. After learning that a decade before this visit I was working at the same institu-
tion as Prof. Petryanov, who even attended my PhD defense, the plant manager was happy to
hear more details about the history behind discovery of the aerosol filter.

As could be seen from Fig. 4.89 no recycle is needed as high conversion levels are
achieved. Operating salt temperatures are in the range 325–425 °C. Initial start-up
temperature is ~375 °C and it takes approximately 2 weeks to stabilize the catalyst
504 4 Engineering technology

K
Air Off-gas
E

o-Xylene p
SC
R C

T
Phthalic
Overheads anhydride

ST DI
Reside

Fig. 4.89: PA production technology [66]. K, air compressor; P, o-xylene pump; E, evaporator;
R, reactor; C, salt bath cooler; SC, switch condensers; T, crude phthalic anhydride tank;
D, predecomposer; ST, stripper column; DI, distillation column.

activity and operate at 350–355 °C. When the reaction takes off a hot spot is gener-
ated in the reactor, which exceeds salt bath temperature by 50–60 °C as it is in the
range 410–430 °C. The temperature declines along the reactor length as the gener-
ated heat is related to reaction rate, which obviously declines with increased con-
version. With operation time the catalyst deactivates and the hot spot moves
downwards along the reactor. In order to compensate for the activity loss the salt
bath temperature is increased during operation to 360–365 °C at the middle of the
catalyst lifetime and to 370–380 °C at the end. Higher operation temperature results
in lower selectivity, higher phthalide yields, worsening of the product quality,
which is seen as increase of the color number, and finally lower phthalic anhydride
yields. Thus after 4–5 years of operation the catalysts have to be replaced.
Fig. 4.90 illustrates a typical behavior of o-xylene oxidation catalysts with time-
on-stream, while Fig. 4.91 displays operation with mixed o-xylene-naphthalene
feed with an equal ratio between reactants.
Loading of catalysts in industrial multitubular reactors is a time-consuming
procedure, as it should guarantee uniform flow distribution. With such multizone
catalysts loading is arranged by filling first one zone, measuring pressure drop in
all tubes, levelling off loading and proceeding with another zone.
The reactor effluent gas leaving the reactor is cooled down first in the gas cooler
(Fig. 4.89). Downstream cooler switch condensers equipped with U-type finned
tubes are applied. Such tubes are alternatively heated or cooled with a hot or cold
heat-transfer medium. By cooling the gas the crude PA desublimates on the fin tube
surface.
 4.7 Oxidation 505

70
400
390

Salt bath temperature (°C)

380
o-Xylene load (g/Nm3)

60 370
360
350
340
50
330
320
310
40 300
0 6 12 18 24 30 36 42 48 54 60 0 6 12 18 24 30 36 42 48 54 60
(A) Operation time (months) (B) Operation time (months)

117
116

Phthalide in reaction gas (wt%)

0.2
115
114
PA-yield (wt%)

113
112
0.1
111
110
109
108 0.0
0 6 12 18 24 30 36 42 48 54 60 0 6 12 18 24 30 36 42 48 54 60
(C) Operation time (months) (D) Operation time (months)

Fig. 4.90: Experience in a commercial plant with loading of o-xylene 60 g Nm–3.

110

108
PA yield (wt%)

106

104

102

100
0 40 80 120 160 200 240
Operation time (days)

Fig. 4.91: Experience in a commercial plant with mixed o-xylene-naphthalene feed.

Each switch condenser can be loaded with a certain quantity of crude phthalic
anhydride, which is defined by the surface area of switch condenser. After full load-
ing a switch condenser is isolated from the gas stream and heated up to melt the
506 4 Engineering technology

crude phthalic anhydride from the fin tubes and direct it to the purification section,
first through a crude PA tank. The heat removed from the reactor process gas during
the loading phase is absorbed by a heat transfer oil circulating through the fin
tubes. The same oil at a higher temperature is used for unloading. A scheme of a
Rollechim switch condenser is presented in Fig. 4.92.

Gas in
117 °C
Bundle 1
Oil out
75 °C
37%
Oil out
23% Bundle 2
66 °C
Oil out
Bundle 3 23%
60 °C
Oil out 57 °C
23% Bundle 4
Oil in
100%
Gas out
(A) (B)

Fig. 4.92: Switch condensers. (Figures from [67]).

As visible from this Fig. 4.92 the downflow mode of operation is applied, as the reac-
tion gas enters at the top. Such operation allows deposition of most of the solid prod-
uct on the upper most bundle. During the unloading stage the lower tube bundle is
washed by molten PA, and in this way removing highly corrosive maleic anhydride
and phthalic acid from the tubes, which was challenging in up-flow operations.
High efficiency of the switch condenser affords more than 99.5% of phthalic an-
hydride recovery and thus less than 0.5% of the crude PA escapes with the off-gas.
The off-gas leaving the switch condensers is passed through a scrubbing tower
where the major part of the organic compounds, such as maleic anhydride, is re-
moved by water scrubbing. Carbon monoxide cannot be removed by such scrub-
bers, thus either thermal or catalytic combustion is used.
The product purification is performed first by thermal pre-treatment and then
by vacuum distillation. For naphthalene oxidation the pre-treatment takes place in
a cascade of vessels heated by outside coils with heat transfer oil at temperature
230–300 °C with 10–24 h retention time. Such treatment can be applied for crude
products with low contents of impurities. Chemical treatment was also used to de-
stroy the by-product naphthoquinone.
The purification in o-xylene oxidation is intended to destroy phthalic acid by
dehydration to PA, and removal of water and low-boiling compounds such as
maleic anhydride, o-tolualdehyde and benzoic acid.
 4.7 Oxidation 507

After the predecomposer (where phthalic acid is dehydrated) phthalic anhy-

dride is introduced into the first distillation column (Fig. 4.89). It is separated in
this column from the low-boiling components, which are taken from the column
top. In the subsequent column purified phthalic anhydride is removed from the top
and is delivered to flaking, liquid storage or directly to downstream facilities, while
the residue is discharged from the bottom.
Both distillation columns operate under vacuum, which is generated by ejectors
generally driven by steam.

4.7.4 From alcohols to aldehydes: oxidative dehydrogenation of methanol

to formaldehyde

Formaldehyde is produced industrially from methanol by partial oxidation and de-

hydrogenation with air

CH3 OH ) HCHO + H2 , ΔH = + 84 kJ=mol

CH3 OH + O2 ) HCHO + H2 O, ΔH = − 159 kJ=mol

mainly in the presence of silver catalysts, steam and excess methanol at either
680–720 °C with methanol conversion approximately 99% (so-called complete con-
version) or at 550–650 °C giving incomplete conversion of methanol 70–80%. In
the latter case, the product is distilled and the unreacted methanol is recycled. The
advantage of lower temperature and incomplete conversion is in less prominent
side reactions. The reaction temperature depends on the excess of methanol in the
methanol–air mixture. The composition of the mixture must be outside the explo-
sion limits. Addition of steam is needed to diminish coking on silver catalysts.
Oxidation can be done also without steam only with excess air in the presence
of a modified iron–molybdenum–vanadium oxide catalyst at 250–400 °C with
methanol conversion of approximately 98 – 99%.
Among side reactions, oxidation of hydrogen to water, decomposition of form-
aldehyde to CO and hydrogen, oxidation of methanol and formaldehyde to CO2 and
water should be mentioned.
Another important factor affecting the yield of formaldehyde and methanol
conversion, besides the catalyst and temperature, is the addition of inert com-
pounds to the reactants. Water is added to the spent methanol–water-evaporated
feed mixtures, and nitrogen is added to air and air–off-gas mixtures, which are re-
cycled to dilute the methanol–oxygen reaction mixture.
There are several options for the process based on silver catalysts. One variant
used for many decades included incomplete conversion with a single-stage metha-
nol separation and recycle (Fig. 4.93).
508 4 Engineering technology

Off-gas

Steam Water

Steam
Ag catalyst bed
Steam (10–50mm)

Air BFW 890 K

1.5 bar Steam
Steam
Formaldehyde
solution (50 wt.%)
Methanol recycle
Methannol

Vaporizer Reactor and waste Absorber Distillation

heat boiler coloumn

Fig. 4.93: Silver-based methanol oxidation to formaldehyde [20]. Reproduced with permission.

First, a mixture of methanol and water is evaporated and sent to the adiabatic reactor
containing a shallow bed (25–30 mm) of a silver catalyst. There is a separate flow to
the vaporizer of the fresh process air. The content of the “inert” gases (nitrogen,
water and CO2) should be selected to prevent formation of explosive methanol–oxy-
gen mixtures. A typical methanol-rich feed (40–45% methanol, 20–25% air and
the rest steam) enables safe operation outside the flammability limits. The aver-
age lifetime of a catalyst bed depends on impurities such as inorganic materials
in the air and methanol feed. The catalyst bed is located immediately above a
cooler (water boiler). Temperature of the outlet gases diminishes to 150 oC and
simultaneously superheated steam is produced. Such rapid cooling is needed to
avoid oxidation of formaldehyde to CO2. In the absorption tower the process gas
is flowing countercurrently with water. Methanol is recovered at the top of the
distillation column and is recycled to the bottom of the evaporator. The product
containing up to 55 wt% formaldehyde and less than 1 wt% methanol along with
approximately 0.01 wt% formic acid is taken from the bottom of the distillation
column. A special care should be taken on preventing corrosion as formaldehyde
solutions are corrosive.
Capital and operating costs for distillation and recycling equipment needed
for recovery of unconverted methanol can be reduced by approximately 12% and
20% in the case of a complete conversion process (Fig. 4.94), which also has sev-
eral variants. The fixed-bed adiabatic reactor is, however, essentially the same as
in incomplete conversion with a 15–30 mm deep, 2–3 m diameter bed of Ag cata-
lyst (Fig. 4.95). Several mesh fractions are typically used with the finest on the top
 4.7 Oxidation 509

To off-gas
incineration

Process water
Absorber
dilution water

Steam
distributor

Reactor
Superheater Gas filter
Water pump and flame arrestor

Waste heat
boiler Absorber Absorber
ctolumn 1 column 2
Evaporator

Air blower
Methanol pump Air
Product formaldehyde solution
Methanol

Fig. 4.94: Silver-based complete methanol oxidation to formaldehyde process [68]. Reproduced
with permission.

Fine silver

Medium silver

Coarse silver

Fine mesh

Coarse mesh

Optional base plate

Waste heat boiler

Fig. 4.95: Silver catalyst (left) bed configuration, center: fresh bed, right: in operation [68].
Reproduced with permission.

and the largest granules on the bottom of the bed. The bed is arranged to prevent
the finest silver grains from slipping through the structure.
The feed comprising methanol, air, steam and tail-gas recycle is introduced to
the reactor at 340–350 °C and approximately 1–1.5 bar. Safety considerations are
important in methanol oxidation process. Addition of steam helps to increase selec-
tivity and mitigates catalyst deactivation by lowering coke formation and sintering.
510 4 Engineering technology

There is, however, a limit on the amount of steam, because the target formaldehyde
product strength of typically 55% should be achieved, and moreover, additional
amounts of water increase the downstream treatment and recycling costs.
The residence time is approximately 0.01 s to minimize decomposition of form-
aldehyde. Methanol conversion is 98.5–99% with selectivity to formaldehyde of
90–92% with the rest of products being CO, CO2 and small amounts of formic acid.
Rapid cooling of formaldehyde gas is done using a waste heat boiler system. The
exit gas stream leaves at temperature between 120 and 200 °C, and then passes
through the first absorption tower. Water containing recaptured methanol and
formaldehyde from the second absorption column is sent to the reactor, enhancing
the overall product yield.
The lifetime of the catalyst life is from few months to a year depending on the im-
purities in methanol and air, operation temperature and the plant capacity. Sintering
of silver at high reaction temperatures results in pressure drop elevation. As typical
with many processes where catalyst deactivates at some point, a decision should be
taken to replace the catalyst. Such decision is based on evaluation of the catalyst costs
versus lower productivity, higher operations costs and safety considerations.
Besides silver-based technology an alternative process employs a metal oxide
catalyst.
The catalyst composition in the latter so-called Formox process is rather com-
plex; for example, it can contain iron and molybdenum oxides Fe2(MoO4)3 with an
excess of MoO3 giving Mo:Fe atomic ratio of 1.5–2.0. Iron molybdate structure is sta-
bilized by such additives as V2O5, CuO, Cr2O3, CoO and P2O5. The catalyst lifetime is
typically 18–24 months because the catalyst is more tolerant to poisons.
An excess of air is used in this process with approximately 13:1 air to methanol
molar ratio, thus the concentration of methanol is 6–9%. Too high amounts of meth-
anol would lead to catalyst reduction and subsequent deactivation. Proper heat man-
agement avoiding overheating is needed to prevent sintering. Methanol conversion is
done at atmospheric pressure and much lower temperature (270–400 oC, a typical
value 340 °C), compared to silver-based process. Typical single-pass methanol con-
version is 97–99% with formaldehyde selectivity of 92–95%. Lower temperature is
beneficial from the viewpoint of higher selectivity. All these in addition to a simple
method of steam generation contribute to the advantages of the Formox process. At
the same time, the volume of gas is 3–3.5 higher than in the silver-based technology
resulting in more expensive larger capacity equipment.
Figure 4.96 displays the flow scheme for the Formox process. Methanol is fed
through a steam-heated evaporator and after mixing with the fresh process air
and the recycled off-gas passes through a heat exchanger heated by the product
stream.
The catalyst pellets or rings (Fig. 4.97) of the size 3–4 mm are placed in multi-
tubular reactors of a diameter approximately 2.5 m containing thousands of tubes
(diameter 2–3 cm and 1.0–1.5 m length). The reaction heat is removed by a
 4.7 Oxidation 511

Recycle
Vent
H2O

Absorption tower
Colling
water
Reactor

Air

Methanol
Formaldehyde
solution

Fig. 4.96: Synthesis of formaldehyde by oxidation on Fe–Mo catalyst [69].

Fig. 4.97: Fe–Mo catalysts of different shapes (a) [70]; (b) [71] for synthesis of formaldehyde by
methanol oxidation.

high-boiling heat-transfer fluid (oil, Dowtherm, molten salt) held at approxi-

mately 270–290 °C. This fluid is circulating outside of the tubes generating steam
in a boiler.
A typical feature of highly exothermal reactions in multitubular reactors are hot
spots, which in the case of methanol oxidative dehydrogenation on iron molybdate
reaches 340–350 °C.
The outlet gases are cooled to 110 °C to avoid formaldehyde and sent to the bot-
tom of an absorber column to which the process water is added at the top. The
product containing up to 55 wt% formaldehyde and 0.5–1.5 wt% methanol is fur-
ther purified by ion exchange to diminish the formic acid content. The overall plant
yield is 88–91 mol%. The absorber off gas (N2, O2, CO2 with few percent of dimethyl
ether, carbon monoxide, methanol and formaldehyde) is noncombustible as such
512 4 Engineering technology

and requires catalytic incineration at 450–550 oC or alternatively combustion with

additional fuel at 700–900oC.
Both types of plans using either silver or iron catalysts despite operation at very
different conditions give similar values of methanol conversion and formaldehyde
yields. According to available comparative analysis of both processes [72] if cheap
methanol is available, silver can be considered a better option, especially for
smaller operations. For larger plants with capacity of up to 100,000 tons per year
the metal oxide catalysts can be more suitable compensating a higher capital ex-
penditure related to fivefold higher air flow and subsequently higher compression
costs. Plants using more productive silver catalysts may present lower operating
costs, lower catalyst costs as silver can be regenerated contrary to only Mo from
iron molybdate and lower steam production. The tail gas in the silver process con-
tains about 20% hydrogen and can be burned directly eliminating carbon monoxide
and other environmentally harmful organic compounds. In an alternative iron
oxide process, burning of the tail gas with too low amounts of flammable com-
pounds requires a catalytic incinerator or addition of extra fuel, which adds to the
production costs. Metal oxide catalyst is more resistant to poisons and because of
low temperature is less susceptible to thermal degradation giving a better quality
formaldehyde product with less impurities, such as formic acid, heavy metals and
unreacted methanol. A further advantage of the process using iron oxide with a
substantially longer catalyst life are less downtime problems.

4.7.5 Ammoxidation

4.7.5.1 Acrylic acid

Acrylic acid is produced by air-based heterogeneous catalytic oxidation of propene
in the gas phase using multicomponent mixed metal oxide catalysts. The reaction is
occurring through two exothermic steps with first formation of acrolein and water

CH2 = CH − CH3 + O2 ! CH2 = CH − CH = O + H2 O, ΔH = − 340.8 kJ=mol

and subsequent oxidation of acrolein to acrylic acid

CH2 = CH − CHO + 0.5O2 ! CH2 = CH − COOH, ΔH = − 254.1 kJ=mol

These two steps require different reaction conditions and catalysts to produce opti-
mum conversion and selectivity in each step. For the first step, supported multicom-
ponent systems containing Bi-Mo-Fe-Co-Sb-K mixed oxides are used. Acrolein yields
depend not only on the chemical compositions of these catalysts, but also on their
physical properties, such as shape, porosity, pore-size distribution and specific sur-
face area, as well as on the reaction conditions and construction of the reactor.
There was a significant improvement in the yield of the product because of the
catalyst development. In the 1950s, the yield with Cu2O catalyst was 10–20%, which
 4.7 Oxidation 513

was elevated to 45–50% in 1960s on Bi-Mo-P containing catalyst. Bi-Mo-Fe-Co

mixed oxides allowed a further increase of the yield to 80–85%. A further improve-
ment was obtained with Bi-Mo-Fe-Co-Sb-K mixed oxides.
The catalysts for oxidation of propene to acrylic acid are essentially the same as
applied in propylene ammoxidation (Section 4.7.5.2) when oxidation is done in the
presence of ammonia. In the latter case, the overall yield of acrylonitrile is lower
than the combined yield of acrolein and acrylic acid. Both reactions share the same
mechanistic features. Oxidation of propylene to acrolein (the first step in formation
of acrylic acid) follows Mars-van-Krevelen mechanism with a number of steps, each
requiring their own active sites. Therefore, multicomponent catalysts are applied
and catalytic performance strongly depends on the catalyst structure, which is dis-
cussed in more detail in Section 4.7.5.2.
The process flow diagram of the oxidation section of acrylic acid production is
shown in Fig. 4.98.

Waste-gas
incinerator

Water

Steam
Air
Propene

To purification
section
First- Waste- Second- Waste- Absorbing
stage heat stage heat column
reactor boiler reactor boiler

Fig. 4.98: Schematic diagram of the oxidation section in acrylic acid production.

Air, propylene and steam are fed to reactors consisting of catalyst-filled tubes with
molten salt circulating on the shell side for removal of the reaction heat. The salt
passes through a steam boiler to produce steam for heat recovery. The gas composi-
tion is 5–10% propylene, 30–40% steam with air being the rest. An excess of air is
used to keep the catalyst in the oxidized state. Steam can be replaced by inert gas.
Other reaction parameters: temperature 300–400 oC, contact time 1.5–3.5 s, and
inlet pressure: 150–250 kPa. With time the yield of the product is decreasing and/or
pressure drop is increasing to a point when a catalyst charge is replaced. Such re-
placement can happen after a relatively long catalyst lifetime (up to 10 years)
514 4 Engineering technology

At high propene conversion (up to 98%), acrolein is formed in the yields of

83–90%. The main by-product is acrylic acid (5–10% yields) with acetaldehyde
and acetic acid being other by-products. The reactor effluent is rapidly quenched
at the exit to prevent subsequent reactions of acrolein. The catalysts used in the
second step (multicomponent metal-oxide catalysts containing Mo, V, W and
some other elements) require reaction temperatures from 200 to 300 oC and con-
tact times from 1 to 3 s. They give almost 100% conversion of acrolein and yields
of acrylic acid larger than 90%. The effluent gas from the second-stage multitube
reactor in Fig. 4.98 is cooled to about 200 °C and then fed to the absorbing column
to be scrubbed with water. Because the effluent gas contains a large amount of
steam, acrylic acid is usually obtained as an aqueous solution of 20–70 wt%. The
waste gas (nitrogen, excess oxygen, CO2 and propylene) along with some residual
organic compounds should be incinerated.

4.7.5.2 Acrylonitrile
The worldwide capacity for acrylonitrile (CH2 = CH–CN), an important monomer for
the manufacture of useful plastics, is currently approximately 7 million metric tons
per year. The average capacity of acrylonitrile plants is around 170,000 tons/year.
The main application of acrylonitrile is in production of acrylic fibers, acrylonitrile-
butadiene-styrene, adiponitrile, nitrile–butadiene copolymers as well as acrylamide
for water treatment polymers.
Prior to 1960s the most popular process of acrylonitrile production was addition
of hydrogen cyanide to acetylene. Currently manufacture of acrylonitrile mainly fol-
lows the Standard Oil of Ohio (SOHIO) process, based on ammoxidation of propyl-
ene in fluidized bed reactors using propylene, ammonia and air as reactants and
giving as by-products acetonitrile and hydrogen cyanide

CH2 = CH − CH3 + 3NH3 + 3O2 ! 3HCN + 6H2 O

2CH2 = CH − CH3 + 3NH3 + 3=2O2 ! 3CH3 − CN + 3H2 O

in commercially attractive quantities (ca. 25% total yield). The side reactions are fa-
vored at higher temperature and pressure, and longer residence time. The amounts
of HCN are larger than that of acetonitrile. Besides these products carbon dioxide is
also formed. The complex reaction network in the synthesis of acrylonitrile is
shown in Fig. 4.99.
The overall process has, nevertheless, high selectivity to acrylonitrile not re-
quiring further recycling steps. The costs of acrylonitrile production are closely
linked to the market price of reagents, in particular propylene, which contributed
substantially (> 70%) to the total cost.
A schematic view on the reaction mechanism is shown in Fig. 4.100.
 4.7 Oxidation 515

+O2; +NH3 +O2; +NH3

CH2=CH–CH3
+O2 +O2
+O2
CH2=CH–CHO CH3–CHO + HCHO
+NH3 CO2 +NH3 +NH3
CH2=CHCN CH3CN + HCN

Fig. 4.99: Reaction network in acrylonitrile synthesis.

H
O NH Diimide Complex
Bi Mo
O O
O NH
O Bi NH
Mo
O O
Slow O NH
α-H abstraction Propene adsorption 2 H2O

OH NH
.Bi Mo
O O H O Bi Mo O
O NH C
H2C CH3 O O O O
NH3
π-allyl complex
H
OH . N XH2 [O] H2O
.Bi Mo ....
O O H
O NH Ammoxidation
Water desorption
(X=O, NH)
Nitrogen insertion
OH . N
.Bi Mo OH ..
.Bi . Mo
..

O O H
O O O OH
O
[O] OH . N
H abstraction .Bi Mo
O O OH
O Acrylonitrile desorption
H
C CN
H 2C

Fig. 4.100: Ammoxidation mechanism [73]. Reproduced with permission.

Ammonia interacts with the bifunctional active centers, generating an ex-

tended ammoxidation site containing ammonia as = NH. On this site propylene is
inserted by α-hydrogen abstraction forming an allylic complex. Further nitrogen
insertion and proton abstraction gives acrylonitrile, which then desorbs from the
surface.
Regeneration of the reduced surface takes place by oxygen (O2–) from the lattice
oxygen, which is filled with oxygen from the gas phase. An ammoxidation catalyst
thus possesses multifunctional and redox properties. As mentioned earlier, ammox-
idation catalysts can be used also for synthesis of acrolein, even if in the presence
of ammonia the yields of acrolein are small. An explanation for this is that surface
516 4 Engineering technology

species (e.g., π-allyl complex in Fig. 4.100) which are precursors for acrolein react
in a fast way to acrylonitrile. Thus, the rate of acrylonitrile formation is higher than
that of acrolein.
The bismuth phosphomolybdate catalysts supported on silica are the dominant
catalysts used commercially with such promoters as K, Cs, Mg, Mn, Co, Ni, Fe, W
and Re. Historically to the mixed metal oxide Fe-Bi-Mo-O, first divalent cobalt and
nickel molybdates were added improving activity and selectivity, followed by intro-
duction of alkali metal promoters [74].
The catalyst should fulfil several functions, such as α-H abstraction (provided
by Bi3+, Sb3+ or Te4+ in the composition), olefin chemisorption and nitrogen inser-
tion (Mo6+, Sb5+), while a redox couple (Fe2+/Fe3+ or Ce3+/Ce4+) enhances the trans-
fer of lattice oxygen between the bulk of the catalyst and its surface.
In the chemical composition of catalysts, multimetal molybdates are used com-
prising bismuth molybdate as α-Bi2Mo3O12 phase, divalent metal (Ni, Co, Mg, Fe)
molybdates M2+MoO4 of α and β structure and trivalent Fe2Mo3O12 dispersed in sil-
ica (50% w/w). The catalysts also contain antimonate with the rutile structure,
made up of four metal antimonate cations and a redox couple of Fe, Ce, U, and Cr.
As mentioned earlier, alkali metals (e.g., Rb, K, Cs) are added to the active phase
along with other promoters (P, B, Mn). An example of a complex ammoxidation cat-
alyst chemical composition of the active phase is given below: (K, Cs)0.1(Ni, Mg,
Mn)7.5 (Fe,Cr)2.3Bi0.5Mo12Ox.
A coprecipitation procedure with aqueous solutions of bismuth nitrate and am-
monium molybdate is used in catalyst preparation, followed by heat treatment of
the dried particles at approximately 500 °C to crystallize the bismuth molybdate
phase.
Because of high reaction exothermicity (ΔHo = −515 kJ/mol), ammoxidation re-
action is conducted in a fluidized-bed reactor needed to remove the excess of heat.
Such operation imposes several restrictions on the catalyst size which should be in
the range 40–100 μm (Fig. 4.101) and mechanical properties, which should account

Fig. 4.101: Ammoxidation catalyst [71].

 4.7 Oxidation 517

for the abrasive environment in a fluidized bed reactor. The catalyst should be hard
enough and resistant to attrition. To meet these requirements, bismuth molybdate
catalyst is supported on silica.
The steady-state kinetics of propene oxidation is first order in propene and zero
order in oxygen when oxygen concentration is above stoichiometry. The apparent
activation energy for acrylonitrile formation is the same as for acrolein formation
(~83 kJ/mol); moreover, kinetic regularities are similar (first order in propene and
zero order in oxygen and ammonia) pointing out on a similar mechanism.
The SOHIO process (currently INEOS Technologies ammoxidation technology)
operates with near stoichiometric amounts of reactants in temperature range of 400–
510 °C and at 50–200 kPa (0.5 to 2 bar) giving propylene conversion above 95%, and
the yield of acrylonitrile approximately 80%. In the SOHIO process (Fig. 4.102), pro-
pylene, ammonia after evaporation and air are passed through a fluidized bed reac-
tor containing the catalyst.

Light ends
Propane Azeotrope column HCN
nitrogen

Oxalic
acid Purifying column Acrylonitrile
Water s crubber

Fluid bed
Product splitter

H2O
catalytic
reactor
1.5–3 atms
400–500 °C
H2O
Heavy ends
Azeotrope column

Purifying column

Acetonitrile

Steam

Ammonia Propylene Air

propane
feedstock
H2O Heavy ends

Fig. 4.102: Flow scheme of acrylonitrile synthesis.

The ratio between reactants is 1: (0.9 ÷1.1): (1.8 ÷ 2.4) for propylene, ammonia and
oxygen, respectively. An excess of oxygen is required to keep the catalyst in a
proper state needed for maintaining high activity and selectivity. The fluidized bed
518 4 Engineering technology

reactor is used to control temperature during this exothermal reaction (enthalpy –

670 kJ/mol). As a general comment, it can be stated that typically a reaction heat
above 500 kJ/mol should not be handled in fixed-bed multitubular reactors.
Usually the heat recovered is used to produce high-pressure steam.
Catalyst deactivation is a serious problem; however, by stabilization of the Mo
phase, addition of excess of MoO3 phase during preparation and periodic addition
of the catalyst during operation it is possible to extend the catalyst lifetime to 10
years. Fine particles (Fig. 4.101) are required for proper fluidization and the optimal
product yield. The contact time is 6–10 s.
The internal heat transfer coils in the reactor are used as baffles to improve fluidi-
zation, minimizing back-mixing. This finally leads to higher selectivity. An expansion
chamber at the reactor top is used for separation of smaller and larger catalyst par-
ticles, the former ones are returned through the cyclones to the reactor inlet. The re-
actants pass through the reactor only once. A heat exchanger located downstream
the reactor is used for generation of medium pressure steam. Thereafter, the reactants
are quenched, and unreacted ammonia is neutralized in water to remove remaining
ammonia. The product mixture containing water and organic compounds is split into
two phases in the product splitter. Acrylonitrile, acetonitrile and hydrogen cyanide
are separated by multiple distillations. The column where HCN is removed operates
under slight vacuum to prevent release of toxic HCN to atmosphere.
The diameter of the reactor for acrylonitrile synthesis is approximately 9–10 m
(Fig. 4.103) and the height of the fluidized bed is 6500 mm.

4.8 Oxychlorination

4.8.1 Overview

Oxychlorination of ethylene to dichloroethylene over CuCl2 catalyst (ΔH = −295 kJ/

mol) followed by thermal cracking of the latter

!
+ 2HCL + 0.5O2

CH2 = CH2
− H2 O
CH2 Cl − CH2 Cl
! CH = CH Cl + HCl
2 2

is a key step in synthesis of vinyl chloride, a monomer for the second largest plastic –
polyvinyl chloride. Oxychlorination reaction is strongly exothermic, and a careful
temperature control is essential both to ensure high selectivity and prevent rapid cat-
alyst deactivation.
In the so-called balanced process (Fig. 4.104), ethylene oxychlorination and di-
chloroethylene pyrolysis are coupled along with direct chlorination of ethylene in a
single process to increase the throughput of vinyl chloride monomer (often abbrevi-
ated as VCM). In this way there is no net consumption or production of HCl. Vinyl
 4.8 Oxychlorination 519

Fig. 4.103: Fluidized bed reactor for ammoxidation.

Cl 2
Direct
chlorination EDC recycle

C2H4 EDC Thermal VCM

purification cracking purification VCM

Waste
Oxy- O2(air) recovery
O2(air) chlorination HCl recycle

Fig. 4.104: Balanced process of VCM manufacturing.

520 4 Engineering technology

chloride is made only by thermal cracking of 1,2-dichloroethane, formed by either

direct chlorination or oxychlorination of ethylene.

The first exposure of the author to industrial chlorination was almost 35 years ago as a trainee at
a chemical plant in a remote location where a range of pesticides were produced using various
chlorination reactions. Apart from personal memories, related to daytime temperature of −30 °C
to −35 oC for a whole month of traineeship and accommodation in a factory dormitory with a very
noisy life 24/7 as neighbors were working in shifts, the author remembers that wearing a gas
mask was obligatory in the chlorination section because of green clouds of chlorine. This was
certainly not that sort of green chemistry, which one would like to use industrially.

4.8.2 Catalysts and reactors

In industry the reaction is carried out in the vapor phase in either fluidized bed reac-
tors (Fig. 4.104a) or fixed-bed reactors (Fig. 4.105b) with oxygen being supplied as
pure gas (oxygen-based process) or as conventional air (air-based process). Fixed-
bed oxychlorination generally operates at 230–300 °C and 150–1,400 kPa, while
lower temperatures can be used in fluidized bed reactors (220–245 °C) operating at
150–500 kPa. More efficient heat removal can be achieved in fluidized bed reactors.

EDC + by-products Steam

Ethylene
HCI, oxygen

RO
Boiling
Water

TIC Catalyst

Ethylene HCI

Ait/oxygen

EDC and
by-products

Fig. 4.105: Reactors for ethylene oxychlorination [75].

As catalysts, copper chloride as an active phase with promoters on porous supports

such as alumina (e.g. 8–16 wt% CuCl2/γAl2O3) is applied. It is almost generally ac-
cepted that the mechanism of oxychlorination involves a three-step redox process:
(i) reduction of CuCl2 to CuCl with simultaneous chlorination of ethylene
 4.8 Oxychlorination 521

CH2 = CH2 + 2CuCl2 ! ClCH2 − CH2 Cl + Cu2 Cl2

(ii) oxidation of CuCl to give an oxychloride and (iii) closure of the catalytic circle
by rechlorination with HCl, restoring the original CuCl2. The last two steps give the
overall reaction

Cu2 Cl2 + 2HCl + 0.5O2 ! 2CuCl2 + H2 O

The active site probably involves an isolated CuxCly complex, which is anchored to
the high-surface-area γ-Al2O3 support.
A particular challenge with copper chloride is its volatility at reaction tempera-
ture, which can be decreased by adding potassium chloride. The latter results, how-
ever, in lower activity, improving at the same time selectivity by decreasing
formation of ethyl chloride. The control over copper chloride content is needed as
an excess would lead to catalyst caking. KCuCl3 and K2CuCl4 have low melting
points and the eutectic mixture with copper chloride melts at 150 °C. Other alkali
metals and rare earth chlorides (e.g., MgCl2 or LaCl3) can be also added to the cata-
lyst formulation as promoters or to inhibit by-product formation. Their role is also
in maintaining a high Cu2+ concentration, as it is critical for high activity, selectivity
and stability.
Catalysts are usually prepared on a large scale by impregnating a suitable alu-
mina support with aqueous solutions of cupric and potassium chlorides. In the clas-
sical method the solution containing cupric chloride and other salts is added to the
support followed by holding for 30 min and drying of the catalyst at approximately
110 °C for 16–20 h in dry air. Even if good distribution of the active phase is
achieved during impregnation, drying results in nonuniform crystallization and for-
mation of the catalytically active phase with poor dispersion. This in turn nega-
tively influences catalytic performance. Alternative methods to achieve better
dispersion include impregnation during fluidization or by support immersing in a
melt of cupric-potassium chlorides, application of nonaqueous solutions, dry im-
pregnation with copper chloride or with a salt of copper, which can be decomposed
followed by annealing with HCl and oxygen.
In fluidized bed reactors microspheroidal catalyst powder with a particle size of
40–60 µm is used (Fig. 4.106a,b), while in fixed-bed reactors loaded with catalyst
pellets a much larger size of catalyst particles in the mm range should be utilized
(Fig. 4.106c) to avoid too high pressure drop.
During operation because of attrition the average size is decreasing which is
compensated by adding the fresh catalyst. As an example, a vinyl chloride plant
with a capacity 300,000 tons per year with 50 tons of catalyst hold-up requires 15
tons per of catalyst to compensate for the losses. For adequate fluidization the frac-
tion of the particles below 40 µm should be 25–35%. The copper content does not
practically change.
522 4 Engineering technology

(a) (b) (c)

Fig. 4.106: Ethylene oxychlorination catalysts for fluidized bed reactors from different suppliers (a)
[76] and (b) [75] and (c) for fixed bed applications [75].

One of the main process problems with oxychlorination is a possible tendency

for the catalyst particles to cake or to stick to the cooling bundles within the reactor.
This results in poor temperature control, poor fluidization of agglomerated par-
ticles, local overheating and a loss in selectivity. Erosion in the industrial reactor
walls leads to a release of iron, whose penetration into the catalyst worsens selec-
tivity by increasing the yield of the total oxidation products.
Temperature control for a very exothermic reaction is easier to organize in fluid-
ized beds, compared to fixed-bed tubular reactors. Higher temperature besides pro-
moting dehydrochlorination of ethylene dichloride to vinyl chloride and formation
of other chlorinated products also leads to increased oxidation of ethylene, elevated
formation of coke and more losses of copper by sublimation.
In fluidized bed reactors (Fig. 4.105a) the heat is controlled by adding cold reac-
tants and having an internal heat exchanger generating steam. Due to high gas
flow rates, industrial fluidized bed reactors often operate in the turbulent fluidiza-
tion regime. These reactors despite periodical makeup of the catalyst to compensate
for losses in cyclones and mechanical degradation allow effective reaction heat re-
moval, near isothermal operations, negligible impact of external mass- and heat-
transfer limitations, as well as minimal internal mass transport limitations. The lat-
ter is important because, as discussed in Chapter 3, operation in the internal mass
transfer regime negatively influences selectivity to the intermediate product.
The reaction conditions for oxychlorination, which are optimum from the view-
point of the ratio between reactants, require operation within the explosive limits.
An appropriate design of fluidized bed reactors with certain measures against radi-
cal chain propagation is thus needed. Special devices for mixing reactants to ex-
clude formation of large volumes of explosive mixtures outside the fluidized bed,
allow safe operation and low by-products selectivity making strict feed control less
 4.8 Oxychlorination 523

critical. Besides adequate mixing of reactants, conditions for fluidization should be

properly ensured, which impose limitations not only on the flow rate and the cata-
lyst particles size, but also on the reactor diameter and thus production capacity.
Clear advantages of fluidized bed reactors are their isothermal operation and a
possibility to compensate the catalyst losses. This allows to maintain constant cata-
lyst activity, and therefore a stable reactor performance, product quality and feed
consumption. The drawbacks include corrosion of the reactor internals needed for
heat removal and lower conversion compared to the tubular reactor.
Tubular reactors (Fig. 4.105b) allow higher conversion, especially HCl exceed-
ing 99%. Heat is removed by evaporation of steam condensate; thus, a uniform tem-
perature profile is difficult to achieve. Formation of hot spots with too high
temperature can be mitigated by diluting the catalyst with inert diluents of different
types and materials (Fig. 4.107) and/or by varying the copper chloride content in
the catalyst.

(a) (b)

Fig. 4.107: Inert materials for dilution of catalyst fixed beds in ethylene oxychlorination (a)
standard and high performance diluents from INEOS [75], (b) graphite pellet 5 × 5 mm and alumina
ring 6 × 6 × 2.5 mm diluents from BASF [77].

Several options of catalyst loading in tubular reactors can be used. For instance, in
a single reactor up to four different layers of a catalyst containing the same amount
of CuCl2 (8.5 wt%) can be loaded with different degree of dilution. The first layer
contains 7% of the catalyst, followed by 15%, 40% and finally 100% of the catalyst
without any dilution.
Mixing of the catalyst and the inert material should be properly done to avoid
any dust and local overheating, minimize nonuniform pressure drop along the reac-
tor and ensure good performance in terms of catalysts life and selectivity. This was
in particular important when silica particles with irregular shape were used as dilu-
tants. Mixing can be done by a catalyst manufacturer, which delivers catalyst mix-
tures for direct loading in the reactor.
A better control with a diluted catalyst can be obtained in a series of several
separate reactors, which could be further improved by having different catalyst
compositions with varying amounts of CuCl2. Spherical catalysts supplied previ-
ously by Stauffer Chemical Company did not require dilution with an inert. Dilution
of catalysts containing 6%, 10%, 18% CuCl2, and promoted with 3%, 3%, and 2%
524 4 Engineering technology

KCl with inerts was also reported in the first two beds, while the last bed contained
only the more concentrated catalyst [3].
Conditions in different reactors are different. Obviously, they are more severe in
the first reactor, thus the catalyst life is shorter (one year) and can be increased to
18 months and 3 years, respectively, in the second and third reactors. When several
types of catalysts are applied hot spots are present close to the position in tubes
where the catalyst type changes.
In tubular reactors, which can operate with air or oxygen, control of the oxygen
content is required to ensure operation outside of explosive conditions. Another op-
tion is to split air between several beds. Half of the overall amount of the air can be
added to the first reactor, while the rest is split between the second and third reac-
tors to minimize hot spots. Replacement of air with oxygen reduces the amount of
vented gas.

4.8.3 Reaction network and kinetics

The main reactions shown in Fig. 4.108 follow the route with oxygen involvement
(A) or without it (B).

(A1)
+ 2 HCI +0.5 O2 C2H4CI2
(A) ΔHr=–242 kj/mol
C2H4 (A2) (B2)
+CI2
(B) ΔH =–69 kj/mol
+ HCI r C2H5CI ΔHr=–110 kj/mol
(B1) ΔHr=+72 kj/m

Fig. 4.108: Reaction network in ethylene oxychlorination [78].

All chloroorganic by-products except ethyl chloride result from secondary 1,2-di-
chloroethane transformations, with dichloroethane dehydrochlorination and partial
oxidation being the primary reactions in such transformations. Vinyl chloride and
chloroethanol are converted further to 1,1,2-trichloroethane, trichloroacetaldehyde,
trichloromethane, tetrachloromethane, dichloroethene, CO2 and CO. Ethyl chloride
is formed by ethene hydrochlorination.
Typically, the reaction kinetics follows first order dependence on ethylene
and oxygen partial pressures, while HCl does not influence the reaction rate, af-
fecting, however, selectivity in total oxidation and also selectivity to dichloroeth-
ane, which decreases upon increasing partial pressure. The detailed kinetic
analysis of oxychlorination was performed in [79] where a following mechanism
was proposed:
 4.8 Oxychlorination 525

1. CuCl2 * + 2HCl ! CuH2 Cl4 * 2

2. CuH2 Cl4 * + C2 H4 ! CuH2 Cl4 . C2 H4 * 2
. .
3. CuH2 Cl4 C2 H4 * ! CuH2 Cl3 C2 H4 Cl* 2
4. CuH2 Cl4 * + CuH2 Cl3 . C2 H4 Cl* ! 2CuH2 Cl3 * + C2 H4 Cl2 ðfastÞ 2 (4:41)
5. CuH2 Cl3 * + O2 ! CuH2 Cl3 O2 * 1
6. CuH2 Cl3 O2 * + 3CuH2 Cl3 * ! 2CuH2 Cl4 * + 2CuCl2 * + 2H2 OðfastÞ 1
4HCl + 2C2 H4 + O2 ! 2C2 H4 Cl2 + 2H2 O

On the right-hand side, the stoichiometric (Horiuti) numbers of a particular step are
given. The following kinetic equation corresponds to this scheme
2
K1 K2 k3 PC2 H4 PHCl
rC2 H4 Cl2 =   (4:42)
k3 k2 k3 k3
1+ ð1 + PC H Þ + K1 PHCl 1 +
2
+ K2 PC2 H4 ð1 + Þ
k−2 k−1 2 4 k−2 2k5 PO2

Equation (4.42) can be simplified to

K2 k3 PC2 H4 2k5 PO2

rC2 H4 Cl2 = (4:43)
K2 k3 PC2 H4 + 2k5 PO2

4.8.4 Technology

Oxychlorination can be done either in air or oxygen atmosphere. The flow scheme
of an air-based process is given in Fig. 4.109. In this process, ethylene and air are
fed in slight excess of stoichiometry enabling high conversion of HCl. As the vent
system is large because of the presence of nitrogen with the highest molar flow rate
such operation is aimed at minimizing the loss of excess ethylene in the vent
stream. In the oxygen-based process only a small portion of the vent gas is purged,
being significantly lower than in the air-based process. Typical feedstock conver-
sion is approximately 94–99% based on ethylene and 98.0–99.5% for HCl.
Selectivity to dichloroethane is 94–97%.
Direct chlorination is done in reactor 1 where there is a constant level of liquid
containing the catalyst (FeCl3). Heat of the reaction is removed by evaporation of
1,2-dichloroethane, which is partly condensed and partly goes to distillation.
Oxychlorination is done, for example, in a fixed-bed reactor 5 operating under
0.5 MPa at 260–280 °С. Ethylene, recycle gas and HCl are preliminary mixed,
thereafter oxygen (as air) is added to the reactor. The mixing sequence and com-
position are selected to avoid expositions. The heat of reaction is removed by gen-
eration of steam. The outlet gases (unconverted ethylene, oxygen and HCl, as well
526 4 Engineering technology

Gas
2 Recycle 13
10 Gas

2 23
1 5 6
H2O
3 Steam
air 4
C2H4 C2H4 12
7 9 8
CI2 Condensate 8
23 HCI NaOH
HCI

H2O 2 2 2
2 11
2 20

12 19
VCM
14 21
16 17

18 23 22
15 15
15
8
Higher 7 8
15 23
chlorides Dichloroethane

Fig. 4.109: Synthesis of vinyl chloride: 1, chlorator; 2, condenser; 3, collecting vessel; 4, mixer; 5,
reactor; 6, 20, direct heat exchanger; 7, 10, heat exchangers; 8, circulation pump; 9, scrubber; 11,
12, separators; 13, compressor; 14, drying column; 15, boiler; 16, 21, 22, distillation columns; 17,
vessel; 18, pump; 19, furnace; 23, expansion valve [54].

as 1,2 dichloroethane) are cooled in a heat exchanger 6 with a mixture of water

and 1,2 dichloroethane. Partially cooled mixture is treated in a scrubber 9 with al-
kali removing НСl and СО2. After cooling in a heat exchanger 10, the condensate
is separated (pos.11), while the recycled gases (ethylene, oxygen and inerts) after
compression are returned to oxychlorination. Part of this stream is purged.
Condensate from 11 flows to separator 12, where heavier 1,2-dichloroethane is sep-
arated from water. It still contains some water, thus drying is done in column 14.
Two streams of dichloroethane (one from direct chlorination and another un-
reacted dichloroethane after pyrolysis) are then combined and distilled (pos. 16).
Pyrolysis of 1,2-dichloroethane into vinylchloride and HCl is done in a furnace 19
at 1.5 – 2.0 MPa and 500 °С. The gases are cooled in a heat exchanger 20 by a
recycle of 1,2-dichloroethane and thereafter cooled with water. Distillation in
column 21 allows to separate HCl, which is then send to oxychlorination, and the
bottom part containing vinyl chloride and unreacted 1,2-dichloroethane. The
 References 527

latter mixture is separated in column 22 operating under pressure. The bottom

part – unreacted 1,2 dichloroethane – is sent back to distillation (pos. 16).
Utilization of oxygen allows a significant reduction in the volume of the vent
system and operation at lower temperatures resulting in improved operating effi-
ciency and product yield. Larger than stoichiometric excess for ethylene is used in
oxygen-based process compared to air based. Downstream processing includes
cooling of the reaction effluent, its purification from the traces of unconverted HCl,
separation of dichloroethane and water also by condensation as in the air-based
process, recompression, reheating and recycling. Higher selectivity to dichloroeth-
ane is achieved by lowering per pass conversion of ethylene. A purge stream of ap-
proximately 2–5% is needed to avoid accumulation of such impurities as, for
example, carbon oxides, argon or light hydrocarbons, either introduced with the
feed or formed during the reaction.
In fluidized bed reactors, the preheated gas mixture enters the base of the ves-
sel and temperature is controlled either by cooling coils or a bundle of cooling
tubes in the fluid bed. The oxychlorination reaction takes place at a relatively low
temperature, around 220–225°C and at a low pressure of about 2 bar.
At higher temperature, more by-products are formed as both ethylene oxidation
and dichloroethane cracking become more prominent. The latter gives by-products
with higher levels of chlorine substitution. Moreover, elevation of temperature en-
hances sublimation of CuCl2 and thus catalyst deactivation.
By-products in oxychlorination are chloral CCl3CHO, 1,1,2-trichloroethane, chlo-
roform, 1,2-dichloroethylene and ethyl chloride. Removal of chloral is especially im-
portant as it can polymerize in the presence of strong acids. Caustic washing is
done for this purpose preventing thereby polymerization and formation of solids,
which can otherwise clog processing equipment downstream.

References
[1] https://condorchem.com/files/catalogos/Air%20Treatment%20-%20ONLINE.pdf
[2] Beskov, V.S., Safronov, V.S. (1999) General Chemical Technology and the Fundamentals of
Industrial Ecology. Moscow, Russia: Khimia.
[3] Lomic, G, Kis, E., Grabovac, M., Marinkovic-Neducin, R. (2002), Investigation of unusual
inactivity of industrial platinum-rhodium catalyst gauze, Scanning 24: 140–143.
[4] Chernyshev, V. I., Zjuzin, S. V. (2001) Improved start-up for the ammonia oxidation reaction,
Platin. Met. Rev. 45: 3440.
[5] http://www.catalysis.ru/resources/institute/Publishing/buclet/BIC-Booklet-2014_en.pdf
[6] Lloyd, L. (2011) Handbook of Industrial Catalysts. Springer.
[7] Maurer, R., Bartsch, U. (2001) Krupp Uhde Nitric Acid Technology, Enhanced plant design for
the production of azeotropic nitric acid. Heraeus Nitric Acid Conference, Johannesburg, 2001.
http://www.cheresources.com/azeotrop.pdf.
528 4 Engineering technology

[8] Hömmerich, U., Rautenbach, R. (1998) Design and optimization of combined pervaporation/
distillation processes for the production of MTBE. J. Membr. Sci. 146:53.
[9] Noskov, A.S. (2006) Industrial catalytic reactors and their peculiarities, Industrial Catalysis in
Lectures. Moscow, Russia: Kalvis.
[10] http://www.uop.com/reforming-ccr-platforming/.
[11] Bartolomew, C.H., Farrauto, R.J. (2006) Fundamentals of Industrial Catalytic Processes. Wiley.
[12] Krishna, R., Sie, S.T. (1994) Strategies for multiphase reactor selection. Chem. Eng. Sci.
49:4029.
[13] Dudukovic, M.P. (2009) Frontiers in reactor engineering. Science 325:698.
[14] Müller, H. Sulfuric Acid and Sulfur Trioxide, Ullmann’s Encyclopedia of Industrial Chemistry.
DOI: 10.1002/14356007.a25_635.
[15] Peters, T.A., van der Tuin, J., Houssin, C., Vorstman, M.A.G., Benes, N.E., Vroon, Z.A.E.P.,
Holmen, A., Keurentjes, J.T.F. (2005) Preparation of zeolite-coated pervaporation membranes
for the integration of reaction and separation. Catal. Today 104:288.
[16] Koch Modular Process Systems, LLC. Pilot Plant Services Group, http://www.pilot-plant.com/
reactions.htm.
[17] www.cdtech.com/techProfilesPDF/CDTECHEB.pdf.
[18] https://inside.mines.edu/~jjechura/Refining/07_Catalytic_Cracking.pdf. Copyright © 2017
John Jechura
[19] http://en.wikipedia.org/wiki/File:FCC_Chemistry.png
[20] Moulijn, J.A., Makkee, M., Van Diepen, A. (2013) Chemical Process Technology, 2nd Edition.
Chichester, Wiley.
[21] Zeolites and Catalysis: Synthesis, Reactions and Applications, Cejka, J., Corma, A., Zones, S.
(Ed.) Wiley, Weinheim, 2010.
[22] BASF (2011) Multistage Reaction Catalysts (MSRC): A Breakthrough Innovation in Fluid
Catalytic Cracking (FCC) Technology. http://www.catalysts.basf.com/refining.
[23] https://refiningcommunity.com/wp-content/uploads/2017/11/Boron-Based-Technology-an-
Innovative-Solution-for-Resid-FCC-Unit-Performance-Improvement-Clark-BASF-FCCU-
Galveston-2018.pdf
[24] http://www.uop.com/fcc
[25] Vogt, E.T.C., Weckhuysen, B.M. (2015) Fluid catalytic cracking: recent developments on the
grand old lady of zeolite catalysis, Chem. Soc. Rev. 44: 7342–7370.
[26] Fernandes, J.L., Verstraete, J.J., Pinheiro, C.I.C., Oliveira, N.M.C., Ribero, F.R. (2007) Dynamic
modeling of an industrial R2R FCC unit. Chem. Eng. Sci. 62:1184.
[27] http://www.treccani.it/portale/opencms/handle404?exporturi=/export/sites/default/
Portale/sito/altre_aree/Tecnologia_e_Scienze_applicate/enciclopedia/italiano_vol_2/273-
298ITA3.pdf&%5D
[28] Zecevic, J., Vanbutsele, G., de Jong, K.P., Martens, J.A. (2015) Nanoscale intimacy in
bifunctional catalysts for selective conversion of hydrocarbons, Nature, 528: 245–248.
[29] Kolesnikov, I.M., Catalysis and production of catalysts, Moscow, Tekhnika, 2004, 400p.
[30] http://www.treccani.it/export/sites/default/Portale/sito/altre_aree/Tecnologia_e_Scienze_
applicate/enciclopedia/inglese/inglese_vol_2/273-298_ING3.pdf
[31] http://www.treccani.it/portale/opencms/handle404?exporturi=/export/sites/default/
Portale/sito/altre_aree/Tecnologia_e_Scienze_applicate/enciclopedia/inglese/inglese_vol_
2/309-324_ING3.pdf&%5D
[32] Rana, M.S., Samano, V., Ancheyta, J., Diaz, J.A.I. (2007), A review of recent advances on
process technologies for upgrading of heavy oils and residua, Fuel 86: 1216–1231.
 References 529

[33] https://www.eni.com/en_IT/attachments/innovazione-tecnologia/technological-answers/
scheda-est-eng.pdf
[34] http://csd.newcastle.edu.au/chapters/Fig1_2.png
[35] Bodrov, I.M., Apelbaum, L.O., Temkin, M.I. (1967) Kinetics of the reaction between
methane and water vapor catalyzed by nickel on a porous support. Kinetics and Catalysis
8: 821–829.
[36] Khomenko, A.A., Apel’baum, L.O., Shub, F.S., Snagovsky, Yu.S., Temkin, M.I. (1971) Kinetics
of the reaction of methane with water vapor and a reversible reaction of carbon monoxide
hydrogenation on nickel. Kinetika i Kataliz 12: 423–431.
[37] Xu., J, Froment, G.F. (1989) Methane steam reforming, methanation and water-gas shift:
I. Intrinsic kinetics. AIChE J. 35: 88–96.
[38] Aparicio, L.M. (1997) Transient isotopic studies and microkinetic modeling of methane
reforming over nickel catalysts. J. Catal. 165: 262–274.
[39] Avetisov, A.K., Rostrup-Nielsen, J.R., Kuchaev, V.L., Bak Hansen, J.-H., Zyskin, A.G.,
Shapatina, E.N. (2010) Steady-state kinetics and mechanism of methane reforming with
steam and carbon dioxide over Ni catalyst. J. Mol. Catal. A. 315: 155–162.
[40] Beyer, F., Brightling, J., Farnell, P., Foster, C. (2005) Steam reforming – 50 years of
development and challenges for the next 50 years. AIChE 50th Annual Safety in Ammonia
Plants and Related Facilities Symposium, Toronto.
[41] Dul’nev, A.V., Obysov, A.V. (2012) Experience from the industrial operation of natural gas
reforming supported Ni catalysts and ways to improve them. Catal. Ind. 4: 19.
[42] Nielsen, A., Kjaer, J., Hansen, B. (1964) Rate equation and mechanism of ammonia synthesis
at industrial conditions. J. Catal., 3: 68–79.
[43] Temkin, M.I. (1979) The kinetics of some industrial heterogeneous catalytic reactions. Adv.
Catal. 28: 173–281.
[44] Aparicio, L.M., Dumesic, J.A. (1994) Ammonia synthesis kinetics: Surface chemistry, rate
expressions, and kinetic analysis. Top. Catal. 1: 233.
[45] Shepetovsky, D.B., Brodskaya, I.G., Ozolin, P.A. (2007) Loading, reduction and examination
of catalysts “CA-KJ” in industrial column of ammonia synthesis with productivity NH3 1360 t/
d on “OAO AZ0T” (Berezniki). Catal. Ind. N3, 42–47.
[46] Brown, D.E., Edmonds, T., Joyner, R. W., McCarroll, J.J., Tennison, S.R. (2014) The genesis and
development of the commercial BP doubly promoted catalyst for ammonia synthesis, Catal.
Lett. 144: 545–552.
[47] Luzzi, A., Lovegrove, K., Filippi, E., Fricker, H., Schmitz-Goeb, M., Chandapillai, M., Kaneff, S.
(1999) Technoeconomic analysis of a 10 MWe solar thermal power plant using ammonia-
based thermochemical energy transfer. Solar Energy 66:91.
[48] Fertilizer Manual. Kluwer, 1998.
[49] http://www.thyssenkrupp-uhde.de/de/kompetenzen/technologien/fertiliser/ammonia
kharnstoff/75/94/ammonia-conversion.html
[50] http://www.iffcokandla.in/data/polopoly_fs/1.2465622.1437673182!/fileserver/file/507742/
filename/020.pdf
[51] http://www.iffcokandla.in/data/polopoly_fs/1.3253588.1472738687!/fileserver/file/681710/
filename/4F.pdf
[52] https://www.prnewswire.com/news-releases/global-ethylene-oxide-market-report-2018—
forecast-to-2023-the-growing-demand-for-pet-bottles-from-the-packaging-industry-
300755179.html
[53] Makhlin, V.A. (2009) Development and analysis of heterogeneous catalytic processes and
reactors, Theor. Found. Chem. Eng., 43: 261–275.
530 4 Engineering technology

[54] Lebedev, N.N. (1988) Chemistry and Technology of basic organic and petrochemical
synthesis, Chimia.
[55] Krupnov, P.V., Deryugin, A.V., Chesnokov, B.B. (2007) Experience and technical revamp of
Dzerzhinsk ethylene oxide and glycols plant of OAO Sibur Neftekhim, Catal. Ind., 2:43–45.
[56] https://www.digitalrefining.com/article/1001406,Enhancements_in_Ethylene_Oxide_
Ethylene_Glycol_manufacturing_technology.html#.XUG1RW8zb4c
[57] http://www.metrohm-applikon.com/Downloads/Process_Application_Note_AN-PAN-1007-
HPPO-propylene-oxide-production.pdf
[58] http://www.treccani.it/portale/opencms/handle404?exporturi=/export/sites/default/
Portale/sito/altre_aree/Tecnologia_e_Scienze_applicate/enciclopedia/inglese/inglese_vol_
2/615-686_ING3.pdf
[59] Lorz, P.M., Towae, F.K., Enke, W., Jäckh, R., Bhargava, N., Hillesheim, W. Phthalic Acid and
Derivatives. In: Ullmanns Encyclopedia, vol. 27, page 135, DOI: 10.1002/14356007.
[60] Marx, R., Wölk, H.-J., Mestl, G., Turek, T. (2011) Reaction scheme of o-xylene oxidation on
vanadia catalyst. Appl. Catal. A.Gen. 398:37.
[61] Cavani, F., Caldarelli, A., Luciani, S., Cortelli, C., Cruzzolin, F. (2012) Selective oxidation of
o-xylene to phthalic anhydride: from conventional catalysts and technologies toward
innovative approaches. Catalysis (RSC) 24:204–222.
[62] https://catalysts.basf.com/public/files/literature-library/32012BF-9796-7152_BR_PUE_
Catalysts_Ansicht.pdf
[63] US Patent 8153825 (2012) Preparation of phthalic anhydride by gas phase oxidation of
o-xylene, BASF.
[64] Calderbank, P.H., Chandrasekharan, K., Fumagalli, C. (1977) The prediction of the
performance of packed-bed catalytic reactors in the air-oxidation of o-xylene. Chem. Eng. Sci.
32:1435.
[65] Anastasov, A.I. (2003) Deactivation of an industrial V2O5–TiO2 catalyst for oxidation of
o-xylene into phthalic anhydride. Chem. Eng. Proc. 42:449.
[66] http://d-nb.info/1030112800/34
[67] http://www.rollechim.com/switch.htm
[68] Miller, G.J., Collins, M. (2017) Industrial production of formaldehyde using polycrystalline
silver catalyst, Ind. Eng. Chem. Res. 56: 9247–9265.
[69] http://www.treccani.it/portale/opencms/handle404?exporturi=/export/sites/default/
Portale/sito/altre_aree/Tecnologia_e_Scienze_applicate/enciclopedia/inglese/inglese_vol_
2/615-686_ING3.pdf&%5D
[70] https://www.clariant.com/en/Solutions/Products/2019/04/15/07/20/FAMAX-200-Series
[71] http://www.glp.com.au/wp-content/uploads/2014/02/General-SC-Overview-BCT.pdf
[72] https://www.phxequip.com/resource-detail.40/comparing-the-different-formaldehyde-pro
duction-processes.aspx
[73] Burrington, J. D., Kartisek, C. T., Grasselli, R. K. (1984) Surface intermediates in selective
propylene oxidation and ammoxidation over heterogeneous molybdate and antimonate
catalysts, J. Catal. 87: 363–380.
[74] Brazdil, J. F. (2017) A critical perspective on the design and development of metal oxide
catalysts for selective propylene ammoxidation and oxidation, Applied Catal. A., 543: 255–
233.
[75] https://www.ineos.com/globalassets/ineos-group/businesses/ineos-technologies/presenta
tions/edc-oxychlorination-catalysts.pdf
[76] http://www.honorshinechemc.com/photo/pc3635343-catalyst_for_ethylene_oxychlorina
tion_honor_1c_high_copper_content_oc_catalyst_nc_200.jpg
 References 531

[77] https://catalysts.basf.com/public/files/literature-library/1120115938_BR_Oxicloration-FBC_
USLetter.pdf
[78] Vajglová, Z., Kumar, N., Eränen, K., Peurla, M., Murzin, D. Yu., Salmi, T. (2018) Ethylene
oxychlorination over CuCl2/γ-Al2O3 catalyst in micro- and millistructured reactors, J. Catal.,
364: 334–344.
[79] Bakshi, Yu.M., Gel’bstein, A.I., Gel´perin, E.I., Dmitrieva, M.P., Zyskin, A.G., Snagovkii, Yu.S.
(1991) Investigation of mechanism and kinetics of additive oxychlorination of olefins. V.
Kinetic model of the reaction, Kinet. Catal., 32: 740–748.
Acknowledgments
I would like to express my gratitude to very many colleagues who shared their ex-
perience on various aspects of catalysis, in particular to the late Prof. M.I. Temkin
and Dr. N.V. Kul’kova, as well as Dr. A.K. Avetisov (all from Karpov Institute of
Physical Chemistry), Dr. R. Touroude (University of Strasbourg), Dr. E. Schwab
(BASF) and Dr. I.L. Simakova (Boreskov Institute of Catalysis). The everyday inter-
actions with colleagues from the Laboratory of Industrial Chemistry and Reaction
Engineering at Åbo Akademi University (Dr. K. Eränen, Dr. N. Kumar, Prof. J.-P.
Mikkola, Dr. P. Mäki-Arvela and Prof. J. Wärnå) and collaboration with numerous
colleagues and PhD students from Åbo Akademi University and other institutions
are gratefully acknowledged. Please, forgive me for not mentioning you all. Very
special thanks go to my colleague, friend and brother-in-science Prof. T. Salmi,
who has influenced me through the decades in very many ways.
The moral support of my family was very important in writing the first and
the second edition of this book. I am especially grateful to E. Murzina for her techni-
cal and spiritual contribution and especially endless patience and tolerance.

https://doi.org/10.1515/9783110614435-005
Recommended reading
Anderson, B.G., de Jong, A.M., van Santen, R.A. (2002) In situ measurement of heterogeneous
catalytic reactor phenomena using positron emission. In: Haw, J.F., editor. In-situ
Spectroscopy in Heterogeneous Catalysis. Weinheim, Germany: Wiley-VCH, pp. 195–237.
Atherton, J., Houson, I., Talford, M. (2011) Understanding the reaction. In: Houson, I., editor.
Process Understanding for Scale-Up and Manufacture of Active Ingredients. Chapter 4,
Weinheim, Germany: Wiley-VCH, pp. 87–125.
Aris, R. (1975) The Mathematical Theory of Diffusion and Reaction in Permeable Catalysts. Vol. 1.
Oxford: Clarendon Press.
Augustine, R.L. (1996) Heterogeneous Catalysis for the Synthetic Chemist. New York: Marcel
Dekker, Inc.
Bartolomew, C.H., Farrauto, R.J. (2006) Fundamentals of Industrial Catalytic Processes. Wiley.
Hoboken, NJ.
Beller, M., Renken, A., Rutger A. van Santen (2012) Catalysis. From Principles to Applications.
Weinheim, Germany: Wiley-VCH.
Berger, R.J., Stitt, E.H., Marin, G.B., Kapteijn, F., Moulijn, J.A. (2000) Chemical reaction kinetics in
practice. CATTECH 4: 2.
Boudart, M., Djega-Mariadassou, G. (1984) Kinetics of Heterogeneous Catalytic Reactions.
Princeton, NJ: Princeton University Press.
Boudart, M. (1995) Turnover rates in heterogeneous catalysis. Chem. Rev. 95: 661.
Butt, J.B. (2000) Reaction Kinetics and Reactor Design. New York: Marcel Dekker.
Campanati, M., Fornasari, G., Vaccari, A. (2003) Fundamentals in the preparation of heterogeneous
catalysts. Catal. Today 77: 299.
Chorkendorff, I., Niemanstverdriet, J.W. (2003) Concepts of Modern Catalysis and Kinetics.
Weinheim, Germany: Wiley-VCH.
Clark, A. (1970) The Theory of Adsorption and Catalysis. New York: Academic Press.
De Jong, K.P. (2009), Synthesis of Solid Catalysts, Weinheim, Germany: Wiley-VCH
Dittmeyer, R., Emig, G. (2008) Simultaneous heat and mass transfer and chemical reaction. In: Ertl,
G., Knözinger, H., Schüth, F., Weitkamp, J., editors. Handbook of Heterogeneous Catalysis.
Weinheim, Germany: Wiley-VCH, pp. 1727–1784.
Doraiswamy, L.K., Sharma, M.M. (1984) Heterogeneous reactions: Analysis, examples and reactor
design. Vol. 1, Gas-Solid and Solid-Solid Reactions, Vol. 2, Fluid-Fluid and Fluid-Fluid-Solid
Reactions. New York: John Wiley.
Dumesic, J.A., Rudd, D.F., Aparicio, L. (1993) The Microkinetics of Heterogeneous Catalysis.
American Chemical Society, Washington, DC.
Dutta, S., Gualy, R. (2000) Build robust reactor models. Chem. Eng. Prog. 37–51.
Eigenberger, G. (1992) Fixed Bed Reactors, Ullmann’s Encyclopedia of Industrial Chemistry.
Vol. B4, 199.
Euzen, J.P., Trambouze, P., Wauquier, J.P. (1993) Scale-up Methodology for Chemical Processes.
Paris: Technip.
Fogler, H.S. (1998) Elements of Chemical Reaction Engineering (3rd edn). Englewood Cliffs,
NJ: Prentice Hall.
Geus, J.W., van Dillen, A.J. (2008) Preparation of supported catalysts by deposition-precipitation.
In: Knözinger, H., Schueth, F., Weitkamp, J., editors. Handbook of Heterogeneous Catalysis,
2.4.1. Weinheim, Germany: Wiley-VCH, pp. 428–467.
Gladden, L.F., Mantle, M.D., Sederman, A.J. (2006) Magnetic resonance imaging of catalysis and
catalytic processes. Adv. Catal. 50: 1.

https://doi.org/10.1515/9783110614435-006
536 Recommended reading

Frank-Kamenetskii, D.A. (1955) Diffusion and Heat-transfer in Chemical Kinetics. Princeton,

NJ: Princeton University Press.
Froment, G., Bischoff, K. (1990) Chemical Reactor Analysis and Design (2nd edn). New York: Wiley.
Haber, J., Block, J.H., Delmon, B. (1995) Methods and procedures for catalyst characterization. Pure
Appl. Chem. 67: 1257–1306.
Hagen, J. (2006) Industrial Catalysis: A Practical Guide. Weinheim, Germany: Wiley-VCH.
Hammer, B., Norskov, J.K. (2000) Theoretical surface science and catalysis – calculations and
concepts. Adv. Catal. 45: 71.
Ertl, G., Knözinger, H., Weitkamp, J., editors. (1997) Handbook of Heterogeneous Catalysis. 5 Volume
Set. Weinheim, Germany: Wiley-VCH.
Hanefeld, U., Lefferts, L., editors (2018). Catalysis. An Integrated Textbook for Students. Weinheim,
Germany: Wiley-VCH.
Hessel, V., Löwe, H., Mueller, A., Kolb, G. (2005) Chemical Micro Process Engineering. Weinheim,
Germany: Wiley-VCH.
Horiuti, J., Nakamura, T. (1967) Theory of heterogeneous catalysis. Adv. Catal. 17: 1.
Horiuti, J. (1973) Theory of reaction rates as based on the stoichiometric number concept. Ann. NY
Acad. Sci. 213: 5.
Ioffe, I.I., Reshetov, V.A., Dobrotvorskii, A.M. (1985) Heterogeneous Catalysis. Leningrad: Khimia.
Kiperman, S.L. (1979) Foundations of Chemical Kinetics in Heterogeneous Catalysis. Moscow: Khimia.
Kolasinski, K.W. (2002) Surface Science. Foundations of Catalysis and Nanoscience. Chichester: Wiley.
Kolesnikov, I.M. (2004) Catalysis and Production of Catalysts. Moscow: Teknika.
Laidler, K.J. (1987) Chemical Kinetics. New York: Harper and Row.
Levenspiel, O. (1998) Chemical Reaction Engineering (3rd edn). Weinheim, Germany: Wiley.
Lloyd, L. (2011) Handbook of Industrial Catalysts. Berlin: Springer.
Logan, S.R. (1996) Fundamentals of Chemical Kinetics. Edinburgh Gate: Longman.
Marceau, E. (2006) Characterizing a supported catalyst: A methodological approach. In: Murzin, Yu.
D., editor. Nanocatalysis. ISBN: 81-308-0056-X. Kerala, India: Research Signpost. pp. 127–175.
Marceau, E., Carrier, X., Che, M., Clause, O., Marcilly, C. (2008) Ion exchange and impregnation. In:
Knözinger, H., Schueth, F., Weitkamp, J., editors. Handbook of Heterogeneous Catalysis, 2.4.2..
Weinheim, Germany: Wiley-VCH. DOI: 10.1002/9783527610044.hetcat0022, pp. 467–484.
Marceau, E., Carrier, X., Che, M. (2009) Impregnation and drying. In: de Jong, K.P., editor.
Synthesis of Solid Catalysts. Weinheim, Germany: Wiley-VCH, pp. 59–82.
Masel, R.I. (2001) Chemical Kinetics and Catalysis. New York: Wiley-Interscience.
Missen, R.W., Mims, C.A., Saville, B.A. (1999) Chemical Reaction Engineering and Kinetics.
New York: John Wiley & Sons.
Deustchmann, O., editor. (2012) Modelling and Simulations of Heterogeneous Catalytic Reactions:
From the Molecular Process to the Technical System. Weinheim, Germany: Wiley-VCH.
Moulijn, J.A., Makkee, M., van Diepen, A.E. (2013) Chemical Process Technology. Second Edition.
Chichester: John Wiley and Sons.
Murzin, D., Salmi, T. (2016) Catalytic Kinetics. Science and Engineering. Amsterdam: Elsevier.
Murzin, D. Yu. (2015) Catalytic Reaction Technology. Berlin: Walter de Gruyter.
Ostrovskii, N.M. (2001) Kinetics of Catalyst Deactivation. Moscow: Nauka.
Petrov, L.A. (1992) Application of graph theory to study of the kinetics of heterogeneous catalytic
reactions. Math. Chem. 2: 1.
Petrov, L.A., Alhamed, Y., Zahrani, A.Al., Daous, M. (2011) Role of chemical kinetics in the
heterogeneous catalytic studies. Chinese J. Catl. 32: 1085.
Pilling, M.J., Seakins, P.W. (1995) Reaction Kinetics. Oxford: Oxford University Press.
Poling, B.E., Prausnitz, J.M., O’Connell, J.P. (2001) The Properties of Gases and Liquids. McGraw-Hill.
New York.
 Recommended reading 537

Post, T. (2010) Understanding the real world of mixing. Chem. Eng. Prog. 25.
Ross J.R.H. (2019), Contemporary Catalysis. Fundamentals and Current Applications. Amsterdam:
Elsevier.
Rothenberg, G. (2008). Catalysis. Concepts and Green Applications. Weinheim, Germany: Wiley-VCH.
Salmi, T.O., Mikkola, J.-P., Wärnå, J.P. (2010) Chemical Reaction Engineering and Reactor
Technology. CRC Press, Boca Raton, Fl.
Satterfield, C.N. (1970) Mass Transfer in Heterogeneous Catalysis. Cambridge, MA: MIT press.
Schmidt, F. (2001) New catalyst preparation technologies observed from an industrial viewpoint.
App. Catal. A. Gen. 221: 15.
Schmidt, L.D. (1998) The Engineering of Chemical Reactions. Oxford: Oxford University Press.
Smith, J.M. (1981) Chemical Engineering Kinetics. McGraw-Hill. New York.
Schüth, F., Hesse, M. (2008) Catalyst forming. In: Ertl, G., Knözinger, H., Schüth, F., Weitkamp, J.,
editors. Handbook of Heterogeneous Catalysis. Weinheim, Germany: Wiley-VCH, pp. 676–699.
Snagovskii, Yu.S., Ostrovskii, G.M. (1976) Modelling of Kinetics of Heterogeneous Catalytic
Reactions. Moscow: Khimia.
Somorjai, G.A. (1994) Introduction to Surface Chemistry and Catalysis. New York: Wiley-
Interscience.
Stolze, P. (2000) Microkinetic simulation of catalytic reactions. Prog. Surf. Sci. 65: 65.
Stitt, E.H., Simmons, M.J.H. (2011) Scale-up of chemical reactions. In: Houson, I., editor. Process
Understanding for Scale-up and Manufacture of Active Ingredients. Weinheim, Germany: Wiley-
VCH, pp. 155–198.
Thomas, J.M., Thomas, W.J. (1997) Principles and Practice of Heterogeneous Catalysis. Weinheim,
Germany: Wiley-VCH.
Temkin, M.I. (1979) The kinetics of some industrial heterogeneous catalytic reactions. Adv. Catal.
28: 173.
Twygg, M. (1997) Handbook of Industrial Catalysis. Manson Publishing, London.
van Santen, R.A., van Leeuwen, P.W.N.M., Moulijn, A.A., Averill, B.A. (2002) Catalysis: An
Integrated Approach. Elsevier, Amsterdam.
van Santen, R.A., Neurock, M. (2017) Modern Heterogeneous Catalysis. An Introduction. Weinheim,
Germany: Wiley-VCH.
van Santen, R.A. (2006) Molecular Heterogeneous Catalysis. Weinheim, Germany: Wiley-VCH.
Önskan, Z.P, Avci, A.K. (2016) Multiphase Catalytic Reactors, Theory, Design, Manufacturing and
Application. Weinheim, Germany: Wiley-VCH.
Index
3D printing 220–222 atomic layer deposition 119, 125, 181, 201–202
AAS 62 attenuated total reflectance 91
absorber 96, 97, 399, 400 Avrami-Erofeev 142, 168
activated carbon 55, 67, 146–149, 153, 195 axial dispersion 372, 373
activation energy 3, 4, 9, 19, 80, 81, 114, 137, axial flow 330, 343, 363, 383, 476, 480,
242, 272–274, 283, 286, 290, 305, 306, 481, 503
317–319, 484, 496, 517
activity 1, 4–16, 21, 38, 40, 49, 50, 53, 57, 59, back-mixing 330, 334, 339–342, 453, 476,
108, 110, 115–118, 121, 122, 133, 136, 138, 484, 518
146, 147, 165, 184, 197, 202, 215, 223, 229, baffles 294, 343, 518
230, 232–235, 275–279, 288, 294, 303, Berty 323, 324, 327
311, 313, 314, 319–321, 334, 344, 346, Balandin 4
347, 350–352, 358, 359, 381, 403, Berzelius 1
409–411, 416, 424, 426, 433–435, 437, BET 63, 67, 68, 70
443, 445, 446, 453, 464, 465, 467, 470, beta zeolite 94–95
475, 476, 479, 481, 485, 486, 490, 500, binder 38, 131, 132, 158, 169, 170, 177, 203,
501, 504, 516, 517, 521, 523 206, 210, 212, 214, 215, 219, 221, 224, 335,
acrylic acid 512–514 352, 359, 434, 445, 446
acrylonitrile 514–518 biographical non-uniformity 257
adsorption 1, 3, 4, 6, 16–19, 21, 23–26, 59, Biot number 306
61–64, 66–75, 78, 82, 83, 88–90, 95, Bodenstein 237
100, 114, 116, 118, 123, 125, 141–143, 146, body-centered cubic 22
148, 182–194, 199, 213, 239, 240, 242, boehmite 160, 214
243, 245, 246, 248, 249, 252, 257–259, Boltzmann 269, 296
262, 263, 265, 266, 267, 270, 271, 274, Bosch 468, 469
275, 284, 434, 450, 453, 460, 461, Boudart 266, 319, 473
472–475, 479 Boudouard reaction 346, 455
AFM 104–105 Brønsted acidity 29, 162, 163, 165, 176
ALD 201, 202 bubble column 327, 335–337, 340–342, 344,
alumina 14, 15, 38, 40, 43, 53–55, 68, 84, 90, 352, 355, 356, 453
102, 109, 119, 121, 128, 130–132, 134, 146, bulk density 13, 14, 49, 52, 172, 235, 263, 371,
150, 152, 154, 155, 157, 159–163, 166, 167, 372, 437, 448, 477
172, 177, 179, 183, 184, 190, 198, 202, 203,
213, 214, 219, 220, 351, 357, 361, 424, calcination 57, 119, 121, 123, 125, 128–131, 135,
433–436, 444, 445, 447, 452, 475, 520 143–145, 160, 169, 171–173, 175, 178, 180,
aluminophosphate 162 187, 195, 197, 204, 206, 213–220, 223,
ammonia synthesis 5, 52, 134, 383, 407, 408, 351, 352, 435, 445, 446, 466
456, 457, 463, 467–483 calcium aluminate 465, 466
ammoxidation 512 calorimetry 82–83
antibonding 20–22 capillary condensation 63, 64, 67, 71, 72
approach to equilibrium 4, 229, 364, 381, 456, capillary impregnation 188, 189
465, 467, 468 carbenium ions 29, 427, 428
Archimedes 336 Carberry 323, 324, 327
Arrhenius equation 236, 272, 283, 376 catalysis 1, 4–7, 10, 15–17, 21, 22, 24, 27,
atom efficiency 44, 496 29–45, 49, 50, 57–117, 121, 124, 145, 154,

https://doi.org/10.1515/9783110614435-007
540 Index

164, 173, 176, 190, 232–235, 237, 239, 274–280, 288, 320, 321, 327, 334, 346,
245, 250, 256, 257, 268, 274, 275, 351, 352, 354, 358, 359, 361, 367, 392,
284–286, 299, 303, 310, 315, 320, 331, 396, 398, 404–411, 429, 431, 435, 437,
351, 365, 398, 443, 461, 466 442, 448, 451, 455, 464, 465, 479, 484,
catalyst deactivation 12, 13, 43, 49, 321, 352, 509, 510, 518, 527
358, 359, 367, 411, 429, 431, 442, 451, deposition-precipitation 119, 123–125,
455, 464, 479, 509, 518, 527 198–200
catalyst forming 202–203 desorption 17, 62–64, 66, 71–73, 75, 77–81,
catalyst preparation 10, 50, 58, 98, 107, 134, 116, 127, 164, 215, 239, 259, 260, 267, 271,
143–145, 153, 183–185, 193, 196, 199, 348, 430, 443, 461, 500
201–203, 205, 217, 219, 221–224, 351, DFT 59, 115, 116
435, 516 diatomaceous earth 14, 159
catalyst reduction 77, 222, 337, 383, 384, differential selectivity 12, 261, 311–313
476–478, 510 diffusion coefficient 189, 191, 192, 287, 290,
catalytic cracking 7, 29, 81, 82, 109, 164, 169, 292, 295–300, 306, 315, 316, 318, 375
203, 274, 334, 335, 401, 410, 423–441 distillation column 359, 394, 406, 421, 422,
catalytic cycle 2, 4, 7, 262, 264, 265, 275 494, 507, 508
ceramic foam 178–180 DRIFTS 88
cetane number 426 Dubinin 71
Chapman-Enskog 290, 295, 296 Dumesic 271, 473
chemical technology 391, 392
chemical vapor deposition 98, 125, 150, 181 E-factor 44
chemisorption 10, 15–22, 62, 74–77, 133, 430, effectiveness factor 234, 288, 300, 303–311,
472, 473, 516 313, 315, 316, 319, 354, 368, 370, 371, 465,
chemoselectivity 261 466, 481
Chilton-Colburn 291 egg yolk 190
chromatography 110, 112–114 egg-shell distribution 190
CNT 149–152 electric double layer 153
coke 7, 12, 77, 81, 82, 95, 107, 146, 275–279, Eley-Rideal 17, 241, 248, 262, 273, 305,
346, 383, 409–411, 423–426, 429–436, 371, 372
439, 440, 443, 446, 451, 452, 455, 465, emissions 5, 32, 35, 36, 38–40, 88, 136, 197,
496, 509, 522 214, 223, 400, 418–420, 435, 496
coking 15, 165, 197, 278, 279, 351, 409, 411, enantioselectivity 261
426, 431, 443, 451, 507 energy of dissipation 291, 292, 294, 319
complete oxidation 15, 16, 484, 486, 487, enzyme 30
489, 496 EPR 94, 95
conceptual process design 394–406 ethylene epoxidation 25, 485–490
configurational diffusion 300, 314, 315 EXAFS 61, 84, 98, 99
cordierite 15, 53, 54, 177, 179, 180 exhaust gas cleaning 40
crushing 50, 218, 223, 465 explosions 334, 367, 368, 409, 414, 484, 487,
crystallography 21–24 489, 493, 494, 497, 498, 507
CVD 125, 201 external diffusion 16, 17, 49, 223, 286–295,
cyclone 157, 334, 438, 439, 484, 518, 522 303, 306, 316, 324, 395
extrudate 6, 7, 52, 53, 124, 128, 131, 132, 147,
Damköhler number 288 206–208, 210, 211, 213–218, 328, 378,
Darcy’s law 188 381, 452
De Donder 231 extrusion 52, 128, 131, 132, 150, 177, 202,
deactivation 12, 13, 33, 38, 43, 49, 59, 62, 89, 206–218, 221, 224, 351, 352
109, 110, 115, 118, 224, 229, 234, Eyring 3, 269
 Index 541

face-centered cubic 23 HAADF 102

FCC 109, 171, 205, 220, 275, 403, 423–426, Haber 91, 468, 469
429, 432, 434, 435, 437, 439–443, 484 Harkins-Jura-de Boer 70
Fick 188, 287, 298, 301, 373 heat capacity 231, 291, 311, 332, 376, 377, 498
filtration 52, 119, 125–127, 144, 170, 175, heat exchanger 326, 332, 361, 363, 394, 399,
179, 400 400, 414–416, 419, 447, 449, 471, 482,
Fischer-Tropsch 135, 345–357 487, 489, 510, 518, 522, 526
fixed-bed 8, 51, 52, 56, 58, 107, 146, 158, 196, heat transfer 12, 16, 54, 57, 179, 220, 223, 229,
223, 234, 276, 277, 278, 280, 292, 303, 232, 234, 281, 284, 285, 287–291, 317,
321, 327, 328–331, 334, 337, 338, 340, 318, 326, 329, 330, 334, 341, 344, 356,
345, 347, 350, 352, 354, 361, 364, 368, 367, 368, 370, 374, 376, 449, 456, 464,
373, 407, 409, 410, 414, 418, 424, 452, 465, 467, 503, 504, 506, 510, 518
453, 458, 492–494, 496, 497, 508, 518, hemicellulose 33
520–522, 525 Herschel-Bulkley model 209
flame hydrolysis 118, 154, 157 heteropoly acids 145–146
flow circulation reactor 325 high throughput experimentation 4, 57, 58
flow diagram 386, 394, 398–400, 513 Hinshelwood 17, 27, 240, 241, 248, 264, 265,
flow reversal 332, 414 271, 305
fluid catalytic cracking 81, 82, 109, honeycomb 132, 177, 210, 216, 221
169, 275, 401, 410, Horiuti 27, 28, 247–250, 252, 525
423–441 Horiuti-Polanyi 27, 28
fluidized bed reactor 52, 131, 150, 157, hot spot 119, 179, 330, 367, 377, 378, 404, 415,
203, 333–336, 344, 354, 367, 368, 374, 416, 449, 464, 465, 484, 492, 494, 496,
410, 414, 433, 484, 494, 497, 514, 501, 504, 511, 523, 524
516–523, 527 HREELS 60
fluidized catalytic cracking 7, 203, 334 hydrocracking 14, 348, 441–454
fluorescence microscopy 108 hydrodesulfurization 129, 383, 408, 411,
Fourier 88, 93, 98, 287, 308 457, 458
formaldehyde 507–512 hydrodynamics 53, 57, 229, 292, 327, 341, 365,
Fowler-Guggenheim isotherm 260 366, 368, 370, 371
freeze drying 128 hydrogenation 5, 8, 14–16, 27–28, 33, 52, 84,
Freundlich 257 121, 146, 154, 173, 247, 255, 262, 263, 327,
Froment 460, 461 336, 348, 357, 367, 383, 384, 392, 393,
FTIR 88, 89, 91 407, 409, 430, 442–445, 447, 448, 457
fuel cell 45, 46 hydrotalcite 176
fugacity 472 hydrogenolysis 243, 246
hydrotreating 14, 407, 410, 425, 435, 442, 443,
gasoline 15, 32, 33, 169, 348, 354, 404, 410, 446–450, 453, 457
424–426, 429–438, 441–445 hysteresis 63–65, 72, 73
gauze 395, 396–398
gelation 119, 120, 158, 204, 206, 214–215 ICP-MS 105, 106
Gibbs free energy 2, 45, 231, 257 ideal adsorbed layer 17
gradientless 261, 274, 324, 327 iLEM 108, 109
grafting 54, 119 immobilization 119
granulation 131, 202, 205–207, 352 impregnation 57, 98, 118, 122, 123, 125, 147, 178,
granulation pan 206, 207 181–183, 186–197, 222, 351, 352, 466, 521
graphite 132, 144, 148, 150, 152, 218, 219, incipient wetness impregnation 123, 195
359, 479 induced non-uniformity 257, 260
grinding 144, 197, 205, 206, 219 inhibitor 13, 15, 485, 487
542 Index

interfacial energy 139, 141, 198, 200 Mars-van Krevelen 26, 27, 255, 513
internal diffusion 16, 17, 49, 286, 300, 303, mass action law 16, 17
305, 306, 312–315, 319, 343, 370, 372, mass transfer limitations 52, 53, 288, 292,
373, 375 293, 318, 319, 320, 327, 334, 343, 354,
ion exchange 55, 57, 98, 106, 112, 118, 123, 374, 453, 486, 500
125, 162, 170, 171, 176, 181, 182, 433, 511 MCM-41 84, 85, 100, 106, 174, 175
IR 60, 87–89, 91–94, 97, 164, 270 mechanical strength 7, 10, 13, 49, 52, 53, 55,
isoelectric point 123, 153, 154, 161, 182, 183 146, 149, 179, 187, 206, 216, 465, 467
isomerization 15, 28, 29, 91, 238, 242, 357, membrane 420, 421
426, 428–431, 442–445, 493 mesoporous materials 75, 95, 100, 146,
isotope labeling 26 174–175
metal dispersion 74–77, 89, 125, 133, 154, 201,
KAAP 478–480 268, 319, 351, 359, 466
Keggin 145 metal organic frameworks 173–174
Kelvin equation 72, 73, 141 metallic foams 178, 179
kieselguhr 14, 159 methanation 5, 116, 346, 350, 456
kinetic polynomial 268–269 microchannels 93, 326, 374
kinetics 1, 7–9, 17, 23, 24, 26, 28, 49, 80, 110, microkinetic modeling 271, 461
133, 135, 137, 141, 183, 216, 229, 232–284, microreactors 58, 220, 326, 327, 374
286, 288, 298, 299, 301, 303, 305, 306, microscopic reversibility 237
313, 315, 317, 320, 327, 329, 330, 342, Miller index 19, 23
358, 365–368, 372, 392, 405, 431–433, millisecond catalytic cracking 440
459, 470–474, 502, 517, 524 Mittash 468, 475
Knudsen diffusion 299, 300, 314 modeling 7, 49, 59, 114, 184, 190, 200, 221,
Kolmogorov 291 224, 229, 232, 242, 271, 284, 285, 315,
Kunii-Levenspiel 374 330, 365–369, 374, 431, 474
molecular orbital 20, 21
LA-ICP-MS 105, 106 monoethanolamine 456
Langmuir 16, 17, 27, 66, 70, 185, 188, 239, monolayer capacity 63–66, 68
240, 241, 243, 248, 260, 264, 265, monolith 36, 42, 43, 52–55, 107, 124, 125, 132,
271, 305 177–180, 210, 215, 216, 220, 324, 331,
Langmuir-Hinshelwood mechanism 17, 27, 240, 340, 394, 398
241, 247, 264, 265, 271, 305 Mössbauer 61, 96–98
lateral interactions 19, 62, 63, 185, 260, 284 moving-bed 52, 334, 340–344, 424
leaching 7, 47, 121, 133, 275, 490 MTBE 401, 403, 404
length scale 6 multitubular 329, 330, 332, 378, 380–382,
Lennard-Jones 296 394, 404, 456, 482, 484, 486, 492, 500,
Levenberg-Marquardt 281 502, 504, 510, 511, 518
Lewis acidity 160, 163, 176 multitubular reactor 329, 330, 332, 378, 380,
light scattering 93 381, 382, 394, 404, 456, 484, 486, 500,
lignin 33, 246 502, 504, 510, 511, 518
lignocellulosic 33
Lowenstein’s rule 163 nanocluster 11, 133, 200
nanotechnology 10
machine learning 60 naphthalene 406, 494–501, 504–506
maldistribution 327, 339, 449, 464 natural gas 14, 15, 32, 33, 52, 162, 172, 275,
maleic anhydride 26, 368, 414, 415, 491–494, 347, 348, 350, 385, 408, 411, 423,
499, 501, 506 454–468, 476, 480
 Index 543

NMR 61, 99–101, 106 pore diffusion 7, 298–307, 311, 313, 318, 330,
nucleation 78, 115, 119, 129, 133, 136, 137, 334, 367
139–143, 155, 156, 168, 198, 202, 206 pore size 14, 19, 49, 59, 63, 70–74, 135, 159,
Nusselt 291 165, 175, 179, 206, 221, 300, 303, 318, 512
pore volume impregnation 122, 187, 192
octane number 169, 430–432, 435, 448 porogenes 202, 213, 214, 218
oil refining 29, 31, 32, 105, 146, 335, 336, porosity 7, 14–16, 49, 51, 62, 63, 118, 121, 125,
407, 441 131, 132, 146, 148, 158, 162, 189, 202, 203,
operando 91, 92 208, 214, 218, 219, 223, 298, 299, 303,
Ostwald- de Waele equation 209 308, 318, 319, 375, 485, 512
oxidation 1, 14–16, 25–27, 33, 35, 40, 43, 44, potential energy 2, 3, 18, 59, 116
47, 52, 74, 77, 81–82, 86, 87, 94–98, 128, potential energy diagram 2, 3, 18
147, 148, 154, 159, 196, 214, 219, 243, 247, power law 209, 259, 276, 277
251, 252, 255, 265, 275, 333, 359, 368, power number 294
379, 380, 382, 385, 386, 397–401, 409, Prandtl 291, 311
411, 414, 415, 418, 423, 476, 479, Prater 308–310, 317, 318
483–518, 521, 522, 524, 527 precious metals 10, 47, 134, 147, 385
oxidative dehydrogenation 507 precipitation 118, 119, 123–125, 132, 135–146,
oxychlorination 518–527 157, 160, 176, 181, 187, 197–201, 223,
o-xylene 165, 275, 380, 382, 490, 495–506 352, 466
precursor 57, 119, 122, 123, 128, 133, 141, 147,
palladium 15, 38, 76, 84, 173, 383, 384, 396, 155, 166, 167, 181–183, 186–193, 198, 199,
397, 444, 446, 447 201, 202, 270, 276, 383, 479
parameter estimation 246, 268, 280–282, pressure drop 40, 43, 50–52, 54, 125, 131, 178,
284, 372 179, 190, 202, 223, 303, 324, 328, 330,
partition function 269, 270, 399 331, 337, 338, 340, 354, 356, 361, 363,
patent 118, 181, 423, 501 364, 378, 381, 382, 464, 465, 467, 481,
Peclet 368, 372, 373 504, 510, 513, 521, 523
pelletizing 144, 207 promoter 13, 15, 218, 359, 419, 436, 475, 479
PET 107, 108 propylene oxide 491
petrochemistry 32 pulsed flow 339
phthalic anhydride 406, 407, 494–507 pzc 182–187, 199, 200
physisorption 16–20, 62, 63, 65, 67, 73, 74
Plank 269 quasi-equilibrium 239, 252, 273
plastic deformation 219 quench 359, 418, 439, 449, 453
platforming 410
platinum 1, 15, 36, 42, 43, 45, 152, 183, 190, radial flow 330, 343, 364, 379, 476, 477,
191, 395–398, 444 479, 481
plug flow 209, 211, 320, 321, 330, 334, 335, Raman 60, 87–93
337, 340, 344, 370–374, 376, 377, 404, Raney nickel 121
415, 460 rare earth zeolite 170, 171, 172, 434
point of zero charge 153, 182, 184 Raschig ring 51, 467
poison 128, 276, 279, 403 rate-limiting step 237
poisoning 14, 15, 190, 275, 288, 313, 314, 352, reaction order 231, 236, 256, 259, 267, 269,
365, 411, 465 306, 313, 316, 431, 460
Polanyi parameter 257, 267 reaction rate 3, 4, 7, 12, 229, 232, 233, 235,
pollution 32 240, 241, 243, 245, 249, 250, 252, 254,
polymerization 14, 16, 158, 348, 349, 426, 258–260, 264, 267–269, 276–278, 280,
430, 527 287, 288, 290, 298, 300–302, 305, 306,
544 Index

308, 311, 315, 317, 319, 321, 337, 376, shear viscosity 208
377, 395, 405, 407, 449, 461, 467, 472, Sherwood 288, 290, 291
473, 524 sieving 197, 377
reaction route 248, 249, 251, 265, 266, 278 silica 14, 29, 84, 109, 111–113, 119, 121, 134,
reactor 5–8, 13, 14, 49–58, 62, 107, 110, 124, 146, 152, 154, 155, 157–160, 163, 165, 167,
129–131, 142, 144, 146, 150, 155, 157, 158, 175, 177, 182, 183, 184, 186, 199, 203, 214,
178, 179, 190, 194–197, 202, 203, 218, 351, 352, 357, 414, 419, 424, 433,
218–220, 223, 229, 232–235, 247, 434, 444, 445, 447, 466, 475, 492, 497,
274–276, 278–280, 285, 286, 288, 499, 500, 516, 517, 523
291–295, 303, 311, 319, 320–386, simplex algorithm 281
391–394, 398–400, 404–426, 432–434, sintering 10, 14, 15, 53, 84, 128, 129, 133, 150,
437–440, 442, 446, 449, 450, 452–454, 180, 214, 221, 275, 276, 278, 352, 354,
456–458, 460, 464, 465, 469, 471, 476, 358, 359, 409, 411, 446, 466, 486,
477, 479–490, 492–494, 496–504, 506, 509, 510
508–514, 516–527 sol-gel 118, 120, 121, 157–159, 352
refinery 348, 403, 423, 441, 453, 457, 463 solubility curve 138, 198
regeneration 7, 13, 43, 59, 81, 117, 118, 128, specific mixing power 292, 295
146, 276–278, 331, 334, 351, 352, 366, spillover 14, 75, 243
368, 383, 384, 392, 405, 408–411, 424, sponge catalysts 121, 134
425, 433, 439, 443, 446, 515 spray drying 131, 137, 202–205, 352, 414, 500
regenerator 274, 334, 368, 410, 411, 424, 432, spray pyrolysis 155
434, 438, 439, 494 steady-state approximation 237, 238, 245,
regioselectivity 261 247, 254
residence time 52, 155, 157, 234, 292, 293, steam reformer 382, 456, 463, 464, 467
326, 330, 340, 371, 372, 375, 395, 432, steam reforming 14, 15, 52, 115, 116, 162, 255,
433, 434, 439, 442, 445, 448, 510, 514 275, 302, 380, 381, 407, 408, 411,
Reynolds 288–292, 294, 319, 336, 368, 372 454–468, 476
rotational 54, 126, 206, 269, 270 steam to carbon 456, 463, 465
steepest descent 281
Sabatier 4, 5, 124 stereoselectivity 28
Scherrer equation 83 STM 104, 105
Schmidt 288, 290, 291, 311 stoichiometric number 235, 247, 250, 252,
selective oxidation 26, 44, 154, 484, 487, 491, 253, 266
499, 501 strong metal support interactions 154
selectivity 1, 7–13, 15, 16, 28, 32, 36, 38, 44, structure sensitivity 12, 266–268
57, 59, 110, 114, 117, 118, 122, 162, sulfur dioxide 14, 379, 418
164–166, 173, 179, 202, 223, 229, 232, sulfuric acid 157, 172, 214, 217, 218, 380, 386,
235, 261–267, 276, 311, 312, 313, 315, 320, 393, 418, 419
321, 330, 337, 347–349, 351, 352, 358, superficial gas velocity 341, 439
361, 391, 393, 395, 397, 407, 416, 424, supersaturation 119, 123, 136, 138–142, 199
425, 434, 440, 442, 483–487, 490, 492, support 6, 8, 10, 11, 13–15, 53, 57, 59, 63, 75,
493, 494, 496, 499, 500, 501, 503, 504, 76, 78, 84, 85, 87, 102, 111, 119, 122–124,
509, 510, 512, 514, 516–518, 521–525, 527 128, 130, 133, 135, 146–150, 153, 154,
SEM 61, 103, 105 158–160, 177, 179, 180–184, 186–189,
SFG 60, 93 191–201, 206, 213, 216, 218, 222, 261,
shape selectivity 162, 165, 173, 261, 300, 276, 278, 351, 352, 357, 391, 416, 419,
315, 435 445, 465, 466, 479, 485, 486, 497,
shear stress 208 500, 521
 Index 545

supported metal catalysts 76, 81 transport phenomenon 7

surface area 7, 10, 11, 13–15, 19, 38, 43, 49, trickle bed 107, 327, 328, 337–340, 352, 449
52–55, 59, 62–71, 74–76, 84, 117, 119, 121, tube wall temperature 465, 467
122, 124, 125, 133, 134, 144, 146, 149, 152, turnover frequency 5, 7, 233, 267
154–157, 159, 160, 162, 173, 176, 177, 179, turnover number 7, 10, 233
180, 208, 213, 221, 287, 288, 299, 301, two-step sequence 244, 245, 258–260,
318, 326, 376, 396, 465, 467, 475, 476, 267, 273
486, 492, 505, 512, 521
surface charge 93, 123, 152, 153, 159, 161, Van der Waals 16, 18, 116, 120, 148, 203
182, 185 vibrational 60, 87, 92, 93, 269, 270
surface heterogeneity 62, 284, 469 volcano plots 4
surfactants 132, 133, 172, 174, 485
switch condenser 504–506 Washburn equation 74
syneresis 172 washcoat 38, 43, 177
washcoating 38, 177
Tamman temperature 129 washing 43, 118, 119, 125–127, 143, 144, 147,
TEM 61, 102, 103, 109, 133 158, 159, 169, 171, 172, 175, 182, 185, 197,
Temkin 83, 248–250, 252, 257, 259, 269, 352, 393, 445, 527
322–324, 459, 472, 473 water-gas shift 119, 177, 245, 345, 406,
Temkin-Kul’kova reactor 322, 323 456, 459
temperature programmed 62, 75, 77–82, 128, Weisz modulus 306, 307, 310, 311, 318
164, 215 wetting 107, 119, 123, 179, 187, 189, 206,
texture 6, 120, 128, 146 337–339
the reaction coordinate 2, 269 Wheeler equation 72
the transition state theory 3, 269, 271, Wilke-Chang 296
284, 399 wustite 134, 474, 475
thermal cracking 423, 442, 518, 520
thermal treatment 118, 129, 131, 164, 193, 445 XPS 60, 61, 75, 84–87, 105
thermodynamics 4, 229–232, 256, 257, 266, X-ray diffraction 61, 75, 83–84
268, 271, 329, 392, 405, 469–471, X-zeolite 433
496–498
Thiele modulus 301–306, 308–310, 312, Zeldovich-Roginski 257
314–317, 481, 486 zeolites 7, 10, 15, 29, 33, 67, 70, 75, 81,
titania 15, 36, 143, 146, 154–156, 179, 215, 216, 83, 84, 88, 95, 97, 101, 108, 121,
351, 499, 500 123, 146, 162–175, 221, 261, 300,
TPD 78–81, 215 315, 354, 357, 374, 386,
t-plot 68, 70, 71 433–435, 445
TPR 78 zeta potential 153, 154, 160, 161, 214
transition metals 16, 21, 57, 59, 94, 116, 133, ZSM-5 98, 103, 104, 165, 169, 215, 274, 354,
154–156, 164, 485 424, 433, 435
translational 269, 270