You are on page 1of 397

P PR RO OP PA AN NE E R RE EF FO OR RM MI IN NG G U UN ND DE ER R C CA AR RB BO ON N- -

I IN ND DU UC CE ED D D DE EA AC CT TI IV VA AT TI IO ON N: :
C CA AT TA AL LY YS ST T D DE ES SI IG GN N A AN ND D R RE EA AC CT TO OR R
O OP PE ER RA AT TI IO ON N
Kelfin Martino Hardiman
A thesis submitted in fulfilment
of the requirements for the degree of
Doctor of Philosophy
School of Chemical Sciences and Engineering
The University of New South Wales
February 2007
COPYRIGHT STATEMENT
I hereby grant the University of New South Wales or its agents the right to archive and
to make available my thesis or dissertation in whole or part in the University libraries in
all forms of media, now or here after known, subject to the provisions of the Copyright
Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to
use in future works (such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in
Dissertation Abstract International (this is applicable to doctoral theses only).
I have either used no substantial portions of copyright material in my thesis or I have
obtained permission to use copyright material; where permission has not been granted I
have applied/will apply for a partial restriction of the digital copy of my thesis or
dissertation.'
Signed ...........................
Date ...........................
AUTHENTICITY STATEMENT
I certify that the Library deposit digital copy is a direct equivalent of the final officially
approved version of my thesis. No emendation of content has occurred and if there are
any minor variations in formatting, they are the result of the conversion to digital
format.
Signed ...........................
Date ...........................
ii
C Ce er rt ti if fi ic ca at te e o of f O Or ri ig gi in na al li it ty y
I hereby declare that this submission is my own work and to the best of my knowledge
it contains no materials previously published or written by another person, or substantial
proportions of material which have been accepted for the award of any other degree or
diploma at UNSW or any other educational institution, except where due
acknowledgement is made in the thesis. Any contribution made to the research by
others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in
the thesis. I also declare that the intellectual content of this thesis is the product of my
own work, except to the extent that assistance from others in the projects design and
conception or in style, presentation and linguistic expression is acknowledged.
.
iii
A Ab bs st tr ra ac ct t
Steam reforming is the most economical and widely-used route for the conversion of
light hydrocarbon (such as natural gas) to various valued-added products. This process
is commonly carried out over a low-cost alumina-supported nickel catalyst, which often
suffers from carbon deposition resulting in loss of active sites, flow and thermal
maldistribution, as well as excessive pressure drop.
A bimetallic catalyst with improved anti-coking properties was formulated by
incorporating the nickel-based system (15% loading) with cobalt metal (5% loading).
Two-level factorial design was employed to investigate the effect of major preparation
variables, namely impregnation pH value (2-8), calcination temperature (873-973 K),
heating rate (5-20 K min
-1
) and time (1-5 h).
The catalysts prepared were subjected to various characterisation techniques to
determine key physicochemical properties (i.e. BET area, H
2
-chemisorption and NH
3
-
TPD acidity). X-ray diffraction revealed that NiO, Co
3
O
4
, NiCo
2
O
4
and a proportion of
Ni(Co)Al
2
O
4
aluminates were transformed during H
2
-reduction to active Co and Ni
crystallites. TEM images showed an egg yolk profile in the low-pH catalyst suggesting
that main deposition site was located in the particle centre, while metal deposition
occurred primarily around the particle exterior for the high-pH catalyst. Temperature-
programmed experiments were carried out to examine the extent of conversion, type of
iv
surface species and solid-state kinetics (using the Avrami-Erofeev model) involved
during various stages in catalyst life-cycle (calcination, reduction, oxidation and
regeneration). Steam reforming analysis suggested that enhanced catalyst activity may
be due to synergism in the Co-Ni catalyst. Specifically, the low-pH catalyst exhibited
better resistance towards carbon-induced deactivation than the high-pH formulation.
The study also provided the first attempt to develop a quantitative relation between
catalyst preparation conditions and its performance (activity, product selectivity and
deactivation) for steam reforming reaction.
Deactivation and reforming kinetic coefficients were simultaneously evaluated from
propane reforming conversion-time data under steam-to-carbon ratios of 0.8-1.6 and
reaction temperatures between 773-873 K. The time-dependent optimum operational
policy derived based on these rate parameters gave better conversion stability despite
the heavy carbon deposit. Thermal runs further showed that the catalysts regenerated via
two-stage reductive-oxidative coke burn-off exhibited superior surface properties
compared to those rejuvenated by a single-step oxidation.
v
A Ac ck kn no ow wl le ed dg gm me en nt ts s
I express my sincere gratitude to my supervisor, Professor Adesoji A. Adesina, for the
constant guidance and encouragement throughout the course of my research study. I
highly appreciate his compassion for imparting me with valuable knowledge and skills.
The many sacrifices he made to stay after-hours in the office for brainstorming and
paper writing will remain memorable for many years to come.
My special appreciations are due to my senior colleagues, Dr. Tomasz Safinski, Dr.
Praharso and Dr. Francisco Trujillo, for their professional support and advice. I wish
also to thank professional officers, Mr. Andrew Chau, Mrs. Kate Nasev, Mr. John
Franklin and Mr. Craig Howie, and everyone in the Reactor Engineering and
Technology Group for creating an enjoyable and supportive environment.
To my annual review panel, A/Prof. Michael Brungs, Dr. Frank Lucien, Dr. Kingsley
Opoku-Gyamfi, I would like to acknowledge their kind efforts in providing me with
useful and positive feedbacks.
Finally, I would like to thank the University of New South Wales for awarding me a
University Postgraduate Award (UPA) and the Faculty of Engineering for the grant of a
Supplementary Engineering Award (SEA).
vi
L Li is st t o of f P Pu ub bl li ic ca at ti io on ns s
Journals/Books
1. Hardiman, K.M., Mohammed, M.M. and Adesina, A.A. (2003). Deactivation
kinetics of cobalt-nickel catalysts in a fluidised bed reformer, in: Jackson, S.D.,
Hargreaves, J.S.J. & Lennon, D. (Eds.), Catalysis in Application. The Royal
Society of Chemistry, Cambridge, UK, pp. 16-23.
2. Hardiman, K.M., Tan, T.Y., Adesina, A.A., Kennedy, E.M. and Dlugogorski, B.Z.
(2004). Performance of a Co-Ni catalyst for propane reforming under low steam-to-
carbon ratios. Chem. Eng. J., 102, 119-130.
3. Hardiman, K.M., Cooper, C.G. and Adesina, A.A. (2004). Multivariate analysis of
the role of preparation conditions on the intrinsic properties of a Co-Ni/Al
2
O
3
steam
reforming catalyst. Ind. Eng. Chem. Res., 43, 6006-6013.
4. Hardiman, K.M., Trujillo, F.J. and Adesina, A.A. (2005). Deactivation-influenced
propane steam reforming: reactor analysis and parameter estimation. Chem. Eng.
Proc., 44, 987-992.
5. Hardiman, K.M., Hsu, C.-H., Ying, T.T. and Adesina, A.A. (2005). The influence
of impregnating pH on the postnatal and steam reforming characteristics of a Co-
Ni/Al
2
O
3
catalyst. J. Mol. Catal. A: Chem., 239, 41-48.
6. Hardiman, K.M., Hsu, C.-H. and Adesina, A.A. (2006). A mechanistic model for
propane steam reforming on a bimetallic Co-Ni catalyst in fluidized bed reactor, in:
Rhee, H.-K., Nam, I.-S. & Park, J.M. (Eds.), Studies in Surface Science and
Catalysis. Elsevier B.V., Amsterdam.
7. Hardiman, K.M., Cooper, C.G., Adesina, A.A. and Lange, R. (2006). Post-
mortem characterization of coke-induced deactivated alumina-supported Co-Ni
catalysts. Chem. Eng. Sci., 61, 2565-2573.
vii
8. Hardiman, K.M. and Adesina, A.A. (2007). Comparative evaluation of
gasification of carbon deposits over alumina-supported Co, Ni and Co-Ni catalysts
using O
2
, H
2
and inert gases. In preparation.
9. Hardiman, K.M., Alenazey, F.S.M., Adesina, A.A. and Lange, R. (2007). Optimal
operation of a fluidised bed steam reformer with decaying catalyst. In preparation.
International Conferences
10. Hardiman, K.M., Mohammed, M.M. and Adesina, A.A. (2003). Deactivation
kinetics of Co-Ni catalysts in a fluidised bed reformer. International Symposium on
Applied Catalysis. Glasgow (Scotland), July 16-18.
11. Hardiman, K.M. and Adesina, A.A. (2003). Synthesis and characterisation of Co-
Ni catalysts on -Al
2
O
3
support for light hydrocarbon reforming. EuropaCat-VI.
Innsbruck (Austria), August 31-September 4.
12. Hardiman, K.M., Ying, T.T., Mohammed, M.M. and Adesina, A.A. (2003).
Coking dynamics in a fluidised bed reactor for propane steam reforming. 31st
Annual Australasian Chemical Engineering Conference (Chemeca 2003). Adelaide
(Australia), September 29-October 1.
13. Hardiman, K.M., Trujillo, F.J. and Adesina, A.A. (2003). A generalization of the
Levenspiel coupled reaction-deactivation model for reactor analysis. 31st Annual
Australasian Chemical Engineering Conference (Chemeca 2003). Adelaide
(Australia), September 29-October 1.
14. Hardiman, K.M. and Adesina, A.A. (2004). Monometallic Co/Al
2
O
3
, Ni/Al
2
O
3
and bimetallic Co-Ni/Al
2
O
3
catalyst systems for propane steam reforming:
characterisation, activity and stability studies. 13
th
International Congress on
Catalysis. Paris (France), July 11-16.
15. Hardiman, K.M., Cheng-Han, H. and Adesina, A.A. (2005). A mechanistic model
for propane steam reforming on a bimetallic Co-Ni catalyst in fluidized bed reactor.
The 4
th
Asia-Pacific Chemical Reaction Engineering Symposium. Gyeongju
(Korea), June 12-15.
viii
16. Hardiman, K.M., Cooper, C.G. and Adesina, A.A. (2005). Carbon deposition over
Co-Ni/Al
2
O
3
during hydrocarbon reforming. 7
th
World Congress of Chemical
Engineering. Glasgow (Scotland), July 10-14.
ix
T Ta ab bl le e o of f C Co on nt te en nt ts s
COPYRIGHT AND AUTHENTICITY STATEMENTS ii
CERTIFICATE OF ORIGINALITY iii
ABSTRACT iv
ACKNOWLEDGEMENTS vi
LIST OF PUBLICATIONS vii
TABLE OF CONTENTS x
LIST OF FIGURES xv
LIST OF TABLES xxiv
LIST OF ABBREVIATIONS xxix
Chapter 1 INTRODUCTION 1
References 4
Chapter 2 LITERATURE REVIEW 6
2.1 Introduction 6
2.2 Catalytic steam reforming of hydrocarbons 7
2.2.1 Kinetics 9
2.2.2 Reactions associated with steam reforming 13
2.2.2.1 Propane dehydrogenation 13
2.2.2.2 Boudouard 13
2.2.2.3 Water-gas shift 15
2.3 Steam reforming catalysts 17
2.3.1 Metal activity 17
2.3.2 Effect of support 22
2.3.3 Bimetallic catalysts 28
2.3.4 Deactivation of steam reforming catalysts 32
2.3.4.1 Poisoning 34
2.3.4.2 Sintering 36
2.3.4.3 Fouling 40
x
2.4 Carbon deposition 42
2.4.1 Mechanisms of carbon lay-down 43
2.4.2 Properties and reactivity 46
2.5 Summary 51
References 53
Chapter 3 EXPERIMENTAL 67
3.1 Introduction 67
3.2 Materials 67
3.2.1 Chemicals 67
3.2.2 Gases 68
3.3 Catalyst preparation 68
3.3.1 Catalyst support 69
3.3.2 Treatment of support 70
3.3.3 Impregnation catalyst preparation 72
3.3.4 Monometallic catalysts 73
3.3.5 Bimetallic catalysts 74
3.4 Catalyst characterisation 75
3.4.1 Surface area 75
3.4.2 Hydrogen chemisorption 80
3.4.3 Temperature-programmed desorption 85
3.4.4 Thermal gravimetric analysis 88
3.4.4.1 Calcination 89
3.4.4.2 Temperature-programmed reduction and oxidation 90
3.4.4.3 Reproducibility analysis 93
3.4.4.4 Solid-state kinetic analysis 93
3.4.5 X-ray diffraction 94
3.4.6 Transmission electron microscopy 96
3.4.7 Scanning electron microscopy 99
3.4.8 Total carbon analysis 101
3.5 Experimental apparatus 101
3.5.1 Flow controller units 102
3.5.2 Steam generator 103
xi
3.5.3 Fluidised bed reactor rig 103
3.5.4 Product analysis 104
References 107
Chapter 4 PRELIMINARY WORK AND SYSTEM
CHARACTERISATION 110
4.1 General considerations 110
4.2 Thermodynamics 111
4.3 Preliminary experiments 118
4.3.1 Blank runs 118
4.3.2 Product analysis 119
4.3.3 GC calibration 119
4.3.4 Optimum fluidisation flow rate 121
4.3.4.1 Experimental 122
4.3.4.2 Theoretical 123
4.4 Transport resistance considerations 126
4.4.1 External mass transfer 127
4.4.2 Pore diffusion limitation 128
4.4.3 External heat transfer 129
4.4.4 Intraparticle heat transfer 131
4.5 Experimental data treatment 131
4.5.1 Conversion 131
4.5.2 Reaction rate 132
4.5.3 Selectivity 132
4.6 Range of experimental variables 133
References 134
Chapter 5 BIMETALLIC CATALYST DESIGN 136
5.1 General considerations 136
5.2 Catalyst preparation 137
5.3 Catalyst activation 140
5.3.1 Calcination 140
5.3.2 Reduction 144
xii
5.4 Catalyst characterisation 145
5.4.1 Physicochemical attributes 146
5.4.1.1 BET surface area 146
5.4.1.2 Chemisorptive properties 150
5.4.1.3 Acidity 156
5.4.1.4 Model building and optimisation 163
5.4.2 Structural (bulk and surface) studies 166
5.4.2.1 X-ray diffractograms 166
5.4.2.2 Transmission electron microscopy 168
5.4.2.3 Scanning electron microscopy 170
5.4.3 Thermal analyses 172
5.4.3.1 Calcination 172
5.4.3.2 Reduction-oxidation 182
5.5 Catalyst activity evaluation 201
5.5.1 Effect of impregnation pH 202
5.5.2 Effect of calcination variables 205
5.5.3 Comparison with monometallic Co and Ni catalysts 211
5.6 Post-mortem characterisation 213
5.6.1 Total carbon 213
5.6.2 X-ray diffractions 216
5.6.3 Carbon burn-off runs 217
Nomenclature 226
References 228
Chapter 6 REACTOR STUDIES 234
6.1 General considerations 234
6.2 Steady-state kinetics 235
6.2.1 Power law 236
6.2.2 Langmuir-Hinshelwood and Eley-Rideal 241
6.3 Deactivation runs 246
6.3.1 Development of coupled reaction-deactivation model 246
6.3.1.1 Generalisation of coupled reaction-deactivation model 247
6.3.1.2 Model discrimination 256
xiii
6.3.2 Deactivation-influenced steam reforming analysis 265
6.3.3 Post-mortem catalyst examination 280
6.3.3.1 General characterisation 280
6.3.3.2 Thermogravimetric analysis 285
6.3.3.3 Kinetic investigation of coke burn-off 298
6.4 Reactor optimisation 302
6.4.1 Theoretical considerations 302
6.4.2 Model implementation 305
6.4.3 Post-mortem characterisation 320
Nomenclature 323
References 327
Chapter 7 CONCLUSIONS AND RECOMMENDATIONS 332
7.1 Conclusions 332
7.2 Recommendations 337
Appendix ASAMPLE CALCULATION FOR PREPARATION OF
STOCK SOLUTION DURING CATALYST PREPARATION 340
Appendix B METHODOLODY FOR SOLVING COMPLEX CHEMICAL
-REACTION EQUILIBRIA 342
Appendix CTRANSPORT RESISTANCE CALCULATIONS 347
Appendix DCALCULATIONS OF THEORETICAL CATALYST
WEIGHT CHANGE DURING THERMOGRAVIMETIC
RUN 350
Appendix E DEVELOPMENT OF LANGMUIR-HINSHELWOOD AND
ELEY-RIDEAL KINETIC MODELS 353
Appendix F EXAMPLES OF DERIVATION FOR LEVENSPIEL
COUPLED REACTION-DEACTIVATION MODEL 359
Appendix GCONCENTRATION-TIME DATA FOR VARIOUS S:C
RATIO AND TEMPERATURE 363
xiv
L Li is st t o of f F Fi ig gu ur re es s
Fig. 2.1 Methane conversion as a function of nickel loading on o-
alumina. Reaction conditions: temperature, 1023 K; CH
4
:H
2
O
ratio, 2.2:1; pressure, 2.2 MPa [El Solh et al., 2001].
20
Fig. 2.2 Methane conversion with time-on-stream during methane
steam reforming over Ni/Ce-ZrO
2
. Reaction conditions:
temperature, 1023 K; CH
4
:H
2
O:N
2
ratio, 1:3:1 [Dong et al.,
2002]. 22
Fig. 2.3 Effect of supports on CH
4
conversion and product
concentrations for the steam reforming of CH
4
over nickel-
loaded catalysts. Reaction conditions: temperature, 873 K;
reaction time, 0.5 h; CH
4
:H
2
O:Ar ratio, 1:3:5 [Nakagawa et al.,
2003]. 24
Fig. 2.4 Arrangement of platinum atoms in a cluster of about 1 nm in
size [Sinfelt, 2002]. 28
Fig. 2.5 Arrangement of atoms in a bimetallic cluster about 1 nm in
size. The two kinds of atoms are represented by the black and
white spheres [Sinfelt, 2002]. 29
Fig. 2.6 Time scale of deactivation of various catalytic processes
[Moulijn, 2001]. 33
Fig. 2.7 Effect of H
2
S poisoning on the methanation activity of various
metals [Forzatti and Lietti, 1999]. 36
Fig. 2.8 Two conceptual models for crystalline growth due to sintering
by: (a) atomic migration; and (b) crystalline migration
[Bartholomew, 2001]. 37
xv
Fig. 2.9 Normalised nickel surface area (based on H
2
adsorption) versus
time data during sintering of Ni/SiO
2
in hydrogen atmosphere
[Bartholomew, 2001]. 38
Fig. 2.10 Illustration of pore closure at support due to sintering [Adapted
from Fogler, 1999]. 38
Fig. 2.11 Illustration of carbon deposition at support [Adapted from
Fogler, 1999]. 41
Fig. 2.12 Fouling, crystallite encapsulation and pore plugging of a
supported metal catalyst due to carbon deposition
[Bartholomew, 2001]. 42
Fig. 2.13 Formation, transformation and gasification of carbon on nickel
(a, g, s refer to adsorbed, gaseous and solid states, respectively)
[Bartholomew, 2001]. 43
Fig. 2.14 Formation and transformation of coke on metal surfaces (a, g, s
refer to adsorbed, gaseous and solid states, respectively)
[Bartholomew, 2001]. 44
Fig. 2.15 Carbon formation on supported metal catalyst [Moulijn et al.,
2001]. 45
Fig. 2.16 Equilibrium carbon deposition boundaries for amorphous and
graphitic carbons at 723 K and 1.4 atm [Bartholomew, 1982]. 46
Fig. 2.17 Temperature-programmed gasification of different types of
coked catalysts. Operating conditions: heating rate, 5 K min
-1
;
CO
2
flow rate, 0.01 mol min
-1
[Figueiredo, 1986]. 50
Fig. 3.1 Typical adsorption and desorption peaks during a BET
measurement. 80
Fig. 3.2 Typical desorption profiles of different heating rate. 88
Fig. 3.3 Typical catalyst weight profile during calcination. 89
Fig. 3.4 Typical catalyst weight derivative profile during calcination. 90
xvi
Fig. 3.5 Typical catalyst weight profile during reduction-oxidation. 91
Fig. 3.6 Typical catalyst weight derivative profile during reduction. 92
Fig. 3.7 Typical catalyst weight derivative profile during oxidation. 92
Fig. 3.8 Reproducibility analysis of the TGA experiment. 93
Fig. 3.9 Schematic of a transmission electron microscope (TEM)
[Adapted from Crorkendorff and Niemantsverdriet, 2003]. 98
Fig. 3.10 Schematic of a scanning electron microscope (SEM) [Adapted
from Crorkendorff and Niemantsverdriet, 2003]. 100
Fig. 3.11 Schematic diagram of the experimental set-up. 102
Fig. 3.12 Fluidised bed reactor. 106
Fig. 4.1 Thermodynamic characteristics of propane steam reforming
and related reactions. 112
Fig. 4.2 Effect of pressure on equilibrium composition for propane
steam reforming. Reaction conditions: temperature, 800 K; S:C
ratio, 1. 116
Fig. 4.3 Effect of temperature on equilibrium composition for propane
steam reforming. Reaction conditions: pressure, 1 bar; S:C
ratio, 1. 117
Fig. 4.4 Effect of steam-to-carbon ratio (S:C) on equilibrium
composition for propane steam reforming. Reaction conditions:
temperature, 800 K; pressure, 1 bar. 118
Fig. 4.5 Typical chromatogram. 121
Fig. 4.6 Hydrodynamic behaviour of fluidised bed during fluidisation. 123
Fig. 5.1 potential as a function of the pH of the solution [Heise and
Schwarz, 1985].
138
Fig. 5.2 Transient weight profile during calcination to 873 K (with
heating rate of 20 K min
-1
) for catalyst impregnated at pH 2. 141
xvii
Fig. 5.3 Transient weight profile during reduction of calcined Sample
H8. 145
Fig. 5.4(a) NH
3
-TPD profiles for low-pH (L) catalysts. 159
Fig. 5.4(b) NH
3
-TPD profiles for high-pH (H) catalysts. 160
Fig. 5.4(c) NH
3
-TPD profiles for Catalysts L6, L6Co and L6Ni. 161
Fig. 5.5 XRD patterns for freshly-calcined Catalysts L6, H6, L6Co and
L6Ni. 167
Fig. 5.6 XRD patterns for freshly-reduced Catalysts L6, H6, L6Co and
L6Ni. 168
Fig. 5.7(a) TEM micrograph for reduced Catalyst L6. 169
Fig. 5.7(b) TEM micrograph for reduced Catalyst H6. 170
Fig. 5.8(a) SEM micrograph for reduced Catalyst L6. 171
Fig. 5.8(b) SEM micrograph for reduced Catalyst L6Co. 171
Fig. 5.8(c) SEM micrograph for reduced Catalyst L6Ni. 172
Fig. 5.9(a) Calcination weight derivative profiles for acid L6 and basic H6
catalysts. 173
Fig. 5.9(b) Calcination weight derivative profiles for low-pH (L) catalysts. 175
Fig. 5.9(c) Calcination weight derivative profiles for Catalysts L6, L6Co
and L6Ni. 176
Fig. 5.10 A typical weight-based conversion profile during the
calcination run. 177
Fig. 5.11 Weight profiles of TPR-TPO-TPR-TPO runs for Catalysts L6,
H6, L6Co and L6Ni. 182
Fig. 5.12 TPR-TPO-TPR-TPO weight derivative profiles for acid L6 and
basic H6 catalysts. 185
Fig. 5.13 TPR weight derivative profiles for low-pH (L) catalysts. 188
xviii
Fig. 5.14 TPO weight derivative profiles for low-pH (L) catalysts. 189
Fig. 5.15 TPR weight derivative profiles for Catalysts L6, L6Co and
L6Ni. 192
Fig. 5.16 TPO weight derivative profiles for Catalysts L6, L6Co and
L6Ni. 192
Fig. 5.17 A typical weight-based conversion profile during the reduction
run. 193
Fig. 5.18 A typical weight-based conversion profile during the oxidation
run. 197
Fig. 5.19 Propane conversion profiles for acid L6 and basic H6 catalysts. 203
Fig. 5.20 Propane conversion profiles for Catalysts L6, L6Co and L6Ni. 211
Fig. 5.21 Correlation between k
d
' and carbon content in bimetallic
catalysts.
216
Fig. 5.22 XRD patterns for used Catalysts L6, L6Co and L6Ni. 217
Fig. 5.23 Transient weight profiles during oxidative carbon burn-off runs
of spent Catalysts L6, L6Co and L6Ni. 218
Fig. 5.24 Weight-based conversion profiles during oxidative carbon
burn-off runs of spent Catalysts L6, L6Co and L6Ni. 219
Fig. 5.25 TPO weight derivative profiles for spent Catalysts L6, L6Co
and L6Ni. 220
Fig. 5.26 Correlation between and carbon content.
2
,O reg
k 221
Fig. 5.27 Correlation between and carbon content.
2
,O reg
E 225
Fig. 5.28 Correlation between and carbon content.
2
,O reg
A 225
Fig. 6.1(a) Effect of steam partial pressure on steam reforming rate as a
function of temperature at low S:C range (0.8-1.6). 237
xix
Fig. 6.1(b) Effect of steam partial pressure on steam reforming rate as a
function of temperature at high S:C range (2.4-4.8). 237
Fig. 6.2 Activation energy for steam reforming of various hydrocarbons
(adapted from Praharso et al. [2004]). 239
Fig. 6.3 Comparison between predicted rate and observed rate (power
law). 240
Fig. 6.4 Residual plot for rate of propane steam reforming (power law). 240
Fig. 6.5 Typical propane conversion profile. Reaction conditions:
temperature, 823 K; S:C feed ratio, 0.8:1. 257
Fig. 6.6 Typical propane conversion profile. Reaction conditions:
temperature, 773-873 K; S:C feed ratio, 0.8-1.6. 265
Fig. 6.7(a) Transient selectivity profiles for component: (a) H
2
. 267
Fig. 6.7(b) Transient selectivity profiles for component: (b) CO. 267
Fig. 6.7(c) Transient selectivity profiles for component: (c) CO
2
. 268
Fig. 6.7(d) Transient selectivity profiles for component: (d) CH
4
. 268
Fig. 6.8 Profile of deactivation coefficient, k
d
'. 269
Fig. 6.9 Rates of formation (ln scale) and hydrogenation of C

and C

versus reciprocal temperature [Bartholomew, 1982 & 2001]. 270


Fig. 6.10 Arrhenius plot for the rate of deposition on nickel; 1-butene,
100 torr; hydrogen, 25 torr [Lobo et al., 1972]. 271
Fig. 6.11 Arrhenius plots for the rates of deposition from various
propylene-hydrogen mixtures on nickel [Lobo et al., 1972]. 272
Fig. 6.12 Profile of reaction constant, k'. 273
Fig. 6.13 Surface plots of time-averaged product ratios: (a) H
2
:CO; (b)
H
2
:CO
2
; (c) CO:CO
2
; and (d) H
2
:CH
4
. 276
Fig. 6.14 Surface plots of time-averaged ratios per mole C
3
H
8
consumed
for component: (a) H
2
; (b) CO; (c) CO
2
; and (d) CH
4
. 279
xx
Fig. 6.15 X-ray diffractograms of freshly-calcined, freshly-reduced and
coked catalysts. 283
Fig. 6.16(a) TEM of Co-Ni catalyst: (a) fresh state. 284
Fig. 6.16(b) TEM of Co-Ni catalyst: (b) after reforming at S:C = 0.8. 284
Fig. 6.16(c) TEM of Co-Ni catalyst: (c) after reforming at S:C = 1.6. 285
Fig. 6.17 TPO weight profiles of spent catalysts under reaction S:C and
temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6
and 773 K; and (d) 1.6 and 873 K. 286
Fig. 6.18 TPR weight profiles of spent catalysts under reaction S:C and
temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6
and 773 K; and (d) 1.6 and 873 K. 287
Fig. 6.19(a) TPO weight derivative profiles of spent catalysts under
reaction S:C and temperature of: (a) 0.8 and 773 K; (b) 0.8 and
873 K; (c) 1.6 and 773 K; and (d) 1.6 and 873 K. 291
Fig. 6.19(b) TPR (following TPO) weight derivative profiles of fresh and
spent catalysts under reaction S:C and temperature of: (a) 0.8
and 773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and (d) 1.6
and 873 K. 292
Fig. 6.20(a) TPR weight derivative profiles of spent catalysts under reaction
S:C and temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K;
(c) 1.6 and 773 K; and (d) 1.6 and 873 K. 296
Fig. 6.20(b) TPO (following TPR) weight derivative profiles of spent
catalysts under reaction S:C and temperature of: (a) 0.8 and
773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and (d) 1.6 and
873 K. 296
Fig. 6.20(c) TPR (following TPR-TPO) weight derivative profiles of fresh
and spent catalysts under reaction S:C and temperature of: (a)
0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and (d)
1.6 and 873 K. 297
xxi
Fig. 6.21 New reactor operational policy for various S:C ratio (1.0 and
1.5) and initial temperature cases (773 and 813 K). 307
Fig. 6.22(a) Transient propane conversion profiles under dynamic
operational policy with S:C ratio and initial temperature of: (a)
1.0 and 773 K. 309
Fig. 6.22(b) Transient propane conversion profiles under dynamic
operational policy with S:C ratio and initial temperature of: (b)
1.0 and 813 K. 309
Fig. 6.22(c) Transient propane conversion profiles under dynamic
operational policy with S:C ratio and initial temperature of: (c)
1.5 and 773 K. 310
Fig. 6.22(d) Transient propane conversion profiles under dynamic
operational policy with S:C ratio and initial temperature of: (d)
1.5 and 813 K. 310
Fig. 6.23 Correlation between final propane conversion of dynamic
policy runs with carbon content. 312
Fig. 6.24(a) Transient profiles under dynamic operational policy for product
ratios: (a) H
2
:CO. 314
Fig. 6.24(b) Transient profiles under dynamic operational policy for product
ratios: (b) H
2
:CO
2
. 314
Fig. 6.24(c) Transient profiles under dynamic operational policy for product
ratios: (c) CO:CO
2
. 315
Fig. 6.24(d) Transient profiles under dynamic operational policy for product
ratios: (d) H
2
:CH
4
. 315
Fig. 6.25(a) Transient profiles under dynamic operational policy of
component ratios per mole C
3
H
8
converted for species: (a) H
2
. 318
Fig. 6.25(b) Transient profiles under dynamic operational policy of
component ratios per mole C
3
H
8
converted for species: (b) CO. 318
xxii
Fig. 6.25(c) Transient profiles under dynamic operational policy of
component ratios per mole C
3
H
8
converted for species: (c)
CO
2
. 319
Fig. 6.25(d) Transient profiles under dynamic operational policy of
component ratios per mole C
3
H
8
converted for species: (d)
CH
4
. 319
Fig. 6.26 TPO weight profiles of spent catalysts obtained from optimal
policy dynamic runs at S:C and T
0
of: (a) 1.0 and 773 K; (b)
1.0 and 813 K; (c) 1.5 and 773 K; and (d) 1.5 and 813 K. 322
Fig. 6.27 TPO weight derivative profiles of spent catalysts obtained from
optimal policy dynamic runs at S:C and T
0
of: (a) 1.0 and 773
K; (b) 1.0 and 813 K; (c) 1.5 and 773 K; and (d) 1.5 and 813 K. 322
xxiii
L Li is st t o of f T Ta ab bl le es s
Table 2.1 Reported steam reforming kinetic results over nickel-based
catalysts for various hydrocarbons [Adapted from Rostrup-
Nielsen (1973), as well as more recent studies]. 12
Table 2.2 Effect of metals on steam reforming activity. 19
Table 2.3 Effect of supports on steam reforming activity. 27
Table 2.4 Examples of poisons of industrial catalysts [Forzatti and Lietti,
1999]. 34
Table 2.5 Hutting, Tamman and melting temperatures (K) of common
compounds in heterogenous catalysis [Moulijn et al., 2001]. 40
Table 2.6 Forms and reactivities of carbon species formed by
decomposition of CO on nickel [Bartholomew, 2001]. 51
Table 3.1 List of chemicals used. 68
Table 3.2 List of gases used. 69
Table 3.3 Summary of support characteristics. 71
Table 4.1 Thermodynamic characteristics of propane steam reforming
and related reactions. 112
Table 4.2 Limiting temperatures for propane steam reforming and related
reactions. 113
Table 4.3 Equilibrium constants of propane steam reforming and related
reactions. 114
Table 4.4 GC operating conditions. 120
Table 4.5 Retention times and response factors for various components. 121
xxiv
Table 4.6 Physical properties of catalyst and reaction system at 873 K. 125
Table 5.1 2
4
-factorial design plan for bimetallic catalyst synthesis. 143
Table 5.2 Preparation condition for monometallic catalysts. 144
Table 5.3 BET areas of bimetallic catalysts and corresponding Yates
analysis. 149
Table 5.4 BET areas of monometallic catalysts. 149
Table 5.5 Metal dispersions of bimetallic catalysts and corresponding
Yates analysis. 153
Table 5.6 Metal dispersions of monometallic catalysts. 153
Table 5.7 Metal surface areas of bimetallic catalysts and corresponding
Yates analysis. 154
Table 5.8 Metal surface areas of monometallic catalysts. 154
Table 5.9 Active particle sizes of bimetallic catalysts and corresponding
Yates analysis. 155
Table 5.10 Active particle sizes of monometallic catalysts. 155
Table 5.11 Heats of desorption of bimetallic catalysts and corresponding
Yates analysis. 157
Table 5.12 Heats of desorption of monometallic catalysts. 157
Table 5.13 Amounts of NH
3
desorbed in bimetallic catalysts and
corresponding Yates analysis. 162
Table 5.14 Amounts of NH
3
desorbed in monometallic catalysts. 162
Table 5.15 Coefficients of the regression polynomials for each response
variable. 165
Table 5.16 Values of best factor levels for the bimetallic Co-Ni
catalysts. 166
Table 5.17 Calcination kinetic coefficients, k
s,cal
, and R
2
-values for the
various orders in the Avrami-Erofeev model. 180
xxv
Table 5.18 Yates analysis for calcination kinetic constants (n = 3) of low-
pH (L) catalysts. 181
Table 5.19 Percent weight changes and extents of conversion. 183
Table 5.20 Positions of the TPR peaks as a function of the temperature. 190
Table 5.21 Positions of the TPO peaks as a function of the temperature. 190
Table 5.22 Catalyst reduction kinetic coefficients, k
s,red
, and R
2
-values for
the various orders in the Avrami-Erofeev model. 195
Table 5.23 Yates analysis for reduction kinetic constants (n = 3) of low-
pH (L) catalysts. 196
Table 5.24 Catalyst oxidation kinetic coefficients, k
s,ox
, and R
2
-values for
the various orders in the Avrami-Erofeev model. 199
Table 5.25 Yates analysis for oxidation kinetic constants (n = 3) of low-
pH (L) catalysts. 200
Table 5.26 Deactivation and reforming kinetic parameters for acid L6 and
basic H6 catalysts. 204
Table 5.27 Values for k
d
' of low-pH (L) catalysts and corresponding
Yates analysis.
208
Table 5.28 Values for k' of low-pH (L) catalysts and corresponding Yates
analysis.
209
Table 5.29 Values for k'/k
d
' of low-pH (L) catalysts and corresponding
Yates analysis.
210
Table 5.30 Deactivation and reforming kinetic parameters for Catalysts
L6, L6Co and L6Ni. 212
Table 5.31 Carbon content of low-pH (L) catalysts and corresponding
Yates analysis. 215
Table 5.32 Carbon contents of monometallic catalysts. 216
xxvi
Table 5.33 Kinetic coefficients, , and R
2
,O reg
k
2
-values for oxidative carbon
burn-off runs based on the various orders in the Avrami-
Erofeev model. 221
Table 5.34 Pre-exponential factors, , and activation energies,
, for oxidative carbon burn-off runs.
2
,O reg
A
2
,O reg
E
224
Table 6.1 Power-law parameters. 239
Table 6.2 Langmuir-Hinshelwood (LH) and Eley-Rideal (ER) kinetic
models. 242
Table 6.3 Estimates of kinetic and reaction constants. 243
Table 6.4 Thermodynamic coefficients and BMV assessment. 245
Table 6.5 Arrhenius parameters. 245
Table 6.6 Coupled reaction-deactivation models. 255
Table 6.7 R
2
-values for the coupled reaction-deactivation models. 257
Table 6.8 Deactivation rate coefficients (k
d
' or b) for various coupled
reaction-deactivation models (s
-1
).
259
Table 6.9 Reaction rate coefficients (k') for various coupled reaction-
deactivation models (units depend on reaction order).
260
Table 6.10 Activation energy, Arrhenius coefficients and R
2
-values for
deactivation, k
d
'. 263
Table 6.11 Activation energy, Arrhenius coefficients and R
2
-values for
steam reforming, k'. 264
Table 6.12 Calculated F-values from two-way ANOVA with error
variance as denominator. 274
Table 6.13 BET surface areas, H
2
chemisorptive properties and total
carbon contents of freshly-calcined, freshly-reduced and
regenerated spent catalysts. 281
xxvii
Table 6.14 Weight change analysis of spent catalysts after different
thermal treatments. 288
Table 6.15 Predicted weight changes of spent catalysts with and without
effects resulting from bulk phase oxidative/reductive reactions
upon carbon gasification. 289
Table 6.16 Pre-exponential factors, , activation energies, ,
and kinetic constants, , of the coke burn-off process for
catalysts coked under various conditions.
2
,O reg
A
2
,O reg
E
2
,O reg
k
300
Table 6.17 Parameters used in the Weisz-Prater analysis. 301
Table 6.18 Deactivation and reaction parameters of propane steam
reforming. 305
Table 6.19 Estimate of for the four cases of optimal operational policy
(s
0
d
k
-1
). 306
Table 6.20 Total carbon analysis and final conversion percents following
30 h run time under dynamic temperature operation. 312
Table 6.21 Time-averaged concentration ratios of H
2
:CO, H
2
:CO
2
,
CO:CO
2
and H
2
:CH
4
. 316
Table 6.22 Time-averaged ratios of H
2
, CO, CO
2
and CH
4
per mole C
3
H
8
converted. 320
Table 6.23 Pre-exponential factors, , activation energies, ,
and kinetic constants, , of the coke burn-off process for
catalysts used in optimal policy dynamic runs.
2
,O reg
A
2
,O reg
E
2
,O reg
k
323
xxviii
L Li is st t o of f A Ab bb br re ev vi ia at ti io on ns s
AES Auger electron spectroscopy
AFM Atomic force microscopy
ANOVA Analysis of variance
APS Active particle size
BET Brunauer-Emmett-Teller
BMV Boudart-Mears-Vannice
BOU Boudouard reaction
CAEM Controlled-atmosphere electron microscopy
CSTR Continuous stirred-tank reactor
DHY Dehydrogenation reaction
ER Eley-Rideal kinetic model
GC Gas chromatography
HRSTEM High resolution scanning transmission
electron microscopy
IEP Isoelectric point
LH Langmuir-Hinshelwood kinetic model
LHS Left-hand side
MARI Most abundant reactive species
NMR Nuclear magnetic resonance
PD Percent dispersion
PFR Plug flow reactor
SANS Small-angle neutron scattering
SEM Scanning electron microscopy
SR Steam reforming reaction
TCD Thermal conductivity detector
TEM Transmission electron microscopy
TGA Thermal gravimetric analysis
xxix
TOC Total organic carbon
TOF Turn-over frequency
TON Turn-over number
TPCO
2
Temperature-programmed carbon dioxide
TPD Temperature-programmed desorption
TPO Temperature-programmed oxidation
TPR Temperature-programmed reduction
WGS Water-gas shift reaction
XPS X-ray photoelectron spectroscopy
XRD X-ray diffraction
xxx
Chapter 1: Introduction
C
C
h
h
a
a
p
p
t
t
e
e
r
r
1
1
I
I
N
N
T
T
R
R
O
O
D
D
U
U
C
C
T
T
I
I
O
O
N
N
The necessity for innovative efforts in the process industries has become more
important than ever due to rapidly increasing business competitiveness and tougher
environmental legislations [Rostrup-Nielsen and Alstrup, 1999; Gradassi and Green,
1995]. This is especially pertinent to steam reforming as the most important route for
the conversion of light hydrocarbons (in most cases, natural gas) to a wide range of
value-added products. Steam reforming is a process in which hydrocarbons are
catalytically reacted with steam to produce synthesis gas - a mixture of hydrogen and
carbon monoxide. Although the process is known to be costly in term of capital
investment and operating expenditure, it is industrially well-established and will be
anticipated to take a greater role in the future arising from realistic expectation of the
hydrogen economy.
Even so, Tsang et al. [1995] addressed a number of drawbacks to the current application
of steam reforming associated with the generation of excess superheated steam required
to suppress carbon lay-down and the presence of water-gas shift reaction as dictated by
its thermodynamics resulting in the conversion of high-valued carbon monoxide to
much cheaper carbon dioxide. Consequently, the operation of hydrocarbon steam
1
Chapter 1: Introduction
reformer under lower steam-to-carbon (S:C) ratio represents an attractive proposition
since the high-energy consumption pertaining to steam production is reduced and also,
the formation of CO
2
via water-gas shift is minimised [Yamazaki et al., 1996; Aasberg-
Petersen et al., 2001]. This, however, makes the reactor operation more susceptible to
coke-induced deactivation. Moreover, the propensity of carbon-forming reactions
substantially increases as feed used involves hydrocarbon substrate with higher carbon
number and/or unsaturated bonds. Accordingly, the strategy of decreasing S:C ratio will
be viable only in circumstances where issues related to carbon accumulation can be
effectively dealt with.
From a thermodynamic standpoint, carbon formation is indeed an inevitable
pathological phenomenon which occurs despite of high feed S:C ratio (> 3) during
prolonged operation. This is particularly true over nickel-based catalysts commonly
used in industrial reformers because of their relatively high activity and low cost.
Carbonaceous residues may lead to encapsulation of active metal crystallite (resulting in
activity loss) and formation of filamentous carbon with metal particle on the top
[Moulijn et al., 2001]. Eventually, further growth of coke filament gives rise to creation
of hot spots, pressure drop, flow maldistribution and ultimately, reactor blockage
[Rostrup-Nielsen, 1984]. As a result, carbon lay-down may be regarded as one of the
most significant limitations in a commercial steam reformer. In fact, Rostrup-Nielsen
[2000] has emphasised that coking control holds the essential key for unlocking the full
potential of the steam reforming process.
2
Chapter 1: Introduction
In addition to coke build-up, steam reforming conducted in traditional fixed-bed steam
reformers is also significantly restricted by mass transfer and thermodynamic
constraints [Chen et al., 2003]. Nevertheless, the implementation of fluidised bed
system has been demonstrated to give numerous competitive advantages over the fixed-
bed reactor in minimising these limitations as well as reducing carbon yield [Kunii and
Levenspiel, 1977; Effendi et al., 2002].
Specifically, the overall thrust of this thesis was the development of a catalyst system
with anti-coking attributes and the operation of steam reforming under simultaneous
carbon-induced deactivation. Consequently, the major objectives of this project were to:
1. Obtain a quantitative relationship between preparation factors and
physicochemical attributes of a bimetallic alumina-supported Co-Ni
reforming catalyst synthesised via impregnation.
2. Investigate the occurrence of possible synergistic effects by comparison with
individual Co/alumina and Ni/alumina catalysts.
3. Obtain deactivation and reforming kinetic parameters from reaction runs
under varying reaction conditions and over different catalysts using an
appropriate coupled reaction-deactivation model derived based on
Levenspiels methodology [Levenspiel, 1999].
4. Develop and implement a time-dependent optimal operational policy for
maintaining constant conversion level during low S:C reforming operation.
5. Explore the effect of reaction conditions and catalyst type on the nature,
property and reactivity of carbonaceous residues formed.
3
Chapter 1: Introduction
The outcome of the catalysis study in this project would demonstrate the effect of major
preparation variables on the characteristics and deactivation-influenced activity
performance of a bimetallic catalyst which incorporates the highly-active nickel-based
system with relatively low-cost coke-suppressant cobalt metal. The development of
dynamic operational policy based on deactivation and reaction kinetics would result in
prolonged time-on-stream, even under activity loss. Additionally, these techniques
would have the potential to bring about reduction in steam usage of a conventional
reforming process. Obviously, this is a major economic benefit since the energy-
intensive option for excess steam generation is often required to minimise coking.
References
Aasberg-Petersen, K., Bak Hansen, J.-H., Christensen, T.S., Dybkjaer, I., Seier
Christensen, P., Stub Nielsen, C., Winter Madsen, S.E.L. and Rostrup-Nielsen,
J.R. (2001). Technologies for large-scale gas conversion. Appl. Catal. A: Gen.,
221, 379-387.
Chen, Z., Yan, Y. and Elnashaie, S.S.E.H. (2003). Novel circulating fast fluidized-bed
membrane reformer for efficient production of hydrogen from steam reforming
of methane. Chem. Eng. Sci., 58, 4335-4349.
Effendi, A., Zhang, Z.-G., Hellgardt, K., Honda, K. and Yoshida, T. (2002). Steam
reforming of a clean model biogas over Ni/Al
2
O
3
in fluidized- and fixed-bed
reactors. Catal. Today, 77, 181-189.
Gradassi, M.J. and Green, N.W. (1995). Economics of natural gas conversion processes.
Fuel Process. Technol., 42, 65-83.
Kunii, D. and Levenspiel, O. (1977). Fluidization Engineering. R. E. Krieger Pub. Co.,
Huntington, N.Y.
4
Chapter 1: Introduction
Levenspiel, O. (1999). Commentaries: Chemical reaction engineering. Ind. Eng. Chem.
Res., 38, 4140-4143.
Moulijn, J.A., van Diepen, A.E. and Kapteijn, F. (2001). Catalyst deactivation: is it
predictable? What to do? Appl. Catal. A: Gen., 212, 3-16.
Rostrup-Nielsen, J.R. (1984). Catalytic steam reforming, in: Anderson, J.R. & Boudart,
M. (Eds.), Catalysis: Science and Technology. Springer-Verlag, Berlin, pp. 1-
117.
Rostrup-Nielsen, J.R. and Alstrup, I. (1999). Innovation and science in the process
industry: Steam reforming and hydrogenolysis. Catal. Today, 53, 311-316.
Rostrup-Nielsen, J.R. (2000). New aspects of syngas production and use. Catal. Today,
63, 159-164.
Tsang, S.C., Claridge, J.B. and Green, M.L.H. (1995). Recent advances in the
conversion of methane to synthesis gas. Catal. Today, 23, 3-15.
Yamazaki, O., Tomishige, K. and Fujimoto, K. (1996). Development of highly stable
nickel catalyst for methane-steam reaction under low steam to carbon ratio.
Appl. Catal. A: Gen., 136, 49-56.
5
Chapter 2: Literature Review
C
C
h
h
a
a
p
p
t
t
e
e
r
r
2
2
L
L
I
I
T
T
E
E
R
R
A
A
T
T
U
U
R
R
E
E
R
R
E
E
V
V
I
I
E
E
W
W
2.1 Introduction
The availability of synthesis gas (H
2
/CO mixture) is an important issue in many value-
added petrochemical processes, for instance methanol production, ammonia
manufacture and the Fischer-Tropsch reaction for higher hydrocarbons synthesis
[Rostrup-Nielsen, 1997 & 2000; Twigg, 1994]. There are a number of primary routes
for the conversion of hydrocarbon to synthesis gas [Rostrup-Nielsen, 2000]. The first
and most commonly used pathway is steam reforming of methane, which is given by
2 2 4
3H CO O H CH + = + = 205.8 kJ mol
0
298
H A
-1
(2.1)
or, for a more general class of hydrocarbon feed,
2 2
1
2
H ) ( CO O H H C m n n n
m n
+ + = + = 498.6 kJ mol
0
298
H A
-1
(for C
3
H
8
) (2.2)
The second route is known as CO
2
reforming, where CO
2
is utilised instead of H
2
O to
react with hydrocarbons. Thus,
6
Chapter 2: Literature Review
2 2 4
2H CO 2 CO CH + = + = 247.0 kJ mol
0
298
H A
-1
(2.3)
Unlike the previous two pathways, the third route is slightly exothermic releasing a
small amount of energy for each hydrocarbon converted. This reaction is expressed as,
2 2 2
1
4
2H CO O CH + = + = -36.0 kJ mol
0
298
H A
-1
(2.4)
Steam reforming of light alkanes is the preferred as the most reliable route for the
industrial manufacture of synthesis gas (H
2
/CO mixture). The capital and operational
investments of a typical steam reformer may constitute about 60% of the overall cost of
a petrochemical production [De Groote et al., 1996; Rostrup-Nielsen, 2002]. In order to
reduce costs and increase the current efficiency level, research efforts have been
directed to the development of new catalysts and reactor technologies which would lead
to improvements and hence, increased profit margin.
2.2 Catalytic steam reforming of hydrocarbons
The commercial application of catalytic steam reformer was pioneered by BASF in
1926 [Marschner et al., 2000]. This led to further development of reforming process
using tubular reactor furnaces by 1930. In the United States, first tubular steam
reformers were also commissioned in 1930 by Standard Oil Company for producing
hydrogen, which was used for hydrogenation purposes and as a feedstock for ammonia
production [Byrner and Gohr, 1932; Haslam and Russel, 1930; Murphree et al., 1994].
Around the same time, ICI initiated the reforming technology in the Great Britain with
7
Chapter 2: Literature Review
the operation of the Billingham plant in 1936 [Synetix, undated]. A similar progress was
followed by Haldor Topsoe, a Denmark-based hydrocarbon processing and catalysis
company with their first reformer started in 1956.
Subsequently, the reforming technology has gone through a variety of developments
resulting in stronger reactor materials, better operability, enhanced coking controls and
improved catalysts. One of the most notable breakthroughs is known as the autothermal
reforming in which the high-energy requirement of steam reforming is supplied by heat
released from partial combustion of hydrocarbon with oxygen or air premixed to the
reactor feed. With cheaper and improved performance of both reforming catalysts and
reformer operations, steam reforming is expected to continue to be in strong demand
into the foreseeable future [Rostrup-Nielsen, 2000].
The steam reforming of propane is given by
2 2 8 3
7H 3CO O 3H H C + = + = 498.6 kJ mol
0
298
H A
-1
(2.5)
while carbon deposition may occur via propane dehydrogenation
2 4 8 3
2H 2C CH H C + + = = 30.2 kJ mol
0
298
H A
-1
(2.6)
or CO disproportionation (Boudouard) reaction
8
Chapter 2: Literature Review
2
CO C 2CO + = = -172.5 kJ mol
0
298
H A
-1
(2.7)
with possible water-gas shift reaction (WGS)
2 2 2
H CO O H CO + = + = -41.2 kJ mol
0
298
H A
-1
(2.8)
2.2.1 Kinetics
Literature results of steam reforming kinetics using various hydrocarbons over nickel-
based catalysts are summarised in Table 2.1. It is manifest that kinetic attributes are a
strong function of support type, hydrocarbon nature and choice of reforming conditions
(temperature and pressure) [Al-Ubaid, 1986]. Fuelled by the economic potential of
cheap and widely available natural gas, it is not surprising that a large number of
reforming studies utilise methane as hydrocarbon substrate. The pioneering work by
Akers and Camp [1955] revealed a first-order reforming dependence on methane with
an E-value of 36.8 kJ mol
-1
. Ross and Steel [1973] investigated methane reforming at
lower pressures in the range 0-10 mmHg. The orders of reaction rate were 1 and -0.5
with respect to CH
4
and steam respectively, with a corresponding activation energy of
29 kJ mol
-1
. These authors concluded that methane adsorption was the rate-limiting step
and methane competes with steam for adsorption sites. In another study by Numaguchi
and Kikuchi [1988], higher temperatures and pressures were employed for methane
steam reforming in a fixed bed reactor. This study showed a substantially-higher
activation energy (106.9 kJ mol
-1
) with reaction orders of 0 and 0.6 with respect to
methane and steam respectively. Recently, Laosiripojana and Assabumrungrat [2005]
also carried out methane steam reforming using nickel metal on Ce-ZrO
2
support. The
9
Chapter 2: Literature Review
reaction order in methane was 1, whilst steam possessed a negative effect on the
reforming rate with a reaction order of -0.4. The associated activation energy from their
experimental data was 142 kJ mol
-1
.
Rostrup-Nielsen [1975] and Kneale and Ross [1981] observed similar results for steam
reforming of ethane in a differential reactor. From the investigation of Rostrup-Nielsen
[1975], rate dependency orders of 0.54 and -0.33 on ethane and steam respectively were
obtained along with an activation energy of 75.7 kJ mol
-1
. In comparison, Kneale and
Ross [1981] suggested that ethane reforming was characterised by reaction orders of 0.7
and -0.5 with respect to ethane and steam respectively. Furthermore, a slightly-lower
activation energy of 69.8 kJ mol
-1
was reported compared to the value given by
Rostrup-Nielsen [1975].
Bhatta and Dixon [1969] conducted an investigation on steam reforming of n-butane at
30 atm over a temperature range of 693-753 K. The reaction rate has first-order
dependence on steam with an E-value of 54.3 kJ mol
-1
. Moreover, a study by Phillips,
Mulhall and Turner [1969] using n-heptane and n-hexane at 633-723 K and 15 atm
demonstrated that steam reforming rate was proportional to 0.3 and 0 with respect to
hydrocarbons and steam respectively, where the corresponding activation energy was
87.8 kJ mol
-1
. Tottrup [1982] has also performed steam reforming of n-heptane at
isothermal conditions in a single pellet string reactor. In his study, the pressure was
varied between 5-30 atm and the temperature in the 723-823 K range. The intrinsic
reaction order was found between 0.1-0.3 for n-heptane, while a negative order of -0.2
10
Chapter 2: Literature Review
11
for steam was obtained. The associated intrinsic activation energy for this reaction was
67.8 kJ mol
-1
.
More recently, Praharso et al. [2004] studied the steam reforming kinetics of iso-octane
at 583-623 K and 1 atm. These researchers observed a reaction order of 0.2 for iso-
octane indicating strong coverage of nickel by iso-octane. On the other hand, the
reaction order with respect to steam was 0.5 which suggested dissociative adsorption of
steam. Furthermore, a regression of their rate data to a power-law model yielded an
activation energy of 44 kJ mol
-1
.
eview
12
Kinetic coefficients
Authors Catalyst Hydrocarbon T (K) P (atm)
m n
H C
o
O H
2
o
Activation
energy
(kJ mol
-1
)
Akers and Camp [1955] Ni/Kiesulguhr Methane 609-911 1 1 0 36.8
Ross and Steel [1973] Ni/Al
2
O
3
Methane 773-953 0-10 mmHg 1 -0.5 29
Numaguchi and Kikuchi [1988] Ni/Al
2
O
3
Methane 674-1160 1.2-25.5 0 0.6 106.9
Laosiripojana and
Assabumrungrat [2005]
Ni/Ce-ZrO
2
Methane 923-1123 1 1 -0.4 142
Rostrup-Nielsen [1975] Ni/MgO Ethane 773 1-31 0.54 -0.33 75.7
Kneale and Ross [1981] Ni/Al
2
O
3
Ethane 476-713 4 mmHg 0.7 -0.5 69.8
Bhatta and Dixon [1969] Ni/-Al
2
O
3
n-Butane 693-753 30 0 1 54.3
Phillips et al. [1969] Ni/-Al
2
O
3
n-Hexane
n-Heptane
633-723 15 0.3 0 87.8
Tottrup [1982] Ni/MgO n-Heptane 723-823 5-30 0.1-0.3 -0.2 67.8
Praharso et al. [2004] Ni/o-Al
2
O
3
iso-Octane 583-623 1 0.2 0.5 44
Table 2.1. Reported steam reforming kinetic results over nickel-based catalysts for various hydrocarbons [Adapted from Rostrup-
Nielsen (1973), as well as more recent studies].
Chapter 2: Literature R
Chapter 2: Literature Review
2.2.2 Reactions associated with steam reforming
Steam reforming of propane is accompanied by several side reactions, such as
dehydrogenation, Boudouard and water-gas shift reactions, affecting composition of
product gas. In this section, important aspects related to these reactions were briefly
discussed.
2.2.2.1 Propane dehydrogenation
The slightly endothermic propane dehydrogenation is represented by:
2 4 8 3
2H 2C CH H C + + = = 30.2 kJ mol
0
298
H A
-1
(2.6)
As shown later from the Gibbs free energy graph (cf. Fig. 4.1), propane
dehydrogenation is thermodynamically feasible under the operating conditions for
propane steam reforming. While carbon formation is generally a slow process at low
temperature, however, its rate accelerates substantially with increasing temperature.
Bartholomew [1982] illustrated the decomposition reaction for a general form of
hydrocarbon as follows
z n y x m n
H C ... H C CH H C H C
2
+ + - + - + - + - -
(2.9)
2.2.2.2 Boudouard
The Boudouard reaction, also known as CO disproportionation, is another carbon-
forming reaction. This is described by:
13
Chapter 2: Literature Review
2
CO C 2CO + = = -172.5 kJ mol
0
298
H A
-1
(2.7)
Although Boudouard reaction may also be catalysed by nickel catalysts; unlike
dehydrogenation, it is thermodynamically favoured at lower temperature.
Mechanistic study of the Boudouard reaction has been studied by Tottrup [1976] and
Rosei et al. [1983]. They postulated that this reaction occurs based on the following
steps
- = - + CO CO (2.10)
- + - +- - O C CO (2.11)
- + - = - + -
2
CO O CO (2.12)
- + = -
2 2
CO CO (2.13)
where - symbolises a surface site and C-, O-, CO- and CO
2
- correspond to
chemisorbed species. Both studies concluded that the dissociation of CO- (cf. Eq. 2.11)
was indeed the rate-determining step. On the other hand, in another investigation Sakai
[1985] substituted the two steps in Eqs. 2.11 and 2.12 by the following reaction
- + - = - + -
2
CO C CO CO (2.14)
which has also been identified as the rate-limiting step by this author.
14
Chapter 2: Literature Review
2.2.2.3 Water-gas shift
Water-gas shift is particularly important where composition of CO, CO
2
and H
2
in the
product gas is critical, for instance in Fischer-Tropsch reaction, methanol and ammonia
syntheses. It is a slightly exothermic reaction, and given by,
2 2 2
H CO O H CO + = + = -41.2 kJ mol
0
298
H A
-1
(2.8)
Thus, the equilibrium position is shifted toward the products under lower temperature;
while, as suggested by Le Chateliers principle, the effect of pressure is not expected to
be critical.
Previous investigators have reported that nickel catalysts possess a moderate water-gas
shift activity under steam reforming environment [Rostrup-Nielsen, 1973;
Ramachandran et al., 1983; Ming et al., 2002; Xu and Froment, 1989]. Mechanism
underlying the water-gas shift reaction has been studied as far back as 1920 [Armstrong
and Hilditch, 1920]. Even so, up till date, there are still two contrasting mechanistic
details: (A) formate intermediate mechanism, or (B) redox mechanism.
The formate intermediate mechanism was supported by researchers who found evidence
for the formation of formate intermediate species. These investigators have used a
variety of techniques, such as chemical probe [Armstrong and Hilditch, 1920], trapping
reagent [Diagne et al., 1990] and infrared spectroscopy [Shido and Iwasawa, 1993;
Jacobs et al., 2004; Noto et al., 1967; Millar et al., 1993; Amenomiya and Pleizer,
1982]. The elementary steps proposed are [Gorte and Zhao, 2005]:
15
Chapter 2: Literature Review
- - + CO CO (2.15)
- + - - + OH H 2 O H
2
(2.16)
- + - - + - HCOO CO OH (2.17)
- + - +- - H CO HCOO
2
(2.18)
- + - 2 H 2H
2
(2.19)
Shido and Iwasawa [1993] and Jacobs et al. [2004] have proposed that the rate-limiting
step is the decomposition of the formate, [HCOO-].
Several studies, however, have argued that the observed formate species was in fact
only a spectator species and not involved in the actual reaction sequences [Deluzarche
et al., 1985; Tibiletti et al., 2004]. Consequently, a second mechanism was proposed
where H
2
O oxidises the surface producing H
2
, followed by subsequent reduction by CO
to yield CO
2
. Accordingly, this is essentially a cyclic redox process as indicated in the
following simplified form [Gorte and Zhao, 2005]:
- - + CO CO (2.20)
2 2
H O O H + - - + (2.21)
- + - + - 2 CO O CO
2
(2.22)
This mechanism was supported by various researchers based mostly on transient kinetic
experiments [Salmi et al., 1988; Gorte and Zhao, 2005; Campbell and Daube, 1987; Li
et al., 2000].
16
Chapter 2: Literature Review
2.3 Steam reforming catalysts
Typical reforming catalysts contain one or more metallic active species deposited on a
support material. Although support is supposedly inert, metal species may possibly
interact with the support in some form of metal-support interaction. As a consequence,
the choice of support may be a significant factor in determining the activity and stability
of a catalytic material.
This section reviews the performance of various metals and supports on steam
reforming activity. The role of bimetallic catalysts in steam reforming will also be
examined along with their application in the mitigation of catalyst deactivation.
2.3.1 Metal activity
Table 2.2 shows the activity ranking for various metals. The numeral preceding the
metal symbol indicates the corresponding loading (in wt.%). As may be seen, most
species used in these studies belong to Group VIII metals. Many of these metals have
earlier been identified as good steam reforming catalysts [Rostrup-Nielsen, 1973].
However, the differences in activities may be attributed to the variation in metal loading
and particular steam reforming conditions employed by the investigators.
Alumina (Al
2
O
3
) was often used as the support due to its low cost, high surface area and
stable mechanical properties. Li et al. [2005] showed the superiority of Ni over Pt and
Pd in alumina-supported systems for oxidative steam reforming of methane. Further
studies by Rostrup-Nielsen [1973], Engler et al. [1991] and Murata et al. [2004]
17
Chapter 2: Literature Review
18
revealed that Ru and Re exhibited better activity compared to Pt, Pd and Rh during
reforming of methane and higher alkanes. Interestingly, Wang et al. [2004a] also
suggested that Ni and Re gave higher conversions than Ru and Rh when
methylcyclohexane was used as the hydrocarbon feed. For steam reforming of
chlorocarbon [Intarajang and Richardson, 1999], ethanol [Breen et al., 2002; Aupretre,
2002; Liguras et al., 2003; Fatsikostas et al., 2002] and toluene [Grenoble, 1978a], there
is a clear indication that Rh is indeed the most active species, whilst Pd and Pt also gave
good conversions.
When metals are deposited on zirconia (ZrO
2
) or ceria (CeO
2
) supports, Hegarty et al.
[1998] and Engler et al. [1991] also found that Pt, Pd and Ni were more active than
other metals studied under steam reforming with S:C > 2. However, for an equimolar
feed (S:C = 1), the Pt catalyst maintained a stable conversion while the activity of the Ni
catalyst dropped rapidly due to coking. Moreover, for MgO-supported catalysts Qin and
Lapszewicz [1994] showed that Ru, Rh and Ir gave higher and more stable activity
among the noble metals. Nakagawa et al. [2003] recently studied the performance of
oxidised-diamond-supported Group VIII metal catalysts for reforming reactions where
Ru and Ni catalysts were reported to result in higher conversions and improved product
yields. These researchers proposed that synergistic effect between the supposedly
coking-susceptible Ni and oxidised diamond may have played a role in minimising
carbon deposition, thus enhancing the overall catalyst activity and stability.
eview
19
Activity order Hydrocarbon S:C Temperature (K) Sources
1 Al
2
O
3
0.4Ni > 0.6Pt > 0.3Pd
0.5Ru > 0.5Pd > 0.5Pt
5Ru > 5Pd > 5Pt
0.88Pt > 0.48Pd > 0.45Rh
10Re > 10Rh > 10Pd
0.5Pt 0.5Rh 0.5Pd >> 0.5Cu > 0.5Re > 0.5Ir > 0.5Ru > 0.5Ni > 0.5Co
1Rh >> 0.5Pd > 5Ni 1Pt
1Rh > 9.7Ni >> 0.75Pd > 1Pt > 8.7Fe > 9.8Zn > 9.1Cu > 0.67Ru
1Rh >> 1Pt > 1Pd > 1Ru
0.5Rh > 20Ni >> 20 Co
5Ni >> 5Ru >> 5Rh
2Re > 2Ni
1Rh >> 1Ru > 1Pd > 2Pt > 2Ir > 5 Ni > 2 Os
Methane
Ethane
Ethane
Propane
iso-Octane
Chlorocarbon
Ethanol
Ethanol
Ethanol
Ethanol
Methylcyclohexane
Methylcyclohexane
Toluene
1.33
4
4
5.08
1.43
10
1.5
1.5
1.5
1.5
1
1
0.46
973
773
773
673
823
873
673-973
973
873-1123
823-1123
853
853
713
Li et al. [2005]
Rostrup-Nielsen [1973]
Rostrup-Nielsen [1973]
Engler et al. [1991]
Murata et al. [2004]
Intarajang et al. [1999]
Breen et al. [2002]
Aupretre et al. [2002]
Liguras et al. [2003]
Fatsikostas et al. [2002]
Wang et al. [2004a]
Wang et al. [2004a]
Grenoble [1978a]
2 ZrO
2
1Pt >> 1Ni
1Pt > 1Pd > 1Ni >> 1Co > 1Cu > 1Fe
Methane
Methane
1
2.64
1073
673-1073
Hegarty et al. [1998]
Hegarty et al. [1998]
3 CeO
2
0.88Pt > 0.45Rh > 0.48Pd Propane 5.08 673 Engler et al. [1991]
4 MgO
0.5Ru > 0.5Rh > 0.5Ir > 0.5Pt > 0.5Pd Methane 0.67 873-1173 Qin et al. [1994]
5 Oxidised-diamond
5Ru > 5Ni > 5Co > 5Ir > 5Rh > 5Pd >> 5Pt > 5Fe Methane 3 873 Nakagawa et al. [2003]
Table 2.2. Effect of metals on steam reforming activity.
Chapter 2: Literature R
Chapter 2: Literature Review
The effect of metal loading on conversion during steam reforming has also been studied
by a number of investigators. By utilising o-alumina-supported nickel catalysts
prepared at different concentrations, de Lasa and co-workers discovered that methane
conversion increased almost linearly with higher bulk nickel concentration and reached
an optimum level at 2.5 wt.% [El Solh et al., 2001]. Interestingly, nickel content at 4%
resulted in lower conversion while further addition to 20% did not lead to significant
improvement. The relationship between nickel loading and activity is further illustrated
in Fig. 2.1.
Fig. 2.1. Methane conversion as a function of nickel loading on o-alumina.
Reaction conditions: temperature, 1023 K; CH
4
:H
2
O ratio, 2.2:1; pressure, 2.2
MPa [El Solh et al., 2001].
20
Chapter 2: Literature Review
Yanhui and Diyong [2001] reported similar behaviour on Ni/Al
2
O
3
used for steam
reforming of n-octane. The catalyst activity increased with increased nickel content in a
range between 1-5 wt.%, while the improvement in activity was not apparent with
further addition in nickel loading at above 5%. El Solh [2001] explained this behaviour
by relating activity to available metal surface area. It was revealed that at the best nickel
loading (which gives optimum conversion), the nickel surface area was found to be the
highest. Therefore, there seems to be an optimal metallic area which can be formed on a
given catalyst surface regardless of the amount of metal added. In fact, overloading
catalyst may cause particle agglomeration and formation of aluminates leading to poorer
dispersion and unstable catalyst.
In a study by Dong et al. [2002], the effect of nickel content on ceria-supported catalysts
was explored, in which the ceria material is known to play a major role in the reaction
mechanism. Fig. 2.2 depicts conversion-time profiles for methane reforming over
Ni/Ce-ZrO
2
with varying metal composition of 3, 10, 15, 20 and 30 wt.%. The 3Ni/Ce-
ZrO
2
showed approximately 58% conversion, which increased to 89 and 97% for 10Ni
and 15Ni/Ce-ZrO
2
respectively. Further addition in nickel concentration resulted in
lower conversion, as seen for 20Ni and 30Ni/Ce-ZrO
2
catalysts with 85 and 61%
respectively. Constant conversion levels were maintained for all catalysts throughout
the experimental duration (200 min). The authors suggested that the Ce-ZrO
2
support
may be involved by becoming an active site for producing activated oxygen species
from H
2
O molecules which is effectively useful for removal of deactivation-causing
carbon residues. Therefore, in the superior 15Ni/Ce-ZrO
2
the composition of the
21
Chapter 2: Literature Review
dispersed metal and the support material are appropriately balanced enhancing its
integrated ability to activate both methane and steam.
Fig. 2.2. Methane conversion with time-on-stream during methane steam
reforming over Ni/Ce-ZrO
2
. Reaction conditions: temperature, 1023 K;
CH
4
:H
2
O:N
2
ratio, 1:3:1 [Dong et al., 2002].
2.3.2 Effect of support
The activity order of various materials used as supports in hydrocarbon steam reforming
catalysts is summarised in Table 2.3. As may be seen, most support comparison studies
employed active components obtained from Group VIII metals.
Studies involving cobalt supported catalysts have been mainly centred around the steam
reforming of ethanol. According to Haga et al. [1997], Al
2
O
3
-supported cobalt catalyst
achieved the highest conversion, as well as excellent product selectivity. This was
achieved largely due to the ability of alumina in restraining methanation and ethanol
22
Chapter 2: Literature Review
decomposition reactions. Significant methanation, however, occurred concurrently
during ethanol steam reforming over Co/ZrO
2
, Co/MgO and Co/SiO
2
catalysts, whilst
methane and carbon deposits were formed via decomposition of ethanol on Co/C. At
lower steam feed (S:C < 2), Batista et al. [2004] suggested that Co catalyst supported on
SiO
2
may offer better activity than Al
2
O
3
. This was attributed to loss of cobalt active
sites to cobalt aluminate phase in the Co/Al
2
O
3
catalysts. However, the activities of the
two catalysts became comparable at higher cobalt loading.
From Table 2.3, in nickel-based catalysts the use of Al
2
O
3
as a support appeared in
general to give higher activity during methane steam reforming over other unmodified,
single-component supports. For lower temperature range (< 873 K), Matsumura et al.
[2004] and Kusakabe et al. [2004] however reported that Ni/ZrO
2
showed better activity
and stability than both Ni/Al
2
O
3
and Ni/SiO
2
. This was ascribed to the ability of ZrO
2
to
accumulate water which assists in the formation of the hydroxyl groups. Matsumura et
al. [2004] also revealed the tendency of steam in oxidising the metallic nickel particles
on SiO
2
surface to form inactive nickel oxides. A similar phenomenon may occur with
Al
2
O
3
support although with a much lesser extent.
In an effort to improve the performance of ZrO
2
, several researchers have added cerium
oxide (CeO
2
) to this material. The accompanying enhancements were credited to the
role of CeO
2
in stabilising metal dispersion and inducing water decomposition to
generate active oxygen species effective for removing carbonaceous material [Kusakabe
et al., 2004]. In contrast, Laosiripojana and Assabumrungrat [2005] found lower
conversion in Ni/CeZrO
2
, although accompanied with less carbon deposition, compared
23
Chapter 2: Literature Review
to Ni/Al
2
O
3
. They argued that steam reforming activity of Ni/CeZrO
2
is strongly
inhibited by high hydrogen presence which in turn promotes methanation, reverse
water-gas shift and reverse methane steam reforming reactions.
Nakagawa et al. [2003] studied the effect of support material in nickel-based catalysts
for steam reforming of CH
4
. The methane conversion in the decreasing order is: Al
2
O
3
>
oxidised diamond Y
2
O
3
> TiO
2
>> MgO > SiO
2
> La
2
O
3
. As illustrated in Fig. 2.3,
both product concentrations and CH
4
conversion were influenced by the choice of
support.
Fig. 2.3. Effect of supports on CH
4
conversion and product concentrations for the
steam reforming of CH
4
over nickel-loaded catalysts. Reaction conditions:
temperature, 873 K; reaction time, 0.5 h; CH
4
:H
2
O:Ar ratio, 1:3:5 [Nakagawa et
al., 2003].
24
Chapter 2: Literature Review
For steam reforming of ethanol, the role of support in activating steam as hydroxyl
groups has a more crucial implication. Not surprisingly, as reported by Aupretre et al.
[2002] the support activity of nickel catalysts is in the decreasing order of: CeZrO
2
>
CeO
2
> CeO
2
-Al
2
O
3
> Al
2
O
3
. Furthermore, in a recent study by Wang et al. [2004a] the
performances of newly-developed Ni/CeZSM-5 and Ni/HZSM-5 zeolite catalysts were
compared with Ni/Al
2
O
3
. The utilisation of zeolite supports were shown to bring about
improved conversion by almost ten times higher compared to the Ni/Al
2
O
3
catalyst
during steam reforming of methylcyclohexane. Additionally, both zeolite catalysts also
demonstrated greater resistance from sulphur poisoning.
Liguras et al. [2003] investigated the effect of ruthenium (Ru) metal dispersed with
equal loading on different supports, i.e. Al
2
O
3
, MgO and TiO
2
. The corresponding
activity for ethanol steam reforming is in the decreasing order: Al
2
O
3
> MgO > TiO
2
.
The high activity of Al
2
O
3
was assigned to the ability of Ru particles to disperse on this
type of support. The measured dispersion of Ru on Al
2
O
3
was 21% compared to 7 and
2% for TiO
2
and MgO respectively. The Ru/Al
2
O
3
catalyst was also reported to yield
higher selectivity toward reforming products. The poorly-dispersed, MgO-supported
catalyst however resulted in a moderate and stable activity probably due to its basic
nature which prevented production of ethylene and hence, reduced coke formation.
As for their Ni-based catalysts, Aupretre et al. [2002] observed a similar trend on the
support activity order of Rh-based catalysts in ethanol steam reforming. These authors
further showed that the ceria-containing supports enhanced water gas-shift reaction
affecting the CO/CO
2
product composition. The advantage of CeO
2
compared to Al
2
O
3
25
Chapter 2: Literature Review
26
and ZnO as a support material for noble Rh, Pd and Pt metals was also substantiated by
the results of Engler et al. [1991] and Ranganathan et al. [2005] for steam reforming of
propane and methanol respectively. Moreover, Grenoble [1978b] revealed that Rh
catalysts dispersed on SiO
2
and C supports were less active during toluene steam
reforming compared to its Al
2
O
3
counterpart. This author suggested that SiO
2
yields
only tightly-bound OH groups at its surface, while carbon support failed to assist the
steam reforming reaction being unable to activate water.
eview
27
Activity order Metal loading (%) Hydrocarbon S:C Temperature (K) Sources
1 Co
Al
2
O
3
> ZrO
2
> MgO > SiO
2
> C
SiO
2
> Al
2
O
3
Al
2
O
3
SiO
2
7.4
8
18
Ethanol
Ethanol
Ethanol
2.12
1.5
1.5
673
673
673
Haga et al. [1997]
Batista et al. [2004]
Batista et al. [2004]
2 Ni
ZrO
2
> Al
2
O
3
> SiO
2
CeZrO
2
> ZrO
2
> Al
2
O
3
CeZrO
2
> Al
2
O
3
> ZrO
2
Al
2
O
3
> CeZrO
2
> CeO
2
Al
2
O
3
> oxidised diamond Y
2
O
3
> TiO
2
>> MgO
> SiO
2
> L
2
O
3
CeZrO
2
> CeO
2
> CeO
2
-Al
2
O
3
> Al
2
O
3
CeZSM-5 > HZSM-5 >> Al
2
O
3
20
10
10
5
5
9.7
5
Methane
Methane
Methane
Methane
Methane
Ethanol
Methylcyclohexane
2
2
2
1
3
1.5
1
773
773-873
973-1073
1113
873
973
853
Matsumura et al. [2004]
Kusakabe et al. [2004]
Kusakabe et al. [2004]
Laosiripojana et al. [2005]
Nakagawa et al. [2003]
Aupretre et al. [2002]
Wang et al. [2004a]
3 Ru
Al
2
O
3
> MgO > TiO
2
5 Ethanol 1.5 873-1123 Liguras et al. [2003]
4 Rh
CeZrO
2
> CeO
2
-Al
2
O
3
> CeO
2
> Al
2
O
3
CeO
2
> Al
2
O
3
Al
2
O
3
> SiO
2
> C
1
0.45
1-4
Ethanol
Propane
Toluene
1.5
5.08
0.46
973
673
713
Aupretre et al. [2002]
Engler et al. [1991]
Grenoble [1978b]
5 Pd
CeO
2
> ZnO
CeO
2
> Al
2
O
3
2-10
0.48
Methanol
Propane
1
5.08
503
673
Ranganathan et al. [2005]
Engler et al. [1991]
6 Pt
CeO
2
> Al
2
O
3
0.88 Propane 5.08 673 Engler et al. [1991]
Table 2.3. Effect of supports on steam reforming activity.
Chapter 2: Literature R
Chapter 2: Literature Review
2.3.3 Bimetallic catalysts
The concept of bimetallic systems to reforming catalysis was introduced by Sinfelt
[1973, 1983, 1985, 1987, 1997 & 2002]. Hydrocarbon reforming reactions are
traditionally carried out over transition single-metal catalysts. For instance, a
monometallic platinum reforming catalyst was employed by the Universal Oil Products
Company in 1949 [Haensel, 1955]. At the nanoscale level, in a monometallic system the
atoms assemble groups called clusters (also known as crystallites) which are dispersed
throughout the support. As an illustration, a monometallic cluster consisting of platinum
atoms is presented in Fig. 2.4.
Fig. 2.4. Arrangement of platinum atoms in a cluster of about 1 nm in size [Sinfelt,
2002].
As single-metal steam reforming catalysts often suffer from severe deactivation due to
carbon deposition, it has been suggested that the dilution of the active metal surface by
another metal may produce a bimetallic system with improved resistance to coke-
induced deactivation than its monometallic counterparts. The corresponding atomic
arrangement of a typical bimetallic cluster is presented in Fig. 2.5. The number of metal
28
Chapter 2: Literature Review
atoms present in a cluster and dispersion properties of the bimetallic catalysts are
comparable to the monometallic catalysts [Sinfelt, 2002].
Fig. 2.5. Arrangement of atoms in a bimetallic cluster about 1 nm in size. The two
kinds of atoms are represented by the black and white spheres [Sinfelt, 2002].
Various explanations of the role of the second metal in bimetallic catalysts are
summarised below [Macleod et al., 1998; Bartholomew et al., 1980]:
1. Modification of electronic properties of active metal particles. It was posited
that the altered electronic attributes result in weaker bond strength between
active species and carbon atom during adsorption of hydrocarbon [Burch and
Garla, 1981; Betizeau, 1976].
2. Dilution of active metal surface into smaller entities. As coke formation is
favoured by large clusters, when inactive second components are added,
active metal surface shrinks into smaller ensembles minimising potential sites
for coke formation [Coq and Figueras, 1984; Biloen et al., 1977].
29
Chapter 2: Literature Review
3. Providing sites for hydrogenation of coke deposits. It is known that some
metals exhibits a superior tendency for carbon gasification [Parera and
Beltramini, 1988; Carter et al., 1982; Bartholomew et al., 1980].
4. Decreasing the solubility of carbon within the metal lattice. Formation of
deactivation-causing filamentous carbon involves carbon diffusion through
the metal lattice, followed by growth of filament on the rear side of the
lattice. Consequently, the introduction of a second metal possessing lower
solubility than nickel (e.g. noble metals) slows down the rate of carbon
deposition [Bartholomew, 1980].
In one of the early studies involving bimetallic systems, rhenium (Re) [Klulsdahl,
1969], iridium (Ir) [Sinfelt, 1979; Garten and Sinfelt, 1980], tin (Sn) [Muller et al.,
1975; Volter et al., 1981], germanium (Ge) [Bouwman and Biloen, 1977; McCallister
and ONeal, 1971] and gold (Au) [Biloen et al., 1977] have been added to Pt/Al
2
O
3
catalysts. All these bimetallic catalysts were reported to have enhanced catalytic
properties, in terms of activity and stability, compared to the monometallic Pt catalyst.
Macleod et al. [1998] studied the effects of combining Sn, Ge, Re and Ir to Pt/Al
2
O
3
with an equal composition for each catalyst. These bimetallic systems were
subsequently tested for steam reforming of n-octane, where the corresponding time-
averaged deactivation rate increased in the sequence: Pt-Ir < Pt-Re < Pt-Ge < Pt-Sn <
Pt. It was, however, pointed out that deactivation rate is not a simple function of carbon
content, where the order of coke accumulation follows: Pt-Ir < Pt-Re < Pt-Sn < Pt < Pt-
Ge. Using the same bimetallic catalysts, Beltramini and Trimm [1987] showed that the
deactivation rate during n-heptane reforming increased in the order: Pt-Ir < Pt-Ge < Pt-
30
Chapter 2: Literature Review
Sn < Pt-Re < Pt. Interestingly, these authors also suggested that deactivation rate on
bimetallic catalysts was not merely correlated to the amount of coke, as coke content
increased in the order: Pt-Sn < Pt-Re < Pt-Ge < Pt-Ir < Pt. They concluded that not only
coke content but also the control of carbon deposition on reforming active sites would
be essential for minimising the deactivation effect of coking on bimetallic catalysts.
More recent studies on bimetallic reforming catalysts involve the combination of
industrially-popular nickel catalyst with a second metal. In fact, it has been reported in
literature that the addition of various metals, such as Mg [Lisboa et al., 2005], Ca
[Lisboa et al., 2005; Zhang and Verykios, 1994; Goula et al., 1996], Mo [Kepinski et al.
2000; Borowiecki et al., 2002; Wang et al, 2004a & 2004b], Co [Bartholomew et al.,
1980; Choudhary et al., 1997, 1998a & 1998b; Opoku-Gyamfi et al., 1998 & 1999a;
Wang et al., 2004a; Gonzlez et al., 2005; Takanabe et al., 2005], Cu [Choi et al., 1998],
Zn [Choi et al., 1998], Ru [Crisafulli et al., 1999], Pd [Yanhui and Diyong, 2001;
Crisafulli et al., 1999; Zhang et al., 2003], Re [Wang et al., 2004a & 2004b], Pt [Opoku-
Gyamfi et al., 1998 & 1999b; Avci et al., 2004] and Bi [Trimm, 1997], to Ni-based
catalysts increased coking resistance. More recently, several investigators have also
demonstrated substantial enhancement in catalytic life of reforming catalysts promoted
by rare-earth oxides, namely CeO
2
, PrO
2
, Sm
2
O
3
and La
2
O
3
[Ma et al., 1999; Liu et al.,
2002a; Xiancai et al., 2005; Zhuang et al., 2005; Trimm et al., 2004]. These oxides were
shown to enhance water adsorption which in turn increased tendency for steam
reforming and coke removal processes.
31
Chapter 2: Literature Review
From a commercial standpoint, cobalt in particular may be more promising than other
metals. Prior studies by Opoku-Gyamfi et al. [1998 & 1999a] in our laboratory have
shown that an alumina-supported Co-Ni catalyst exhibited synergistic effects during
methane steam reforming. The catalyst also has superior carbon deposition resistance
compared to monometallic Ni/Al
2
O
3
. Additionally, Wang et al. [2004a] reported that
cobalt-doped nickel catalyst improved activity for steam reforming of gasoline.
Choudhary and co-workers have also employed Co-Ni system in methane partial
oxidation and CO
2
reforming; and observed a substantial reduction in carbon deposition
[Choudhary et al., 1997, 1998a & 1998b]. In agreement, the finding of Gonzlez et al.
[2005] showed that high and stable activity was achieved during reforming of CH
4
with
CO
2
over a Co-Ni catalyst. Moreover, Bartholomew et al. [1980] revealed that nickel
promoted with cobalt was substantially more resilient to carbon-induced deactivation
during methanation compared to nickel. Further, Takanabe et al. [2005] demonstrated
that while Co-Ni catalyst was highly active and stable, appropriate loadings of cobalt
and nickel compounds are needed to achieve the optimum catalytic performance.
2.3.4 Deactivation of steam reforming catalysts
The operation of an industrial heterogenous catalytic reactor is susceptible to activity
loss with time-on-stream due to catalyst deactivation occurring in various modes. The
time scale of deactivation varies distinctly depending on the nature of the process as
shown in Fig. 2.6.
32
Chapter 2: Literature Review
Fig. 2.6. Time scale of deactivation of various catalytic processes [Moulijn, 2001].
Consequently, catalysis deactivation often results in poorer product yield, additional
energy consumption and production time loss associated with catalyst replacement,
regeneration and plant shutdown costing industries billions of dollars annually
[Bartholomew, 2001]. While deactivation is inevitable in the operation of catalytic
reactors; it is, however, possible to slow down the process and to minimise its effects,
for instance, the formulation of deactivation-resistant catalysts, the optimal operation of
catalytic reactors suppressing deactivation or the implementation of specific reactivating
procedures [Forzatti and Lietti, 1999]. In this case, a good understanding on physical
and chemical aspects related to catalyst deactivation must first be acquired.
Accordingly, this section briefly discusses the mechanisms of the three principal classes
of catalyst deactivation, namely poisoning, sintering (aging) and fouling (coking).
33
Chapter 2: Literature Review
2.3.4.1 Poisoning
Poisoning arises via strong chemisorption of impurities (usually contained in the feed)
on the active sites available for desired reaction. A list of typical poisons found in
common industrial processes is presented in Table 2.4.
During steam reforming, poisoning is generally encountered by adsorption of H
2
S at the
nickel catalyst surface. In general, this may be described as,
2 2
H S Ni S H Ni + = + (2.23)
Table 2.4. Examples of poisons of industrial catalysts [Forzatti and Lietti, 1999].
Process Catalyst Poison
Ammonia synthesis Fe CO, CO
2
, H
2
O, C
2
H
2
, S,
Bi, Se, Te, P
Steam reforming Ni/Al
2
O
3
H
2
S, As, HCl
Methanol synthesis, low-
temperature CO shift
Cu H
2
S, AsH
3
, PH
3
, HCl
Catalytic cracking SiO
2
-Al
2
O
3
, zeolites Organic bases, NH
3
, Na,
heavy metals
CO hydrogenation Ni, Co, Fe H
2
S, CO, S, As, HCl
Oxidation V
2
O
5
As
Automotive catalytic
converters (oxidation of CO
and HC, NO reduction)
Pt, Pd Pb, P, Zn
Methanol oxidation to
formaldehyde
Ag Fe, Ni, carbonyls
Ethylene to ethylene oxide Ag C
2
H
2
Many Transition metal oxides Pb, Hg, As, Zn
34
Chapter 2: Literature Review
The adsorption of poison is a rapid process occurring predominantly in the support
pores nearby the catalyst exterior. A number of techniques to avoid catalyst poisoning
were proposed by researchers and some of these have been adopted by the industry.
Feed treatment in order to minimise the poison content below the acceptable level is the
most obvious method. Another effective way is to add a substance which selectively
adsorbs the poison, for instance Mo and B adsorb sulphur thereby increasing resistance
from sulphur poisoning [Bartholomew, 2001]. Altering reaction conditions (e.g.
increasing temperature) may have an effect on the adsorption capacity of poison at the
catalyst surface by weakening bond strength between poison molecules and active sites
[Forzatti and Lietti, 1999].
Furthermore, the sulphur tolerance properties of Co, Ru, Fe and Ni during methanation
are compared in Fig. 2.7. The nickel catalyst suffered smaller activity loss under sulphur
poisoning than the Co, Ru and Fe catalysts. All these metals were, however, observed to
exhibit very weak sulphur resistance, where activity dropped by approximately 3-4
orders of magnitude at H
2
S concentration as low as < 100 ppb.
35
Chapter 2: Literature Review
Fig. 2.7. Effect of H
2
S poisoning on the methanation activity of various metals
[Forzatti and Lietti, 1999].
2.3.4.2 Sintering
The term sintering, or aging, in catalysis refers to the loss of active surface area due to
structural modification. Moulijn et al. [2001] pointed out that sintering occurs in all
stages in the life cycle of a catalyst from preparation (local heating during calcination),
reduction, reaction (hot spots, maldistribution) to regeneration (coke burn-off). In
supported metal catalysts, lower surface area could be due to growth of metallic
crystallites (as implicated by decreasing surface-to-volume ratios). In principle, it takes
place via 2 main processes, namely atomic migration and crystallite migration, which
are illustrated in Fig. 2.8. In atomic migration, active atoms were detached from the
36
Chapter 2: Literature Review
crystallites, where they migrated along the support surface and subsequently captured
by larger crystallites. In crystallite migration, a crystallite entity migrated across the
support surface followed by collision and coalescence [Bartholomew, 2001].
Fig. 2.8. Two conceptual models for crystalline growth due to sintering by: (a)
atomic migration; and (b) crystalline migration [Bartholomew, 2001].
The most common cause of sintering is prolonged exposure to high-temperature
environment. The effect of temperature on the extent of sintering in Ni catalysts as a
function of time is presented in Fig. 2.9. Higher temperature usage caused greater loss
of normalised surface area (defined as the ratio of nickel surface area at instantaneous
time to its initial surface area). The surface areas decreased rapidly in the first 15 h,
followed by steady and marginal deactivation rates up to 50 h.
37
Chapter 2: Literature Review
Fig. 2.9. Normalised nickel surface area (based on H
2
adsorption) versus time data
during sintering of Ni/SiO
2
in hydrogen atmosphere [Bartholomew, 2001].
Similarly, porous catalyst supports also suffer consequences of thermally-induced
sintering. In this case, the flow of solid support material may cause narrowing of pores
contained and ultimately closure of these pores. Consequently, access to the catalyst
interior was blocked and surface area was lost. Fig. 2.10 presents a diagram showing
pore closure due to sintering.
Fig. 2.10. Illustration of pore closure at support due to sintering [Adapted from
Fogler, 1999].
38
Chapter 2: Literature Review
Sintering can be effectively delayed by carefully choosing temperature ranges for the
various processes in the catalyst life cycle. As a guideline, temperature below 0.3-0.5
times the melting point of metal and support has been recommended [Moulijn et al.,
2001]. The 0.3 times melting point marks the point, referred as Huttig temperature, at
which atoms at defects gain mobility. Further, the 0.5 times melting point indicates the
point, known as Tamman temperature, at which atoms from the bulk become mobile.
The Huttig, Tamman and melting temperatures of some common compounds
encountered in heterogenous catalysis are given in Table 2.5.
The presence of strong metal-support interactions has been known to increase catalyst
resistance towards sintering. Additionally, sintering can be minimised by introducing
promoters, such as BaO, CeO
2
, La
2
O
3
, SiO
2
and ZrO
2
, in order to increase support
stability by preventing surface restructuring of alumina-supported metal catalysts
[Burtin et al., 1987; Kato et al., 1987; Machida et al., 1987; Beguin et al., 1991].
39
Chapter 2: Literature Review
Table 2.5. Hutting, Tamman and melting temperatures (K) of common compounds
in heterogenous catalysis [Moulijn et al., 2001].
Compound T
Huttig
T
Tamman
T
melting
Pt 608 1014 2028
PtO 247 412 823
Pd 548 914 1828
PdO 307 512 1023
Rh 677 1129 2258
Ru 817 1362 2723
Fe 542 904 1808
Co 526 877 1753
Ni 518 863 1725
NiO 669 1114 2228
Ag 370 617 1233
Cu 407 678 1356
CuO 480 800 1599
Mo 865 1442 2883
Zn 208 347 693
ZnO 675 1124 2248
Al
2
O
3
695 1159 2318
SiO
2
596 993 1986
2.3.4.3 Fouling
Fouling occurs when active surface area is covered by a deposit of species originated
from the fluid phase. Examples of substances that may cause deposition include
combustion ashes, wear residues from process equipment, asphaltenes during
processing of heavy petroleum fractions and carbonaceous residues from
dehydrogenation in hydrocarbon processing [Moulijn et al., 2001]. Carbon deposition is
particularly regarded as the key challenge in steam reforming operation, where coke
40
Chapter 2: Literature Review
deposits act as barriers denying access of reactant hydrocarbon and steam [Rodriguez et
al., 1997]. At reactor level, interparticle coke growth leads to excessive pressure drop
across the reactor and ultimately cessation of fluid flow requiring shut-down. The
occurrence of coking over the surface of supported metal catalyst is illustrated in Fig.
2.11.
Fig. 2.11. Illustration of carbon deposition at support [Adapted from Fogler, 1999].
The effect of coke on reforming activity does not depend solely on its quantity, but also
the location of deposition. As illustrated in Fig. 2.12, coke deposited on pore mouths
causes more detrimental effect by forming pore blockage compared to a situation in
which the same quantity of coke is distributed on the support inner wall [Richardson,
1972; Ozawa and Bishoff, 1968].
In an attempt to minimise carbon deposition, catalysts with strong anti-coking attributes
have been developed. As previously discussed, the addition of a dopant (notably a rare-
earth oxide or a noble metal) into nickel reforming catalysts may in fact reduce coke
formation. Nevertheless, catalyst mortality is inevitable and hence, it is important to
carry out reactor operation with a view to reduce the effect of coking. Consequently, the
reaction conditions (e.g. temperature, pressure and steam-to-carbon feed ratio) should
41
Chapter 2: Literature Review
be selected in a way which permits the least coke accumulation (by means of
minimisation of carbon formation coupled with maximisation of carbon removal).
Fig. 2.12. Fouling, crystallite encapsulation and pore plugging of a supported metal
catalyst due to carbon deposition [Bartholomew, 2001].
2.4 Carbon deposition
Coke deposition during steam reforming occurs via hydrocarbon dehydrogenation
and/or CO disproportionation [Bartholomew, 2001]. Carbon lay-down on the catalyst
surface leads to loss of adsorption sites for both hydrocarbon and steam [Fortazzi and
Lietti, 1999; Rodriguez et al., 1997]. Significantly, overall reactor operation - which
admits coke minimisation and/or catalyst regeneration - has been a strategic objective.
Therefore, the mechanism by which carbon forms, as well as basic structure and
reactivity of coke residues, would need to be explored as discussed further in this
section.
42
Chapter 2: Literature Review
2.4.1 Mechanisms of carbon lay-down
As mentioned previously, carbon-containing fragments may be obtained from carbon
monoxide (via Boudouard) and/or hydrocarbons (via decomposition). These 2 coke-
forming pathways are summarised in Figs. 2.13 and 2.14 respectively.
As presented in Fig. 2.13, the dissociation of CO produces elemental carbon, C
o
, and
atomic oxygen, O. The C
o
species can polymerise to result in amorphous and graphitic
carbons, C
|
and C
c
, respectively, form a metal carbide, C

, or dissolve into the bulk of


the metal to yield whisker-like or vermicular carbon, C
v
[Bell, 1987]. The graphitic
form, C
c
, is highly stable transformed from C
|
at high temperatures (> 773 K) over a
period of time.
Fig. 2.13. Formation, transformation and gasification of carbon on nickel (a, g, s
refer to adsorbed, gaseous and solid states, respectively) [Bartholomew, 2001].
From Fig. 2.14, the dissociation of hydrocarbon is more complicated in which it also
produces various partially hydrogenated species, i.e. CH
x
and C
n
H
z
. These radicals may
condense to form a high molecular weight coke.
43
Chapter 2: Literature Review
Fig. 2.14. Formation and transformation of coke on metal surfaces (a, g, s refer to
adsorbed, gaseous and solid states, respectively) [Bartholomew, 2001].
An illustration of the growth of carbon filaments on a supported nickel catalyst is
presented in Fig. 2.15. The process is initiated by the adsorption of hydrocarbon
molecules on an active particle surface releasing gaseous products and atomic carbons.
Although surface carbon species can be gasified without difficulty, carbon eventually
diffuses - in a form of a metal-carbide intermediate - through the nickel crystallite to the
rear side [Renshaw et al., 1971]. It is believed that the diffusion of carbon through the
nickel particle is the rate-determining step of the carbon filament formation [Aiello et
al., 2000]. The driving force of the diffusion process may either be concentration
gradient [Rostrup-Nielsen and Trimm, 1977], temperature gradient [Baker and Harris,
1978] or a combination of these 2 factors [Bartholomew, 1982]. As carbon reaches the
support, whisker nucleation forms where filamentous carbon would start to grow as
more carbons diffuse. The filament elongation causes detachment of the nickel particle,
which sits on the head of the carbon whiskers rooted in the support. The catalyst
would remain active during fibre elongation until at a stage when sufficient pyrolytic
44
Chapter 2: Literature Review
coke encapsulates the metal particle thereby preventing further adsorption [Bell, 1987].
However, from an industrial perspective growth of carbon whiskers poses various
operational problems associated with reactor blockage and increased pressure drop.
Moreover, the formation of filamentous carbon on a supported metal catalyst at a given
reaction condition is dictated by metal type, crystallite size and support characteristic.
Fig. 2.15. Carbon formation on supported metal catalyst [Moulijn et al., 2001].
Bartholomew [1982] presented an equilibrium C-H-O diagram showing carbon
isotherms at 723 K and 1.4 atm for graphite and amorphous carbons (cf. Fig. 2.16). As
shown in Fig. 2.16, Points A and B correspond to H
2
O/CH
4
compositions of 0.1 and 2.0
respectively. The location of Point A indicates that carbon would be expected to form
under the equilibrium condition. In contrast, at higher S:C ratio Point B lies below the
boundary lines suggesting no tendency for carbon deposition.
45
Chapter 2: Literature Review
Fig. 2.16. Equilibrium carbon deposition boundaries for amorphous and graphitic
carbons at 723 K and 1.4 atm [Bartholomew, 1982].
2.4.2 Properties and reactivity
Due to its complex nature, coke characterisation has presented one of the most
significant challenges in catalyst deactivation research [Sahoo et al., 2004]. Various
techniques have been utilised to investigate the physicochemical attributes and structure
of carbonaceous residues resulting from hydrocarbon dehydrogenation. These involve a
range of spectroscopic
13
C-NMR [Diez et al., 1990; Sahoo et al., 2004; Martn et al.,
2005], AES [Niemantsverdriet and van Langeveld, 1986], laser Raman [Espinat et al.,
1985; Shamsi et al., 2005], SANS [Acharya et al., 1990], XPS [Ordez et al., 2001;
Shamsi et al., 2005]; microscopic TEM [Tracz et al., 1990], HRSTEM [Gallezot et al.,
1989], CAEM [Chang et al., 1990], AFM [Martn et al., 2005]; chemical fractionation
by solvent extraction [Van Doorn and Moulijn, 1990; Furimsky and Massoth, 1999] and
temperature-programmed [Querini and Fung, 1997; Marafi and Stanislaus, 1997;
Ordez et al., 2001; Liu et al., 2001; Liu et al., 2002b; Sahoo et al., 2004; Martn et al.,
46
Chapter 2: Literature Review
2005; Shamsi et al., 2005] methods. To a great extent, these studies indicate that
quantity, location and characteristics of coke deposits were influenced by the type of
hydrocarbon feed used, nature of the catalysts and choice of operating variables.
Nevertheless, Hardiman et al. [2006] emphasised that better integration of information
derived from various finger-printing techniques would ultimately provide deeper
insights into the mechanism of carbon lay-down and rejuvenation strategy.
Chang et al. [1990] investigated carbon deposition over alumina-supported platinum
reforming catalysts. Using a combination of electron microscopy and thermogravimetric
methods, these investigators suggested that coke deposition occurred predominantly at
specific sites on the support, while carbon lay-down on the platinum particles was not
apparent. It was also noted that filamentous and graphitic carbonaceous species were
absent under the typical naphtha reforming temperature around 773 K. The fact that
metal particles were carbon-free was attributed to the promotional effect of the metallic
component on coke burn-off process. This is consistent with the observation of Shamsi
et al. [2005] where coke deposit on the metal sites is characterised by a lower
temperature peak (during TPO) than the carbon residue located on the support. On a
Pt/alumina catalyst, XPS examination further revealed that graphitic carbon was present
since high reforming temperature (1073 K) was implemented.
Querini and Fung [1994 & 1997] pointed out that additional information on coke
morphology may be obtained from thermogravimetric runs by substituting TPO data
into a kinetic model. They demonstrated that the reaction of coke (possessing a tri-
dimensional structure as revealed from TEM examination) on the alumina support of a
47
Chapter 2: Literature Review
Pt-containing catalyst with oxygen species experienced variable controlling step with
reaction order increasing from nearly 0 to close to 1. In fact, this was a reflection of the
decreasing size of the carbonaceous particles during the burn-off process implicating a
relationship between gasification rate and coke surface area. In a further enquiry,
temperature-programmed experiments utilising inert helium gas (presenting an
environment free of external oxygen) was carried out, where CO and CO
2
species were
interestingly detected in the product gas stream. Based on comparison with TPR and
TPO data, it was revealed that OH groups on the alumina surface played a dominant
role by being the reactive species during reaction with the carbon residues.
On the basis of TPCO
2
, H
2
-TPR and TPO experiments, Liu et al. [2001] also reported 2
distinct types of carbonaceous species on a Mo/HZSM-5 catalyst used for methane
dehydro-aromatisation. The coke residue on the Mo species also showed a lower peak
temperature than the carbon deposited on the support. In addition, kinetic analysis of
carbon burn-off indicated that apparent activation energy was higher during removal of
coke from the support compared to that located on the Mo particles [Liu et al., 2002].
Furthermore, Lobo and Trimm [1973] observed that the C:H ratios of coke residues
from dehydrogenation of several light hydrocarbons in nickel foils at 673-873 K were in
the range between 4.5-7.0 revealing aromatic nature. The reaction pathway towards the
formation of aromatic coke on oxide and sulfide catalysts involves polymerisaton of
olefins, followed by cyclisation and subsequent formation of polynuclear aromatics
which ultimately condense as coke, as detailed by Farrauto and Bartholomew [1997].
48
Chapter 2: Literature Review
In particular, the utilisation of thermal programmed techniques is useful for examining
coke reactivity. Carbon may react with various compounds and produce gaseous
products. These reactions are summarised by [Figueiredo, 1986];
2 2
CO O C + = -394.5 kJ mol
0
800
H A
-1
(2.24)
2 2
H CO O H C + + = 135.6 kJ mol
0
800
H A
-1
(2.25)
2CO CO C
2
+ = 172.5 kJ mol
0
800
H A
-1
(2.26)
4 2
CH 2H C + = -87.3 kJ mol
0
800
H A
-1
(2.27)
Walker et al. [1959] compared carbon gasification rate of various reactant gases in the
absence of catalyst at 0.1 bar and 1073 K, where they found the rate decreased in the
order of: O
2
(10
5
) > H
2
O (3) > CO
2
(1) > H
2
(310
-3
). Figueiredo [1986] displays
different weight behaviours during carbon removal experiment (cf. Fig. 2.17). Catalyst
A demonstrated a fast gasification with maximum removal rate at 800 K. Catalyst B
revealed two-stage coke burn-off processes at 500 and 1100 K respectively. Catalyst C
exhibited highly unreactive carbon deposits, which required high-temperature removal
at > 1200 K. Catalyst D contained somewhat stable coke residues, but showing very
rapid gasification starting at about 900 K.
The weight loss profile of Catalyst A was obtained from gasification of coke deposits
on an alumina-supported nickel catalyst. The mild weight loss shown at < 700K was
attributed to burning of surface carbon [De Deken et al., 1981]. On the other hand, less-
reactive carbon residues (which were diffused and dissolved in nickel) are removed at
higher temperatures (> 900K). Prior to burn-off, the latter type of coke diffused slowly
49
Chapter 2: Literature Review
to the surface and reacted with the gaseous reactant at the surface. Therefore, carbon
gasification reaction in principal is the reverse of the deposition process. Further,
surface reaction between carbon with oxygen, steam or carbon dioxide is generally fast,
where the coke removal rate is controlled by carbon back-diffusion. In contrast, reaction
rate with hydrogen is much slower and so, surface reaction is the limiting-rate step
during the hydrogenation of coke deposits.
Fig. 2.17. Temperature-programmed gasification of different types of coked
catalysts. Operating conditions: heating rate, 5 K min
-1
; CO
2
flow rate, 0.01 mol
min
-1
[Figueiredo, 1986].
Apart from signifying reactivities, peak temperatures during carbon gasification in
temperature-programmed reduction (TPR) runs may also be used as a finger-printing
technique to identify the different types and structures of carbons present, as
summarised in Table 2.6.
50
Chapter 2: Literature Review
Table 2.6. Forms and reactivities of carbon species formed by decomposition of
CO on nickel [Bartholomew, 2001].
Structural type Designation
Temperature of
formation (K)
Peak temperature
during TPR
Adsorbed, atomic (surface
carbide)
C
o
473-673 473
Polymeric, amorphous films or
filaments
C
|
523-773 673
Vermicular filaments, fibers,
and/or whiskers
C
v
573-1273 673-873
Nickel carbide (bulk) C

423-523 548
Graphitic (crystalline) platelets
or films
C
c
773-823 823-1123
2.5 Summary
Steam reforming plays a critical role in providing carbon monoxide and hydrogen
feedstocks to various value-added petrochemical processes. As other hydrocarbon-
related reactions, it suffers from activity loss with time-on-stream attributed to catalyst
deactivation occurring in various means, such as poisoning, sintering and fouling
(especially carbonaceous deposits). Poisoning may usually be overcome by proper feed
treatment and/or addition of poison-adsorbing substances. Sintering is usually
associated with high-temperature exposure, thus it is effectively avoided by maintaining
system temperature below 0.3-0.5 times the melting point of metal and support
51
Chapter 2: Literature Review
components. Nevertheless, deactivation arising from fouling by carbon deposits
probably causes the greatest complication in the operation of a steam reformer using
conventional nickel-based catalysts (favoured due to its low cost and high activity). The
direct consequence is blockage of active sites, while excessive carbon growth would
also result in pressure drop across the reactor and flow blockage. As dictated by
thermodynamics, the entire operational range of a typical steam reforming process
favours carbon formation via CO disproportionation (Boudouard) and/or hydrocarbon
decomposition. Initially, these side reactions produce elemental carbons which
subsequently diffuse through the metallic particle and form nucleation sites on the
support as grounds for filamentous carbon growth. Throughout the whole process,
carbon undergoes transformation involving alteration in structure and morphology,
thereby affecting its chemical properties. The reactivity of coke residues may be
effectively assessed by thermal programmed analysis using suitable active gases.
Moreover, in many cases the current industrial practice to charge excess steam to the
reactor has been proven to be an expensive carbon-suppression approach due to the high
energy cost associated with steam generation. Accordingly, the main key in improving
the sustainability and affordability of steam reforming depends largely on carbon
control. Consequently, modified catalyst formulation was sought to reduce coking at
catalyst surface. The addition of cobalt to a conventional Ni-based catalyst (producing a
bimetallic Co-Ni system) has been shown to improve the activity and stability during
reforming runs. Furthermore, it was also reported that some supports may actually
contribute significantly in reducing carbon deposition by activating steam species.
Additionally, reaction variables can also be manipulated to create a reforming
environment which suppresses coke formation.
52
Chapter 2: Literature Review
References
Acharya, D.R., Allen, A.J. and Hughes, R. (1990). A small-angle neutron scattering
investigation of coke deposits on catalysts. Ind. Eng. Chem. Res., 29, 1119-1125.
Aiello, R., Fiscus, J.E., zur Loye, H.-C. and Amiridis, M.D. (2000). Hydrogen
production via the direct cracking of methane over Ni/SiO
2
: catalyst deactivation
and regeneration. Appl. Catal. A: Gen., 192, 227-234.
Akers, W.W. and Camp, D.P. (1955). Kinetics of the methane-steam reaction. AIChE J.,
1, 471-475.
Al-Ubaid, A.S. (1986). Steam reforming of hydrocarbons catalyzed over nickel
supported catalysts. Arabian J. Sci. Eng., 12, 189-198.
Amenomiya, Y. and Pleizer, G. (1982). Alkali-promoted alumina catalysts: II. Water-
gas shift reaction. J. Catal., 76, 345-353.
Armstrong, E. F. and Hilditch, T. P. (1920). Catalytic actions at solid surfaces. IV. The
interaction of carbon monoxide and steam as conditioned by iron oxide and by
copper. Proc. Roy. Soc., 97A, 265-273.
Aupretre, F., Descorme, C. and Duprez, D. (2002). Bio-ethanol catalytic steam
reforming over supported metal catalysts. Catal. Comm., 3, 263-267.
Avci, A.K., Trimm, D.L., Aksoylu, A.E. and Onsan, Z.I. (2004). Hydrogen production
by steam reforming of n-butane over supported Ni and Pt-Ni catalysts. Appl.
Catal. A: Gen., 258, 235-240.
Baker, R.T.K. and Harris, P.S. (1978). Formation of filamentous carbon, in: Walker,
P.L., Jr. & Thrower, A. (Eds.), Chemistry and Physics of Carbon, vol. 14.
Marcel Dekker, New York, pp. 83-165.
Bartholomew, C.H., Weatherbee, G.D. and Jarvi, G.A. (1980). Effects of carbon
deposits on the specific activity of nickel and nickel bimetallic catalysts. Chem.
Eng. Commun., 5, 125-134.
Bartholomew, C.H. (1982). Carbon deposition in steam reforming and methanation.
Catal. Rev.-Sci., Eng., 24, 67-112.
53
Chapter 2: Literature Review
Bartholomew, C.H. (2001). Mechanisms of catalyst deactivation. Appl. Catal. A: Gen.,
212, 17-60.
Batista, M.S., Santos, R.K.S., Assaf, E.M., Assaf, J.M. and Ticinelli, E.A. (2004). High
efficiency steam reforming of ethanol by cobalt-based catalysts. J. Power
Sources, 134, 27-32.
Beguin, B., Garbowski, E. and Primet, M. (1991). Stabilization of alumina toward
thermal sintering by silicon addition. J. Catal., 127, 595-604.
Bell, A.T. (1987). Characterization of carbonaceous residues on catalysts, in: Petersen,
E.E. & Bell, A.T. (Eds.), Catalyst Deactivation. Marcel Dekker, Inc., New York
and Basel, pp. 235-260.
Beltramini, J. and Trimm, D.L. (1987). Activity, selectivity and coking over mono- and
bi- metallic reforming catalysts. Appl. Catal., 32, 71-83.
Betizeau, C., Leclercq, G., Maurel, R., Bolivar, C., Charcosset, H., Frety, R. and
Tournayan, L. (1976). Platinum-rhenium-alumina catalysts: III. Catalytic
properties. J. Catal., 45, 179-188.
Bhatta, K.S.M. and Dixon, G.M. (1969). Role of urania and alumina as supports in the
steam reforming of n-butane at pressure over nickel-containing catalysts. Ind.
Eng. Prod. Res. Develop., 8, 324-331.
Biloen, P., Dautzenberg, F.M. and Sachtler, W.M.H. (1977). Catalytic dehydrogenation
of propane to propene over platinum and platinum-gold alloys. J. Catal., 50, 77-
86.
Borowiecki, T., Giecko, G. and Panczyk, M. (2002). Effects of small MoO
3
additions
on the properties of nickel catalysts for the steam reforming of hydrocarbons: II.
Ni-Mo/Al
2
O
3
catalysts in reforming, hydrogenolysis and cracking of n-butane.
Appl. Catal. A: Gen., 230, 85-97.
Bouwman, R. and Biloen, P. (1977). Valence state and interaction of platinum and
germanium on -Al
2
O
3
investigated by X-ray photoelectron spectroscopy. J.
Catal., 48, 209-216.
54
Chapter 2: Literature Review
Breen, J.P., Burch, R. and Coleman, H.M. (2002). Metal-catalysed steam reforming of
ethanol in the production of hydrogen for fuel cell applications. Appl. Catal. B:
Environ., 39, 65-74.
Burch, R. and Garla, L.C. (1981). Platinum-tin reforming catalysts: II. Activity and
selectivity in hydrocarbon reactions. J. Catal., 71, 360-372.
Burtin, P., Brunelle, J.P. Pijolat, M. and Soustelle, M. (1987). Influence of surface area
and additives on the thermal stability of transition alumina catalyst supports: I.
Kinetic data. Appl. Catal., 34, 225-238.
Byrne, P.J. and Gohr, E.J. (1932). Recent progress in hydrogenation of petroleum. Ind.
Eng. Chem., 24, 1129-1135.
Campbell, C.T. and Daube, K.A. (1987). A surface science investigation of the water-
gas shift reaction on Cu(111). J. Catal., 104, 109-119.
Carter, J.L., McVinker, G.B., Weissman, W., Kmak, M.S. and Sinfelt, J.H. (1982).
Bimetallic catalysts; application in catalytic reforming. Appl. Catal., 3, 327-346.
Chang, T.S., Rodriguez, N.M. and Baker, R.T.K. (1990). Carbon deposition on
supported platinum particles. J. Catal., 123, 486-495.
Choi, J.-S., Moon, K.-I., Kim, Y.G., Lee, J.S., Kim, C.-H. and Trimm, D.L. (1998).
Stable carbon dioxide reforming of methane over modified Ni/Al
2
O
3
catalysts.
Catal. Lett., 52, 43-47.
Choudhary, V.R., Rane, V.H. and Rajput, A.M. (1997). Beneficial effects of cobalt
addition to Ni-catalysts for oxidative conversion of methane to syngas. Appl.
Catal. A: Gen., 162, 235-238.
Choudhary, V.R., Rajput, A.M., Prabhakar, B. and Mamman, A.S. (1998a). Partial
oxidation of methane to CO and H
2
over nickel and/or cobalt containing ZrO
2
,
ThO
2
, UO
2
, TiO
2
and SiO
2
catalysts. Fuel, 77, 1803-1807.
Choudhary, V.R. and Mamman, A.S. (1998b). Simultaneous oxidative conversion and
CO
2
or steam reforming of methane to syngas over CoO-NiO-MgO catalyst. J.
Chem. Technol. Biotechnol., 73, 345-350.
55
Chapter 2: Literature Review
Coq, B. and Figueras, F. (1984). Conversion of methylcyclopentane on platinum-tin
reforming catalysts. J. Catal., 85, 197-205.
Crisafulli, C., Scire, S., Maggiore, R., Minico, S. and Galvagno, S. (1999). CO
2
reforming of methane over Ni-Ru and Ni-Pd bimetallic catalysts. Catal. Lett.,
59, 21-26.
De Deken, J., Menon, P.G., Froment, G.F. and Haemers, G. (1981). On the nature of
carbon in Ni/-Al
2
O
3
catalyst deactivated by the methane-steam reforming
reaction. J. Catal., 70, 225-229.
De Groote, A.M., Froment, G.F. and Kobylinski, T.H. (1996). Synthesis gas production
from natural gas in a fixed bed reactor with reversed flow. Can. J. Chem. Eng.,
74, 735-742.
Deluzarche, A., Hindermann, J.-P., Kiennemann, A. and Kieffer, R. (1985). Application
of chemical trapping to the determination of surface species and to the study of
their evolution under reaction conditions in heterogeneous catalysis. J. Mol.
Catal., 31, 225-250.
Diagne, C., Vos, P.J., Kiennemann, A., Perrez, M.J. and Portela, M.F. (1990). Water-
gas shift reaction over chromia-promoted magnetite: use of temperature-
programmed desorption and chemical trapping in the study of the reaction
mechanism. React. Kinet. Catal. Lett., 42, 25-31.
Diez, F., Gates, B.C., Miller, J.T., Sajkowski, D.J. and Kukes, S.G. (1990). Deactivation
of a nickel-molybdenum/-alumina catalyst: influence of coke on the
hydroprocessing activity. Ind. Eng. Chem. Res., 29, 1999-2004.
Dong, W.-S., Roh, H.-S., Jun, K.-I., Park, S.-E. and Oh, Y.-S. (2002). Methane
reforming over Ni/Ce-ZrO
2
catalyst: effect of nickel content. Appl. Catal. A:
Gen., 226, 63-72.
El Solh, T., Jarosch, K. and de Lasa, H.I. (2001). Fluidizable catalyst for methane
reforming, Appl. Catal. A: Gen., 210, 315-324.
Engler, B., Koberstein, D., Lindner, D. and Lox, E. (1991). The influence of three-way
catalyst parameters on secondary emission, in: Crucq, A. (Ed.), Catalysis and
56
Chapter 2: Literature Review
Automotive Pollution Control II. Elsevier Science Publishers B.V., Amsterdam,
pp. 641-655.
Espinat, D., Dexpert, H., Freund, E., Martino, G., Couzi, M, Lespade, R. and Cruege, F.
(1985). Characterization of the coke formed on reforming catalysts by laser
Raman spectroscopy. Appl. Catal., 16, 343-354.
Farrauto, R.J. and Bartholomew, C.H. (1997). Fundamentals of Industrial Catalytic
Processes. Chapman & Hall, Kluwer Academic Publishers, London.
Fatsikostas, A.N., Kondarides, D.I. and Verykios, X.E. (2002). Production of hydrogen
for fuel cells by reformation of biomass-derived ethanol. Catal. Today, 75, 145-
155.
Figueiredo, J.L. (1986). Gasification of carbon deposits on catalysts and metal surfaces.
Fuel, 65, 1377-1382.
Fogler, H.S. (1999). Elements of Chemical Reaction Engineering. Prentice Hall PTR,
New Jersey.
Forzatti, P. and Lietti, L. (1999). Catalyst deactivation. Catal. Today, 52, 165-181.
Furimsky, E. and Massoth, F.E. (1999). Deactivation of hydroprocessing catalysts.
Catal. Today, 52, 381-495.
Gallezot, P., Leclerq, C., Barbier, J. and Marecot, P. (1989). Location and structure of
coke deposits on alumina-supported platinum catalysts by EELS associated with
electron microscopy. J. Catal., 116, 164-170.
Garten, R.L. and Sinfelt, J.H. (1980). Structure of Pt-Ir catalysts: Mssbauer
spectroscopy studies employing
57
Fe as a probe. J. Catal., 62, 127-139.
Gonzlez, O., Lujano, J., Pietri, E. and Goldwasser, M.R. (2005). New Co-Ni catalyst
systems used for methane dry reforming based on supported catalysts over an
INT-MM1 mesoporous material and a perovskite-like oxide precursor
LaCo
0.4
Ni
0.6
O
3
. Catal. Today, 107-108, 436-443.
Gorte, R.J. and Zhao, S. (2005). Studies of the water-gas shift reaction with ceria-
supported precious metals. Catal. Today, 104, 18-24.
57
Chapter 2: Literature Review
Goula, M.A., Lemonidou, A.A. and Efstathiou, A.M. (1996). Characterization of
carbonaceous species formed during reforming of CH
4
with CO
2
over Ni/CaO
Al
2
O
3
catalysts studied by various transient techniques. J. Catal., 161, 626-640.
Grenoble, D.C. (1978a). The chemistry and catalysis of the water/toluene reaction: 1.
The specific activities and selectivities of the group VIII metals supported on
Al
2
O
3
. J. Catal., 51, 203-211.
Grenoble, D.C. (1978b). The chemistry and catalysis of the water/toluene reaction: 2.
The role of support and kinetic analysis. J. Catal., 51, 212-220.
Haensel, V. (1955). Aromatization, hydroforming and platforming, in: Brooks, B.T.,
Boord, C.E., Kurtz, S.S. & Schmerling, L. (Eds.), The Chemistry of Petroleum
Hydrocarbons, vol. II. Reinhold, New York, pp. 189219.
Haga, F., Nakajima, T., Miya, H. and Mishima, S. (1997). Catalytic properties of
supported cobalt catalysts for steam reforming of ethanol. Catal. Lett., 48, 223-
227.
Hardiman, K.M., Cooper, C.G., Adesina, A.A. and Lange, R. (2006). Post-mortem
characterization of coke-induced deactivated alumina-supported Co-Ni catalysts.
Chem. Eng. Sci., 61, 2565-2573.
Haslam, R.T. and Russell, R.P. (1930). Hydrogenation of petroleum. Ind. Eng. Chem.,
32, 1030-1037.
Hegarty, M.E.S., OConnor, A.M. and Ross, J.R.H. (1998). Syngas production from
natural gas using ZrO
2
-supported metals. Catal. Today, 42, 225-232.
Intarajang, K. and Richardson, J.T. (1999). Catalytic steam reforming of chlorocarbons:
catalyst comparison. Appl. Catal. B: Environ., 22, 27-34.
Jacobs, G., Patterson, P.M., Graham, U.M., Sparks, D.E. and Davis, B.H. (2004). Low
temperature water-gas shift: kinetic isotope effect observed for decomposition of
surface formates for Pt/ceria catalysts. Appl. Catal. A: Gen., 269, 63-73.
Kato, A., Yamashita, H., Kawagoshi, H. and Matsuda, S. (1987). Preparation of
larnthanum |-alumina with high surface area by coprecipitation. J. Am. Ceram.
Soc., 70, C157-C159.
58
Chapter 2: Literature Review
Kepinski, L., Stasinska, B. and Borowiecki, T. (2000). Carbon deposition on Ni/Al
2
O
3
catalysts doped with small amounts of molybdenum. Carbon, 38, 1845-1856.
Klulsdahl, H.E. (1968). US Patent, 3 415 737.
Kneale, B. and Ross, J.R.H. (1981). The steam reforming of ethane over nickel/alumina
catalysts. Faraday Discuss. Chem. Soc., 72, 157-171.
Kusakabe, K., Sotowa, K.-I., Eda, T. and Iwamoto, Y. (2004). Methane steam
reforming over Ce-ZrO
2
-supported noble metal catalysts at low temperature.
Fuel Process. Technol., 86, 319-326.
Laosiripojanaa, N. and Assabumrungrat, S. (2005). Methane steam reforming over
Ni/Ce-ZrO
2
catalyst: Influences of Ce-ZrO
2
support on reactivity, resistance
toward carbon formation, and intrinsic reaction kinetics. Appl. Catal. A: Gen.,
290, 200-211.
Li, Y., Fu, Q. and Flytzani-Stephanopoulos, M. (2000). Low-temperature water-gas
shift reaction over Cu- and Ni-loaded cerium oxide catalysts. Appl. Catal. B:
Environ., 27, 179-191.
Li, B., Watanabe, R., Maruyama, K., Nurunnabi, M., Kunimori, K. and Tomishige, K.
(2005). High combustion activity of methane induced by reforming gas over
Ni/Al
2
O
3
catalysts. Appl. Catal. A: Gen., 290, 36-45.
Liguras, D.K., Kondarides, D.I. and Verykios, X.E. (2003). Production of hydrogen for
fuel cells by steam reforming of ethanol over supported noble metal catalysts.
Appl. Catal. B: Environ., 43, 345-354.
Lisboa, J.S., Santos, D.C.R.M., Passos, F.B. and Noronha, F.B. (2005). Influence to the
addition of promoters to steam reforming catalysts. Catal. Today, 101, 15-21.
Liu, H., Li, T., Tian, B. and Xu, Y. (2001). Study of the carbonaceous deposits formed
on a Mo/HZSM-5 catalyst in methane dehydro-aromatization by using TG and
temperature-programmed techniques. Appl. Catal. A: Gen., 213, 103-112.
Liu, Y., Hayakawa, T., Suzuki, K., Hamakawa, S., Tsunoda, T., Ishii, T. and Kumagai,
M. (2002a). Highly active copper/ceria catalysts for steam reforming of
methanol. Appl. Catal. A: Gen., 223, 137-145.
59
Chapter 2: Literature Review
Liu, H., Su, L., Wang, H., Shen, W., Bao, X. and Xu, Y. (2002b). The chemical nature
of carbonaceous deposits and their roles in methane dehydro-aromatization on
Mo/MCM-22 catalysts. Appl. Catal. A: Gen., 236, 263-280.
Lobo, L.S. and Trimm. D.L. (1973). Carbon formation from light hydrocarbons on
nickel. J. Catal., 29, 15-19.
Ma, L., Praharso and Trimm, D.L. (1999). Rare-earth oxides promoted nickel based
catalysts for steam reforming. Mater. Sci. Forum, 315-317, 187-193.
Machida, M., Eguchi, K. and Arai, H. (1987). Effect of additives on the surface area of
oxide supports for catalytic combustion. J. Catal., 103, 385-393.
Macleod, N., Fryer, J.R., Stirling, D. and Webb, G. (1998). Deactivation of bi- and
multimetallic reforming catalysts: influence of alloy formation on catalyst
activity. Catal. Today, 46, 37-54.
Marafi, M. and Stanislaus, A. (1997). Effect of initial coking on hydrotreating catalyst
functionalities and properties. Appl. Catal. A: Gen., 159, 259-267.
Marschner, F., Renner, H.-J. and Boll, W. (2005). Gas production: 2. Steam reforming
of natural gas and other hydrocarbons, in: Pelc, H., Elvers, B. & Hawkins, S.
(Eds.), Ullmanns Encyclopedia of Industrial Chemistry, 7
th
Edn. Wiley-VCH
Verlag GmbH & Co, Germany.
Martn, N., Viniegra, M., Zarate, R., Espinosa, G. and Batina, N. (2005). Coke
characterization for an industrial Pt-Sn/-Al
2
O
3
reforming catalyst. Catal. Today,
107-108, 719-725.
Matsumura, Y. and Nakamori, T. (2004). Steam reforming of methane over nickel
catalysts at low reaction temperature. Appl. Catal. A: Gen., 258, 107-114.
McCallister, T.P. and ONeal, K.R. (1971). German Offenlegungsschrift, 2 104 429.
Millar, G.J., Rochester, C.H. and Waugh, K.C. (1993). An FTIR study of the adsorption
of methanol and methyl formate on potassium-promoted Cu/SiO
2
catalysts. J.
Catal., 142, 263-273.
Ming, Q., Healey, T., Allen, L. and Irving, P. (2002). Steam reforming of hydrocarbon
fuels. Catal. Today, 77, 51-64.
60
Chapter 2: Literature Review
Monnerat, B., Kiwi-Minsker, L. and Renken, A. (2001). Hydrogen production by
catalytic cracking of methane over nickel gauze under periodic reactor operation.
Chem. Eng. Sci., 56, 633-639.
Moulijn, J.A., van Diepen, A.E. and Kapteijn, F. (2001). Catalyst deactivation: is it
predictable? What to do? Appl. Catal. A: Gen., 212, 3-16.
Muller, A.C., Engelhard, P.A. and Weisang, J.E. (1979). Surface study of platinum-tin
bimetallic reforming catalysts. J. Catal., 56, 65-72.
Murata, K., Wang, L., Saito, M., Inaba, M., Takahara, I. and Mimura, N. (2004).
Hydrogen production form steam reforming of hydrocarbons over alkaline-earth
metal-modified Fe- or Ni-based catalysts. Energy & Fuels, 18, 122-126.
Murphree, E.V., Brown, C.L. and Gohr, E.J. (1940). Hydrogenation of petroleum. Ind.
Eng. Chem., 32, 1203-1212.
Nakagawa, K., Nishimoto, H., Kikuchi, M., Egashira, S., Enoki, Y., Ikenaga, N.,
Suzuki, T., Nishitani-Gamo, M., Kobayashi, T. and Ando, T. (2003). Synthesis
gas production from methane using oxidised-diamond-supported group VIII
metal catalysts. Energy & Fuels, 17, 971-976.
Niemantsverdriet, J. and van Langeveld, A.D. (1986). Coke formation on platinum
metals studied by Auger electron spectroscopy and secondary ion mass
spectrometry. Fuel, 65, 1396-1399.
Noto, Y., Fukada, K., Onishi, T. and Tamaru, K. (1967). Dynamic treatment of
chemisorbed species by means of infra-red technique. Mechanism of
decomposition of formic acid over alumina and silica. Trans. Faraday Soc., 63,
2300-2308.
Numaguchi, T. and Kikuchi, K. (1988). Intrinsic kinetics and design simulation in a
complex reaction network: steam-methane reforming. Chem. Eng. Sci., 43,
2295-2301.
Opoku-Gyamfi, K., Tafreshi, Z.M. and Adesina, A.A. (1998). Activities of Al
2
O
3
supported bi-metallic Pt-Ni and Co-Ni catalysts for CH
4
oxidation. React. Kinet.
Catal. Lett., 64, 229-238.
61
Chapter 2: Literature Review
Opoku-Gyamfi, K. and Adesina, A.A. (1999a). Forced composition cycling of a novel
thermally self-sustaining fluidised-bed reactor for methane reforming. Chem.
Eng. Sci., 54, 2575-2583.
Opoku-Gyamfi, K. and Adesina, A.A. (1999b). Kinetic studies of CH
4
oxidation over
Pt-NiO/o-Al
2
O
3
in a fluidised bed reactor. App. Catal. A: Gen., 180, 113-122.
Ordez, S., Sastre, H. and Dez, F.V. (2001). Thermogravimetric determination of
coke deposits on alumina-supported noble metal catalysts used as
hydrodechlorination catalysts. Thermochim. Acta, 379, 25-34.
Otsuka, K., Ogihara, H. and Takenaka, S. (2003). Decomposition of methane over Ni
catalysts supported on carbon fibers formed from different hydrocarbons.
Carbon, 41, 223-233.
Ozawa, Y. and Bischoff, K.B. (1968). Coke formation kinetics on silica-alumina
catalyst. Ind. Eng. Chem. Proc. Des. Dev., 7, 67-71.
Parera, J.M. and Beltramini, J.N. (1988). Stability of bimetallic reforming catalysts. J.
Catal., 112, 357-365.
Phillips, T.R., Mulhall, J. and Turner, G.E. (1969). The kinetics and mechanism of the
reaction between steam and hydrocarbons over nickel catalysts in the
temperature range 350-500C, Part 1. J. Catal., 15, 233-244.
Praharso, Adesina, A.A., Trimm, D.L. and Cant, N.W. (2004). Kinetic study of iso-
octane steam reforming over a nickel-based catalyst. Chem. Eng. J., 99, 131-
136.
Qin, D. and Lapszewicz, J. (1994). Study of mixed steam and CO
2
reforming of CH
4
to
syngas on MgO-supported metals. Catal. Today, 21, 551-560.
Querini, C.A. and Fung, S.C. (1997). Coke characterization by temperature
programmed techniques. Catal. Today, 37, 277-283.
Ramachandran, P.A. and Bhattacharya, A. (1983). Naphtha reforming kinetics -
Methane selectivity. Chem. Eng. Sci., 38, 865-870.
Ranganathan, E.S., Bej, S.K. and Thompson, L.T. (2005). Methanol steam reforming
over Pd/ZnO and Pd/CeO
2
catalysts. Appl. Catal. A: Gen., 289, 153-162.
62
Chapter 2: Literature Review
Renshaw, G.D., Roscoe, C. and Walker, P.L. (1971). Disproportionation of CO: II.
Over cobalt and nickel single crystals. J. Catal., 22, 394-410.
Richardson, J.T. (1972). Experimental determination of catalyst fouling parameters. Ind.
Eng. Chem. Proc. Des. Dev., 11, 12-14.
Rodriguez, J.C., Romeo, E., Fierro, J.L.G., Santamaria, J. and Monzon, A. (1997).
Deactivation by coking and poisoning of spinel-type of Ni catalysts. Catal.
Today, 37, 255-265.
Rosei, R., Ciccacci, F., Memeo, R., Mariani, C., Caputi, L.S. and Papagno, L. (1983).
Kinetics of carbidic carbon formation from CO in the 10
6
-torr range on
Ni(110). J. Catal, 83, 19-24.
Ross, J.R. and Steel, M.C.F. (1973). Mechanism of the steam reforming of methane
over a coprecipitated nickel-alumina catalyst. J. Chem. Soc. Faraday Trans. 1,
69, 10-21.
Rostrup-Nielsen, J.R. (1973). Activity of nickel catalysts for steam reforming of
hydrocarbons. J. Catal., 31, 173-199.
Rostrup-Nielsen, J.R. (1975). Steam Reforming Catalysts. Danish Technical Press,
Copenhagen.
Rostrup-Nielsen, J.R. and Trimm, D.L. (1977). Mechanisms of carbon formation on
nickel-containing catalysts. J. Catal., 48, 155-165.
Rostrup-Nielsen, J.R. (1997). Industrial relevance of coking. Catal. Today, 37, 225-232.
Rostrup-Nielsen, J.R. (2000). New aspects of syngas production and use. Catal. Today,
63, 159-164.
Rostrup-Nielsen, J.R. (2002). Syngas in perspective. Catal. Today, 71, 243-247.
Sahoo, S.K., Ray, S.S. and Singh, I.D. (2004). Structural characterization of coke on
spent hydroprocessing catalysts used for processing of vacuum gas oils. Appl.
Catal. A: Gen., 278, 83-91.
Sakai, N., Chida, T., Tadaki, T. and Shimoiizaka, J. (1985). Kinetics of carbon
deposition on nickel metal by decomposition of carbon monoxide. J. Chem. Eng.
Japan, 18, 199-204.
63
Chapter 2: Literature Review
Salmi, T., Bstrom, S. and Lindfors, L.-E. (1988). A dynamic study of the water-gas
shift reaction over an industrial ferrochrome catalyst. J. Catal., 112, 345-356.
Shamsi, A., Baltrus, J.P. and Spivey, J.J. (2005). Characterization of coke deposited on
Pt/alumina catalyst during reforming of liquid hydrocarbons. Appl. Catal. A:
Gen., 293, 145-152.
Shido, T. and Iwasawa, Y. (1993). Reactant-promoted reaction mechanism for water-
gas shift reaction on Rh-doped CeO
2
. J. Catal., 141, 71-81.
Sinfelt, J.H. (1973). Supported bimetallic cluster catalysts. J. Catal., 29, 308-315.
Sinfelt, J.H. (1979). US Patent, 3 953 368.
Sinfelt, J.H. (1983). Bimetallic Catalysts: Discoveries, Concepts and Applications.
Wiley, New York.
Sinfelt, J.H. (1985). Bimetallic catalysts. Scientific Am., 253, 90-98.
Sinfelt, J. H. (1987). Structure of bimetallic clusters. Acc. Chem. Res., 10, 134-139.
Sinfelt, J.H. (1997). Catalysis by alloys and bimetallic clusters. Acc. Chem. Res., 20, 15-
20.
Sinfelt, J.H. (2002). Role of surface science in catalysis. Surf. Sci., 500, 923-946.
Synetix (undated). Steam Reforming Catalysts: Natural Gas, Associated Gas and LPG
for Hydrogen Production. ICI Group, United Kingdom.
Takanabe, K., Katsutoshi, N., Nariai, K. and Aika, K. (2005). Titania-supported cobalt
and nickel bimetallic catalysts for carbon dioxide reforming of methane. J.
Catal., 232, 268-275.
Tibiletti, D., Goguet, A., Meunier, F.C., Breen, J.P. and Burch, R. (2004). On the
importance of steady-state isotopic techniques for the investigation of the
mechanism of the reverse water-gas-shift reaction. Chem. Commun., 14, 1636-
1637.
Tottrup, P.B. (1976). Kinetics of decomposition of carbon monoxide on a supported
nickel catalyst. J. Catal., 42, 29-36.
64
Chapter 2: Literature Review
Tottrup, P.B. (1982). Evaluation of intrinsic steam reforming kinetic parameters from
rate measurements on full particle size. Appl. Catal., 4, 377-389.
Tracz, E., Scoltz, R. and Borowiecki, T. (1990). High-resolution electron microscopy
study of the carbon deposit morphology on nickel catalysts. Appl. Catal., 66,
133-147.
Trimm, D.L. (1997). Coke formation and minimisation during steam reforming
reactions. Catal. Today, 37, 233-238.
Trimm, D.L., Adesina, A.A., Praharso and Cant, N.W. (2004). The conversion of
gasoline to hydrogen for on-board vehicle applications. Catal. Today, 93-95, 17-
22.
Twigg, M.V. (1994). Catalyst Handbook. Manson Publishing, London.
Xiancai, L., Min, W., Zhihua, L. and Fei, H. (2005). Studies on nickel-based catalysts
for carbon dioxide reforming of methane. Appl. Catal. A: Gen., 290, 81-86.
Xu, J. and Froment, G.F. (1989). Methane steam reforming, methanation and water-gas
shift: I. Intrinsic kinetics. AIChE J., 35, 88-96.
Van Doorn, J. and Moulijn, J.A. (1990). Extraction of spent hydrotreating catalysts
studied by Fourier transform infra-red spectroscopy. Fuel Process. Technol., 26,
39-51.
Vlter, J., Lietz, G., Uhlemann, M. and Hermann, M. (1981). Conversion of
cyclohexane and n-heptane on Pt-Pb/Al
2
O
3
and Pt-Sn/Al
2
O
3
bimetallic catalysts.
J. Catal., 68, 42-50.
Walker, P.L., Jr., Rusinko, F., Jr. and Austin, L.G. (1959). Gas reactions of carbon, in:
Eley, D.D., Selwood, P.W. & Weisz, P.B. (Eds.), Advances in Catalysis, vol. XI.
Academic Press, New York, pp. 133-221.
Wang, L., Murata, K. and Inaba, M. (2004a). Development of novel highly active and
sulphur-tolerant catalysts for steam reforming of liquid hydrocarbons to produce
hydrogen. Appl. Catal. A: Gen., 257, 43-47.
65
Chapter 2: Literature Review
Wang, L., Murata, K. and Inaba, M. (2004b). Control of the product ratio of
CO
2
/(CO+CO
2
) and inhibition of catalyst deactivation for steam reforming of
gasoline to produce hydrogen. Appl. Catal. B: Environ., 48, 243-248.
Yanhui, W. and Diyong, W. (2001). The experimental research for production of
hydrogen from n-octane through partially oxidizing and steam reforming
method. Int. J. Hydrogen Energy, 26, 795-800.
Zhang, Z.L. and Verykios, X.E. (1994). Carbon dioxide reforming of methane to
synthesis gas over supported Ni catalysts. Catal. Today, 21, 589-595.
Zhang, T. and Amiridis, M.D. (1998). Hydrogen production via the direct cracking over
silica-supported nickel catalyst. Appl. Catal. A: Gen., 167, 161-172.
Zhang, J., Wang, Y., Ma, R. and Wu, D. (2003). Characterization of alumina-supported
Ni and Ni-Pd catalysts for partial oxidation and steam reforming of
hydrocarbons. Appl. Catal. A: Gen., 243, 251-259.
Zhuang, Q., Qin, Y. and Chang, L. (1991). Promoting effect of cerium oxide in
supported nickel catalyst for hydrocarbon steam-reforming. Appl. Catal., 70, 1-
8.
66
Chapter 3: Experimental
C
C
h
h
a
a
p
p
t
t
e
e
r
r
3
3
E
E
X
X
P
P
E
E
R
R
I
I
M
M
E
E
N
N
T
T
A
A
L
L
3.1 Introduction
This section provides a list of chemicals and gases employed during catalyst preparation
and reaction studies, including their supplier, grade and main applications. Additionally,
it outlines the theoretical foundations and operational procedures of the catalyst
characterisation instruments. Finally, a detailed description of the reactor rig is
presented.
3.2 Materials
3.2.1 Chemicals
List of chemicals employed in this study is provided in Table 3.1. The metallic nitrates
and drierite particles were purchased from Sigma-Aldrich (Sydney, Australia), whilst
gamma alumina was obtained from Saint Gobain Nor-Pro (USA). Nitric acid and
ammonia solutions were supplied by Univar (Sydney, Australia). The preparation of all
solutions throughout this study utilised ultra pure water generated by a NANOpure
Diamond UV system (Barnstead International).
67
Chapter 3: Experimental
Table 3.1. List of chemicals used.
Chemical Formula Purity/conc. Application
Cobalt nitrate
Nickel nitrate
Gamma alumina
Nitric acid
Ammonia
Drierite
Co(NO
3
)
2
6H
2
O
Ni(NO
3
)
2
6H
2
O
Al
2
O
3
HNO
3
NH
4
OH
CaSO
4
> 99%
> 99%
Industrial
70% conc.
30% conc.
> 97%
Catalyst preparation
Catalyst preparation
Catalyst preparation
Catalyst preparation
Catalyst preparation
Moisture absorber
3.2.2 Gases
Table 3.2 displays a list of gases used in this investigation, as well as their purity and
specific application. All gases were supplied by Linde (Sydney, Australia), while
propane was, however, obtained from BOC Gases (Sydney, Australia).
3.3 Catalyst preparation
The properties of a catalyst depend strongly on preparation variables, such as method,
concentration of the reactants and experimental conditions. Consequently, a large extent
of investigation in this research was also devoted to exploring an improved catalyst
formulation. The optimised catalyst often corresponds to the one which has enhanced
activity, stability, selectivity and regenerability [Opoku-Gyamfi, 1999].
This section gives details of preparation procedure carried out in the synthesis of
catalysts used in this study. Further, various characterisation techniques employed for
68
Chapter 3: Experimental
measuring physicochemical properties of both fresh and spent catalysts will also be
discussed.
Table 3.2. List of gases used.
Gas Purity/concentration Application
C
3
H
8
Steam
He
Ar
H
2
CH
4
/Ar mixture
CO/N
2
mixture
CO
2
/Ar mixture
N
2
/He mixture
N
2
Air
NH
3
/He
O
2
> 99.9%
Vaporised from ultra pure
water
> 99.996%
> 99.999%
> 99.99%
CH
4
=1%; Ar=99%
CO=50%; N
2
=balance
CO
2
=49.95%; Ar=balance
N
2
=30.03%; He=balance
> 99.99%
> 99.9%
NH
3
=0.4%; He=balance
> 99.99%
Reactant
Reactant
Diluent
GC carrier gas
Catalyst reduction
GC standard gas
GC calibration
GC calibration
BET analysis
Inert for moisture removal/dilution
during thermal gravimetric runs
Oxidation of coked catalysts
Thermal-programmed desorption
experiments
Total carbon analysis
3.3.1 Catalyst support
High-temperature catalytic hydrocarbon processing often causes degradation of active
metallic crystallites. Metallic species in heterogenous catalysts are therefore anchored to
a support. A support generally must possess a superior thermal stability and be
chemically inert. In addition, the utilisation of support in heterogenous catalysts may
assist to increase anti-poisoning measure, enhance mechanical strength and reduce cost
by diluting the usually more expensive active ingredients by cheaper support materials
69
Chapter 3: Experimental
[Chen, 1995]. Further, as discussed in the previous chapter, the catalytic behaviour of a
catalyst can indeed be influenced by the type of support employed. From Tables 2.5 and
2.6, various investigators have used a range of materials, such as alumina, zirconia,
ceria, magnesia, carbon, titania, silica, lanthana, ceria-zirconia, ceria-alumina and
zeolites, as supports in steam reforming catalysts. Accordingly, the use of magnesia,
carbon, titania, silica and lanthana in most cases results in low steam reforming activity.
On the other hand, supports made of ceria-containing materials and zeolites would bring
about substantial improvement in conversion, but their potential at this time is still
limited by high-cost. Zirconia seems to have a comparable catalytic performance as
alumina, nevertheless in view of economic and availability factors, it is apparent that
alumina support is more popular for application in most industrial reforming catalysts.
Moreover, the operation of a fluidised bed steam reformer in this project also raises the
necessity of a mechanically-strong support material. As a result, alumina was also
adopted as the support of choice in this study.
3.3.2 Treatment of support
The type of phase in which alumina exists depends on the starting hydroxides and the
thermal pre-treatment conditions. Although there are several known phases of alumina,
only three of these are of interest particularly for this investigation the high surface
area -Al
2
O
3
, the porous amorphous o-Al
2
O
3
and the crystalline o-Al
2
O
3
.
The -Al
2
O
3
is generally produced by heating of AlOOH (Boehmite) at a temperature
no more than 573 K. It is known to exhibit higher surface area (50-300 m
2
g
-1
) and
superior mechanical stability compared to other alumina forms. However, application of
70
Chapter 3: Experimental
-Al
2
O
3
for high temperature processes is restricted by its tendency to transform to
another phase leading to structural degradation. In order to avoid phase modification
during subsequent preparation and steam reforming runs, it was therefore essential to
carry out a thermal treatment of alumina prior to metallic coating process. Support pre-
treatment was carried out by heating at 1073 K and holding for 6 h at this temperature.
The support treatment temperature was reasonably higher than the maximum calcination
temperature (973 K), whilst on the other hand it was also low enough to minimise
surface area loss due to thermal degradation. The subsequent XRD analysis (cf. Fig.
5.5) of the calcined sample suggested that the support remained in the -Al
2
O
3
phase.
The o-Al
2
O
3
phase (~ 130 m
2
g
-1
) was not identified as it usually appears at 1173 K.
However, the ultimate o-Al
2
O
3
(< 5 m
2
g
-1
) would form mostly above 1373 K. Despite
having lower areas, o- and o-Al
2
O
3
are useful in a number of applications requiring
high thermal stability such as methane steam reforming and CO
2
reforming of methane.
The characteristics of the alumina support used in this study are summarised in Table
3.3.
Table 3.3. Summary of support characteristics.
Parameter Remark
Phase -Al
2
O
3
Particle size range 180-250 m
BET surface area 169.88 m
2
g
-1
71
Chapter 3: Experimental
3.3.3 Impregnation catalyst preparation
Impregnation involves an addition of a solution of a metal salt to the support which then
undergoes a subsequent aging process. In general, there are two types of impregnation
methods, namely incipient wetness impregnation [Huang and Schwarz, 1987a] and
dipping impregnation [Huang and Schwarz, 1987b].
In the incipient wetness impregnation, a solution containing a calculated quantity of
metal compound giving a desired loading is prepared in a specific volume just sufficient
to fill up the pores of the support particles. However, when supports are in a powdery
nature, a substantially larger amount of solution may be added [Geus and van Veen,
1999]. The mixed slurry was then stirred to ensure homogenous distribution, followed
by drying and calcination. This technique has a major disadvantage in which metal
loadings are limited by solubility of precursors in the solution. This problem may,
however, be overcome by multiple impregnation steps [Opoku-Gyamfi et al., 1998].
In the dipping impregnation, the support, which can be pre-saturated with a solution
(wet impregnation) or dry (capillary impregnation) for desired active component profile
in the support particles [Lee and Aris, 1983], is immersed in a solution of the metal
compound. The resulting slurry is then treated under constant stirring for a specific
duration, followed by filtration, drying and calcination. As adsorption of metal
precursors depends largely on the concentration of the adsorption sites, it is not possible
to prepare a dipping impregnated catalyst with a predetermined metal loading [Opoku-
Gyamfi, 1999]. Since catalysts with specific metal loadings were required, incipient
72
Chapter 3: Experimental
wetness impregnation was employed in the preparation of all catalysts used in this
study.
Conventional nickel and coke-suppressant cobalt species were utilised as the active
metal components in a bid to develop an active, stable and affordable catalyst. Although
other precious metals such as Pt, Rh and Ru are known to be active and exhibit anti-
coking properties, price and availability of these metals have so far limited their
commercial applications [Mattos et al., 2003; Roh et al., 2003; Rostrup-Nielsen and
Bak-Hansen, 1993; Laosiripojana and Assabumrungrat, 2005].
3.3.4 Monometallic catalysts
Impregnation technique was employed to prepare the monometallic 20Co/80Al
2
O
3
and
20Ni/80Al
2
O
3
catalysts. Initially, support pore volume was determined by adding ultra-
pure water to a known quantity of Al
2
O
3
support powders. This was done carefully
drop-wise until incipient wetting of the support was evident. The pore volume was
found to be approximately 0.8 ml g
-1
. Based on this estimate, a quantity of ultra pure
water, which is twice the pore volume capacity, was added to a calculated amount of
Co(NO
3
)
2
6H
2
O salt to yield an aqueous solution of Co(NO
3
)
2
.
For 20Co/80Al
2
O
3
catalyst, impregnation was initiated by mixing of the Co(NO
3
)
2
solution with the thermally conditioned Al
2
O
3
support, followed by 3 h of constant
stirring on a hot plate at 303 K at a pH of 2. The low-pH level was thoroughly
maintained during stirring by addition of 3.0 M HNO
3
solution. Approximately 50 ml of
this acidic solution was required to keep the pH level at 2 throughout the 3 h duration.
73
Chapter 3: Experimental
Stirring rate was adjusted to a certain speed (6 on a 10-speed unit IEC Magnetic Stirrer)
which provided adequate mixing of Al
2
O
3
support particles in the slurry solution as to
ensure homogenous distribution. The resulting solution was dried overnight in an oven
at 393 K. The calcination of the product was carried out by ramping to 973 K at 5 K
min
-1
and holding at this temperature for 5 h with continuous air flow at 200 ml min
-1
.
The calcined solid was then crushed and sieved to 212-250 m for further use.
Preparation of 20Ni/Al
2
O
3
catalyst was done using an aqueous solution of Ni(NO
3
)
2
under conditions identical to those of 20Co/80Al
2
O
3
.
3.3.5 Bimetallic catalysts
The bimetallic 5Co-15Ni catalysts were prepared by impregnation under low (2) and
high (8) pH values. Initially, an aqueous Co(NO
3
)
2
solution was prepared based on the
required amount of cobalt nitrate to give the desired metal loading. A pre-weighed
thermally treated Al
2
O
3
was then transferred to the Co(NO
3
)
2
solution, continued by
gentle stirring on a hot plate at 303 K for 3 h under a constant pH level. In preparing a
typical catalyst specimen, for pH 2 about 50 ml of 3.0 M HNO
3
drops were gently
added during the stirring process, whilst approximately 80 ml of aqueous NH
4
OH (at
20% concentration) was utilised for the higher pH level of 8. Stirring was done using an
IEC Magnetic Stirrer at a stirrer speed scale of 6 (with 10 being the fastest), at which all
Al
2
O
3
particles appeared thoroughly mixed in a homogenous solution. Upon stirring, the
slurry was left overnight in an oven at 393 K for moisture removal. The dried solid was
subsequently impregnated with the required quantity of an aqueous Ni(NO
3
)
2
solution
under conditions similar to those of the first impregnation. The final product was
calcined in an oven using air flow under varying thermal conditions, i.e. temperature
74
Chapter 3: Experimental
(873-973 K), heating rate (5-20 K min
-1
) and holding time (1-5 h). The calcined solid
was crushed and sieved to 212-250 m for further use. An example of stoichiometric
calculation for catalyst preparation is provided in Appendix A.
3.4 Catalyst characterisation
Catalyst characterisation provides useful information on the physicochemical attributes
of the catalyst. Analytical techniques such as spectroscopic, microscopic, diffraction or
chemical analysis can be effectively used to investigate the catalyst surface and bulk
properties. Along with reaction data, information obtained from various characterisation
tools ultimately leads to a better understanding in correlating preparation variables with
the physicochemical attributes and catalytic performance. This section describes the
fundamental concepts of various characterisation techniques employed for this study.
3.4.1 Surface area
Most substances used as catalyst supports are porous in nature. These materials contain
deep and complicated network of pores accountable for internal surface area. Surface
area is obviously a key property of a porous material. Despite the fact that support
surfaces are not uniform in nature, higher surface area is often a good indication of a
more active catalyst. In such a case, metal crystallites can be more homogenously
distributed on the maximum possible surface and hence, an increase in adsorption sites
for reactant molecules.
The nature of adsorption is determined according to the magnitude of the binding force
between the gas entities and the solid atoms. When relatively weak forces are involved,
75
Chapter 3: Experimental
the adsorption resembles liquefaction and is termed physical adsorption or
physisorption. It is non-specific since this phenomenon is known to take place on
practically all solids.
For a unit mass of a given adsorbent, the amount of gas adsorbed, n, is a function of
equilibrium pressure, P, temperature, T, and the nature of the gas-solid system.
n = f(P, T, system) (3.1)
If the adsorbent temperature remains constant, whilst the gas temperature decreases
below its critical temperature, thus;
|
|
.
|

\
|
=
0
f
P
P
n
(3.2)
where P
0
= the saturation pressure of the adsorbate.
The most common technique used to measure solid surface areas is the Brunauer-
Emmett-Teller (BET) method, which stems from the work of Langmuir on adsorption
kinetic model [Brunauer et al., 1938]. In a pioneering work, Langmuir developed a
relationship between the amount of adsorbed gas and its equilibrium pressure at
constant temperature [Langmuir, 1916]. The following assumptions were made in his
model:
1. Gas is only adsorbed on a monomolecular layer,
76
Chapter 3: Experimental
2. The radius of action of the surface is very small,
3. Molecules can only be adsorbed on a free surface,
4. The molecules which collided with already adsorbed surface rebound elastically
and return to the gas phase.
The Langmuirs model for a monomolecular adsorption is generally expressed by:
|
.
|

\
|
+
=
n
n
o
o
o
u
0
1
(3.3)
where u = fraction of surface covered by adsorbed molecules,
o
0
= ratio of number of inelastic collisions resulting in adsorption to total
number of collisions of gas molecules on the surface,
n = number of adsorbed molecules desorbing in unit time,
= number of molecules colliding in unit time with a unit area of the surface.
Based on Langmuirs monomolecular model, Brunauer, Emmett and Teller [1938]
derived an expression for multilayer adsorption, given by,
( )
( )
s m m s
P cV
P c
cV P P V
P 1 1
+ =

(3.4)
where P = gas pressure,
P
s
= saturation pressure of adsorbed gas,
V = volume of gas adsorbed,
V
m
= volume of adsorbed gas corresponding to monolayer coverage,
77
Chapter 3: Experimental
c = a constant, characteristic of the adsorbate.
A plot of P/V(P
s
-P) versus P/P
s
provides a straight line with intercept and slope of
1/cV
m
and (c-1)/cV
m
respectively from which the volume of adsorbed gas corresponding
to monolayer coverage, V
m
, may be calculated. The surface area per weight unit can
subsequently be computed from
20
10

=
M
N a n
S
m m
A
(3.5)
where S
A
= surface area of solid (m
2
g
-1
),
a
m
= average area occupied by a molecule,
N = Avogadros number,
M = molecular weight of adsorbate,
n
m
= monolayer capacity of adsorbate (g adsorbate g
-1
solid).
The laboratory measurement of BET surface areas was performed using the
Micromeritics Autochem 2910 unit equipped with a TCD. Approximately 0.1 g of
catalyst, supported on quartz wool inside a U-shaped sample tube, was utilised for each
analysis. The sample was initially dried in flowing nitrogen at 423 K for 1 h and
subsequently cooled to ambient condition. Following pre-treatment, the flowing gas was
switched to a measuring gas (30% N
2
in He). Once TCD signal has stabilised, the
sample tube was then quickly immersed in a Dewar flask filled with liquid N
2
. As a
result, nitrogen was physisorbed on the support surface. When adsorption has reached
equilibrium, the sample temperature was shifted back to ambient by swapping the liquid
N
2
Dewar flask with a water-filled Dewar flask. This caused immediate release of
78
Chapter 3: Experimental
nitrogen adsorbed by the surface. The adsorption and desorption of nitrogen species was
detected by changes in TCD signal as typically shown in Fig. 3.1. Consequently, the
surface area was determined by equations;
|
|
.
|

\
|

|
|
.
|

\
|
+
=
Hg mm 760 K 273.15
K 273.15
a
a
a
STP
P
T SW
V
V
(3.6)
|
|
.
|

\
|
=
0
1
P
P
V V
STP m
(3.7)
( ) ( ) ( )
2
23
10 023 . 6
414 , 22
N
m
BET
A
V
S =
(3.8)
where V
STP
= volume sorbed at STP (ml g
-1
sample ),
V
a
= volume sorbed at ambient conditions (ml),
V
m
= volume of the monolayer (ml),
T
a
= ambient temperature (C),
P
a
= ambient pressure (mm Hg),
P = absolute pressure of nitrogen (estimated by % N
2
P
a
) (mm Hg),
S
BET
= surface area (m
2
g
-1
),
2
N
A = surface area of the N
2
molecule (m
2
molecule
-1
),
SW = sample weight (g),
P
0
= saturation pressure of nitrogen (estimated by P
a
+ 15 mm Hg) (mm Hg).
Due to its better symmetry, the desorption peak instead of the adsorption peak is most
often used as the basis for surface area determination.
79
Chapter 3: Experimental
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
0 2 4 6 8
Time (min)
T
C
D

s
i
g
n
a
l
50
100
150
200
250
300
350
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 3.1. Typical adsorption and desorption peaks during a BET measurement.
3.4.2 Hydrogen chemisorption
In contrast to physisorption, chemisorption occurs when the binding forces between the
gas molecules and solid atoms are comparable to those characterising chemical
compound formation. This is a specific phenomenon since certain solids have sites that
possess the necessary energy requirements. As a result, information regarding the nature
of metallic sites deposited on a support may therefore be extracted from chemisorption
analysis.
In chemisorption, a known quantity of a selective gas was injected to pass through the
sample until the adsorption has reached equilibrium (i.e. amount of gas injected equals
to amount exiting as detected by TCD detector). Hydrogen is one of the most commonly
used adsorbate gas for chemisorption [Augustine et al, 1996] on Group VIII metals. H
2
chemisorption takes places dissociatively on a metal atom, as shown by,
80
Chapter 3: Experimental
H 2M 2M H
2
+ (3.9)
Based on stoichiometry, one hydrogen atom is chemisorbed on a metal surface atom
(i.e. H:M = 1). Given the assumed stoichiometry and the known quantity of adsorbate
gas uptake, chemisorptive properties, such as dispersion, metallic surface area and
active particle size, may be evaluated.
A Micromeritics Autochem 2910 instrument was employed to perform pulse H
2
chemisorption. Approximately 0.1 g of the specimen was initially loaded into a sample
holder. The catalyst was subsequently reduced under flowing hydrogen at 873 K for 2 h
with a heating rate of 2 K min
-1
. This condition was selected to ensure that the calcined
sample was completely reduced. Complete catalyst reduction may be implicated by a
weight drop of at least 5.57% as suggested by theoretical calculation in Appendix D.2.
As shown later from the TPR analysis (cf. Fig. 3.5), a weight drop larger than 5.57%
signalling complete reduction was observed after ramping at 5 K min
-1
to 973 K and
holding at this temperature for 1 h under 50% H
2
/N
2
flow. Hence, reduction conditions
prior to chemisorption which were conducted at much slower heating rate, longer
duration and using pure hydrogen (albeit at slightly lower temperature) compared to the
TPR analysis would also be expected to reduce the catalyst completely. Upon reduction,
the gas stream was then switched to argon and the sample was cooled down to room
temperature. The temperature was then increased at a constant rate of 2 K min
-1
to 373
K and holding for 30 min. Following TCD signal stabilisation, chemisorption
commenced with a series of injections of 1-ml hydrogen doses which passes through the
81
Chapter 3: Experimental
specimen. The flow of hydrogen gas unadsorbed and hence, exiting the system, was
monitored by the TCD detector. The injections ceased either when a peak possessed the
same size as its predecessor or after 20 injections.
Consequently, the volume removed from the injections by the sample may be computed
from:
( )
na i S
V V l V =
(3.10)
where V
S
= volume sorbed (ml),
l = number of injections,
V
i
= volume per injection (ml),
V
na
= total volume not sorbed (ml).
In the cases of 2 or more metal species (i.e. bi- or multi-metallic catalysts) are
employed, the following correlations are used to estimate required parameters (i.e.
stoichiometric factor, molecular weight, specific surface area and metal density) for the
combined catalyst system.
For calculated stoichiometric factor,
calc
atomicN
N N
atomic atomic
calc
MW
W
SF F
W
SF F
W
SF F
SF
(

|
|
.
|

\
|
+ +
|
|
.
|

\
|
+
|
|
.
|

\
|
= ...
2
2 2
1
1 1
(3.11)
where SF
calc
= calculated stoichiometric factor,
SF
N
= stoichiometry factor for metal N,
82
Chapter 3: Experimental
F
N
= fraction of sample weight for metal N,
W
atomicN
= molecular weight of metal N (g mol
-1
),
MW
calc
= molecular weight (g mol
-1
).
For calculated molecular weight,
|
|
.
|

\
|
+ +
|
|
.
|

\
|
+
|
|
.
|

\
|
=
atomicN
N
atomic atomic
calc
W
F
W
F
W
F
MW
...
1
2
2
1
1
(3.12)
where MW
calc
= molecular weight (g mol
-1
),
F
N
= fraction of sample weight for metal N,
W
atomicN
= molecular weight of metal N (g mol
-1
).
For calculated specific surface area,
calc
atomicN
N N
atomic atomic
calc
MW
W
SA F
W
SA F
W
SA F
SA
(

|
|
.
|

\
|
+ +
|
|
.
|

\
|
+
|
|
.
|

\
|
= ...
2
2 2
1
1 1
(3.13)
where SA
calc
= calculated specific area,
SA
N
= specific area for metal N,
F
N
= fraction of sample weight for metal N,
W
atomicN
= molecular weight of metal N (g mol
-1
),
MW
calc
= molecular weight (g mol
-1
).
For calculated metal density,
83
Chapter 3: Experimental
calc
atomicN
N N
atomic atomic
calc
MW
W
D F
W
D F
W
D F
D
(

|
|
.
|

\
|
+ +
|
|
.
|

\
|
+
|
|
.
|

\
|
= ...
2
2 2
1
1 1
(3.14)
where D
calc
= calculated density,
D
N
= density for metal N,
F
N
= fraction of sample weight for metal N,
W
atomicN
= molecular weight of metal N (g mol
-1
),
MW
calc
= molecular weight (g mol
-1
).
The percent metal dispersion was determined by,
calc
calc S
MW
SW
SF V
PD |
.
|

\
|

=
414 , 22
100
(3.15)
where PD = percent dispersion,
V
S
= volume sorbed (ml at STP),
SF
calc
= calculated stoichiometry factor,
SW = sample weight (g),
MW
calc
= molecular weight (g mol
-1
).
Whilst the effective metallic surface area was computed from:
( ) ( ) ( )
calc calc
T S
m
SA SF
SW
F V
S |
.
|

\
|

=
23
10 023 . 6
414 , 22
(3.16)
where S
m
= metallic surface area (m
2
g
-1
sample),
V
S
= volume sorbed (ml at STP),
F
T
= fraction of sample weight of all metals,
84
Chapter 3: Experimental
SF
calc
= calculated stoichiometry factor,
SA
calc
= calculated specific surface area,
SW = sample weight (g).
Further, the following equation was used to calculate the active particle size,
( ) ( ) ( ) ( ) PD SA
MW
D
APS
calc
calc
calc

|
|
.
|

\
|

=
23
10 023 . 6
1
6
(3.17)
where APS = active particle size,
D
calc
= calculated metal density (g ml
-1
),
MW
calc
= molecular weight (g mol
-1
),
SA
calc
= calculated specific surface area,
PD = percent dispersion.
3.4.3 Temperature-programmed desorption
Temperature-programmed desorption (TPD) measures the amount of desorbed species
under heating at a constant rate. This information, along with desorption peak
temperature, may be used to estimate the number, type and strength of active sites
available on the surface of a catalyst. The catalyst specimen was initially contacted by
the adsorptive gas, and followed by temperature ramping at a specific rate. While
temperature was being increased, the activation energy would, at some stage, be
overcome which caused the bonds between adsorbate and adsorbent to break resulting
in desorption of reacted species. As a result, desorbed gas was released to the main
stream and so, the subsequent change in concentration was monitored by the TCD
85
Chapter 3: Experimental
detector. The areas under desorption peaks correspond to the amounts of the desorbed
gas.
The TPD experiments were conducted in an Autochem 2910 (Micromeritics)
instrument. A typical analysis used approximately 0.1 g of catalyst specimen which was
supported on quartz wool inside a sample holder. Sample pre-treatment involved drying
using helium flow at 573 K for 30 min (5 K min
-1
heating rate). The temperature was
then lowered to 423 K, while the gas flow was switched to 0.4% NH
3
. The catalyst was
treated under this condition for 60 min to ensure complete adsorption of all available
sites by NH
3
molecules. Subsequently, NH
3
flow was terminated and desorption process
started by ramping at a constant rate of 10 K min
-1
to 973 K and holding for 60 min. As
a result, thermal energy caused the bonds between NH
3
and metal atoms to rupture
giving rise to desorption of NH
3
molecules. The released NH
3
was swept with inert
helium stream which was continuously passing through the sample. Once the active
sites had been cleaned from NH
3
residues, the temperature was cooled to 423 K and the
catalyst was again re-charged with NH
3
species. However, in the second cycle,
desorption was performed with a higher heating rate of 15 K min
-1
to 973 K and
maintained for 60 min. Similar procedure was repeated for a third and a fourth cycle,
which employed ramping rates of 20 K min
-1
and 25 K min
-1
respectively. The changes
in TCD response due to desorbed NH
3
molecules at varying heating rates are typically
shown in Fig. 3.2.
The determination of heat of desorption, -AH
des
, from TPD experiments requires at least
2 runs at different heating rates. The general equation is given by,
86
Chapter 3: Experimental
C R
A H
T R
H
T
g
sat des
p g
des
p
) (
ln
) ( ln
2

=
|
(3.18)
where | = ramp rate (K min
-1
),
-AH
des
= heat of desorption (kJ mol
-1
K
-1
),
R
g
= gas constant,
T
p
= temperature at peak maximum (K),
A
sat
= the quantity adsorbed at saturation,
C = a constant, related to the desorption rate.
A plot of (ln |)/T
p
2
versus 1/T
p
yields a slope and an intercept of -AH
des
/R
g
and ln (-
AH
des
A
sat
/(R
g
C)) respectively, from which the value of heat of desorption, -AH
des
, can be
computed.
The heat of desorption, -AH
des
, and desorption peak temperature provide indication
which reflects energetic characteristics of the catalyst under different heating rates.
Bronsted acid sites exhibits -AH
des
in the range 125-145 kJ mol
-1
, while Lewis sites is
characterised by much lower -AH
des
-value [Yaluris et al., 1996]. From the previous
studies of Santacesaria and co-workers [Auroux et al., 2001; Iengo et al., 1998a &
1998b], the strength of acid sites have been classified according to their desorption peak
temperatures into three classes: weak (< 523 K), medium (523-673 K) and strong (> 673
K) acid sites.
87
Chapter 3: Experimental
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
373 473 573 673 773 873 973 1073
Temperature (K)
T
C
D

s
i
g
n
a
l
10 K/min
15 K/min
20 K/min
25 K/min
973
Fig. 3.2. Typical desorption profiles of different heating rate.
3.4.4 Thermal gravimetric analysis
Thermal gravimetric analysis (TGA) measures weight changes of a sample when
subjected to linear thermal programming. During a run, an analysis gas passes through
the sample and reacts with the solid. Depending on the nature of reaction, a new species
may be formed which may lead to either increase or decrease in sample weight. The
reduction in sample weight is usually attributed to release of by-product produced in the
reaction. In a TGA unit, the real-time sample weight is monitored by an on-line
measurement system.
Thermogravimetric run was performed using a ThermoCahn TGA 2121 unit using
approximately 0.1 g sample for each measurement. The types of experiments carried out
are thermally-sensitive processes mimicking various stages in catalyst preparation such
88
Chapter 3: Experimental
as calcination and combined reduction-oxidation [Kapteijn et al., 1999]. Reduction was
followed by oxidation to investigate catalyst regenerability.
3.4.4.1 Calcination
Calcination runs were carried out on uncalcined specimens under continuous flow of
high-purity air (55 ml min
-1
) through the sample chamber. During temperature ramping,
metal nitrate decomposed to give the corresponding oxides. As a result of the evolution
of NO
2
(and/or H
2
O), the process is accompanied by a loss in weight with time-on-
stream. Furthermore, the derivative of the weight percent profile provides valuable
information on the type of oxides formed. Figs. 3.3 and 3.4 show weight and weight
derivative profiles for a typical calcination run. All spectra obtained from TGA
experiments have been normalised based on specimen initial weights.
60
70
80
90
100
110
0 0.5 1 1.5 2 2.5
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
%
)
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 3.3. Typical catalyst weight profile during calcination.
89
Chapter 3: Experimental
0 0.5 1 1.5 2 2.5
Time (h)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 3.4. Typical catalyst weight derivative profile during calcination.
3.4.4.2 Temperature-programmed reduction and oxidation
Calcined catalysts were subjected to combined temperature-programmed reduction
(TPR) and oxidation (TPO) runs. These runs were initiated with a drying pre-treatment
using inert nitrogen flow at 423 K for 1 h. Subsequently, TPR was carried out by
thermal ramping at 5 K min
-1
to 973 K and holding at this level for 1 h, succeeded by
cooling to ambient. Upon completion of TPR, it is immediately followed by TPO via a
similar temperature programming scheme. TPR was conducted with 50% H
2
/N
2
, while
high-purity air was utilised for TPO. The flow rates of both gases were set at 55 ml min
-
1
.
The TPR-TPO runs were carried out to investigate reducibility and re-oxidation
characteristics of the oxide phases present after calcination. During reduction (TPR), H
2
reacted with metallic oxides forming metal species and water molecules. The weight
90
Chapter 3: Experimental
drop profile obtained from this analysis is therefore indicative of catalyst reducibility.
The derivative of the TPR weight spectra reveals the peak reduction temperatures and
different phases of the oxides present. On the other hand, re-oxidation (TPO) runs show
the extent in which the reduced catalysts may be re-oxidised. The various oxidation
stages involved are revealed in the TPO derivative spectra. The combined TPR-TPO run
may also be useful to examine the catalyst stability during reaction since steam
reforming environment is a variable mixture of reducing and oxidising conditions. Figs.
3.5, 3.6 and 3.7 show typical profiles for TPR-TPO weight change, TPR weight
derivative and TPO weight derivative respectively.
90
92
94
96
98
100
102
0 2 4 6 8 10 12
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
%
)
273
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
TPR TPO
N
2
Fig. 3.5. Typical catalyst weight profile during reduction-oxidation.
91
Chapter 3: Experimental
0 0.5 1 1.5 2 2.5 3
Time (h)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 3.6. Typical catalyst weight derivative profile during reduction.
0 0.5 1 1.5 2 2.5 3
Time (h)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
273
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 3.7. Typical catalyst weight derivative profile during oxidation.
92
Chapter 3: Experimental
3.4.4.3 Reproducibility analysis
Reproducibility of the TGA experiment was scrutinised by performing two separate
runs of calcination under the identical thermal condition over the same catalyst. A
comparison of the weight change profiles (cf. Fig. 3.8) shows strikingly identical curves
suggesting that the results obtained from the TGA instrument were highly reproducible.
60
70
80
90
100
110
373 473 573 673 773 873 973 1073
Temperature (K)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
%
)
Fig. 3.8. Reproducibility analysis of the TGA experiment.
3.4.4.4 Solid-state kinetic analysis
Furthermore, transient weight measurements obtained from thermal analyses may be
utilised to evaluate solid kinetic constant which may be correlated with subsequent
catalyst performance. For non-catalytic gas-solid reaction (such as calcination,
reduction or oxidation), the solid conversion, o, may therefore be estimated from,
93
Chapter 3: Experimental
f i
i
w w
w w

= o
(3.19)
where o = solid-state conversion,
w = weight at instantaneous time,
w
i
= initial weight,
w
f
= final weight.
Consequently, conversion rate may be expressed as,
( ) o
o
f
s
k
dt
d
= (3.20)
where k
s
= solid-state kinetic constant,
f(o) = conversion function.
For solid-state reactions, the most commonly-used kinetic models are Avrami-Erofeev
(A), Prout-Tompkins (B), geometrical (R), diffusion (D) and order of reaction (F).
Model selection is usually based on the characteristic of conversion-time profile and
statistical fit as discussed in Brown [2001] and Kapteijn et al. [1999].
3.4.5 X-ray diffraction
X-ray diffraction has been widely used for identifying crystalline phases present in
catalyst. Additionally, from this technique the mean crystallite size in the range 3 to 50
nm can be estimated. When the crystallite size is smaller than 3 nm, the X-ray pattern
would show a broad and diffuse, or even absent, diffraction lines. On the other hand, in
crystallite larger than 50 nm, the change in the line shape would not be apparent
94
Chapter 3: Experimental
[Anderson and Pratt, 1985]. As a result, the application of XRD is principally limited in
two ways: firstly, it does not detect amorphous phases and secondly, optimum analysis
may be achieved when crystallite size is within the effective regime (3 to 50 nm).
In XRD analysis, the catalyst powder is irradiated with an X-ray of known wavelength,
, and hence, diffraction takes place. The angle at which constructive interference
occurs is subsequently measured. Accordingly, the interplanar distances or d spacings
of the crystals may be determined from the Bragg equation [Cullity, 2001]:
u sin 2d n = (3.21)
where n = order of reflection (integer),
= wavelength of incident radiation (),
d = interplanar distance of the crystal (),
u = angle of incidence (degrees).
The d spacings, along with the relative intensity of the diffraction lines, constitute the
diffraction pattern for a given crystal. Now, if assuming the diffraction line shapes are
Gaussian, therefore the squares of the contributing width factors are additive. The line
width due to particle size broadening is thus defined as,
2 2 2
inst obs d
| | | = (3.22)
where |
d
= true line width at half maximum intensity,
|
obs
= observed width at half maximum intensity,
|
inst
= instrumental line width by standard.
95
Chapter 3: Experimental
|
inst
is obtained from a calibration process using a standard of high quality with a
crystalline size greater than 1000 . Consequently, the mean crystallite size may be
obtained by the Scherrer equation expressed as [Liebhiafsky et al., 1972],
u |

cos
d
Sch
k
D =
(3.23)
where D = crystalline size (),
= wavelength of X-ray (),
|
d
= angular width of half maximum intensity (degree),
u = Braggs angle (degree),
k
Sch
= Scherrer constant and equals to 51.
XRD measurements of the freshly-calcined, freshly-reduced and spent catalysts were
carried out on a Philips XPert system using Ni-filtered CuKo ( = 1.542 ) at 40 kV
and 30 mA. The catalyst samples were initially crushed to a fine powder (< 100 m).
The specimens were then placed on a glass specimen holder and pressed using a glass
slide. Scanning of sample was then performed starting from 5 to 80 at a speed of 1
per minute. Peaks obtained from the analysis were identified using the Hanawalt search
match interpretation method to reveal the type of phases present [JCPDS, 1983].
3.4.6 Transmission electron microscopy
The transmission electron microscopy (TEM) technique has several important
applications in catalyst characterisation. These include, determination of average
96
Chapter 3: Experimental
particle size and particle size distribution, observation of topography and morphology,
and examination and interpretation of contrast features shown by specimen images (i.e.
extinction contours, phase contrast features and lattice images).
The operational overview of a TEM unit is illustrated in Fig. 3.9. The electron gun
produces a beam of high energy and high intensity electrons, which passes through
condenser lenses resulting in parallel streams. The produced streams strike the sample,
where portions of electrons are transmitted depending on the thickness and density of
the corresponding specimen [Crorkendorff and Niemantsverdriet, 2003]. The
transmitted portion is focused by the objective lens onto a two-dimensional image on
the image plane. The thicker or denser areas of the catalyst are revealed by dark-field
image, whereas bright-field portion of the image indicate thinner or less dense sections.
The TEM micrographs in this investigation were obtained using a Hitachi H-7000
microscope. The catalyst specimens were ground to a fine powder and ultrasonically
dispersed in ethanol. A few drops of the suspension were transferred onto an electron-
transparent carbon-coated copper grid followed by drying [Freel, 1972]. The focussed
images were taken at desired magnification and recorded by a computer.
97
Chapter 3: Experimental
electron gun
condenser
lenses
sample
aperture
objective
lens
lenses
image plane
Fig. 3.9. Schematic of a transmission electron microscope (TEM) [Adapted from
Crorkendorff and Niemantsverdriet, 2003].
98
Chapter 3: Experimental
3.4.7 Scanning electron microscopy
In contrast to TEM, scanning electron microscopy (SEM) generates contrasted images
with 3-dimensional insights of catalyst surfaces. In addition to morphology (size and
shape), SEM images also provide additional information on topography (surface
features) and crystallographic (atomic arrangement).
All SEM micrographs obtained in this study were taken using a Cambridge S360 unit.
Fig. 3.10 presents a schematic illustration of SEM instrument. In SEM unit, an electron
gun is also utilised for producing an intense electron beam. As the beam strikes the
metal-coated specimen, secondary and backscattered electrons emitted from the area.
The yield of secondary and backscattered electrons captured by specific detectors gives
rise to the imaging process. The contrast of the image is hence due to the difference in
the electron collection efficiency which depends on the angle of emission, surface relief
and atomic number of elements [Bergeret and Gallezot, 1997].
99
Chapter 3: Experimental
electron gun
condenser
lenses
scan coils
objective lens
X-ray
detector
electron detector
sample
Fig. 3.10. Schematic of a scanning electron microscope (SEM) [Adapted from
Crorkendorff and Niemantsverdriet, 2003].
100
Chapter 3: Experimental
3.4.8 Total carbon analysis
Carbon deposition during steam reforming leads to poor catalyst and reactor
performance. As a great deal of this project is dedicated to study deactivation, it is
important that carbon content in used catalysts be determined. A Shimadzu Solid
Sample Module SSM-5000A attached to a Total Organic Carbon (TOC) Analyzer
5000A was employed to measure carbon content in used specimens. This analysis
measures total carbon analysis which includes both organic and inorganic carbonaceous
materials. Each analysis utilised approximately 10 mg of catalyst specimen. Inside the
sample boat, the specimen was covered by a strip of quartz wool to prevent particle
scattering resulting from high-temperature combustion.
Total carbon analysis was performed based on a combustion process which occurred
inside a chamber at 1173 K. For a sample placed in the chamber, the carbon component
in the sample reacted with O
2
resulting in the formation of CO
2
. The product gas was
then swept out by a continuous flow of carrier gas (high-purity oxygen). Further, the
flow was cooled and dried and subsequently passed into a non-dispersive infrared gas
analyser (NDIR) for CO
2
measurement. The NDIR generated a CO
2
peak which is
proportional to the carbon content in the sample. Subsequently, the total carbon may be
computed from the peak area using the pre-determined calibration curve relating peak
area and carbon content.
3.5 Experimental apparatus
The schematic diagram of the reactor configuration is presented in Fig. 3.11. The same
set-up was implemented in all reaction runs. Hydrogen cylinder was connected to the
101
Chapter 3: Experimental
system for reductive catalyst activation performed prior to actual steam reforming
experiment. Following reduction, the hydrogen flow was ceased and the flow to the
reactor was switched to propane and diluent helium.
The experimental configuration was divided into four partitions, which consist of flow
controller unit, steam generator, fluidised bed reactor rig and product analysis unit.
1 3 2
M
SG
SP
F
F
B
R
MT
D
C
VENT
GC
KEY:
1 Hydrogen FBR Fluidized Bed Reactor Manual Valve
2 Propane F Furnace 3-way Manual Valve
3 Helium MT Manometer Actuated Valve
M Mixer C Condenser Teflon Tube
SG Steam Generator D Drierite Bed Insulation
SP Syringe Pump GC Gas Chromatograph
Fig. 3.11. Schematic diagram of the experimental set-up.
3.5.1 Flow controller units
Flow rates of gases were regulated using Brooks electronic mass flow controllers
(model 5850 E). These units were connected to a digital control panel which allows the
102
Chapter 3: Experimental
adjustment of the gas flow on a scale between 0.0 to 100.0%. Each mass flow controller
unit was carefully calibrated using a bubble-flow meter. The measured maximum flow
capacities of the controller units are 200, 70 and 400 ml min
-1
for H
2
, C
3
H
8
and He
respectively. A stainless steel chamber (110-ml volume) is used to mix incoming
propane substrate and diluent helium prior to pre-heating through the steam generator.
3.5.2 Steam generator
The steam generator unit consists primarily of a U-shaped tube and a cylindrical mixing
vessel, both made of stainless steel. Ultra pure water was injected from a 20 or 30 ml
glass syringe by a Razel A-99 (220 V, 0.1 A, 50 Hz) syringe pump into the U-shaped
tube housed inside the steam generator. The water injection rate was adjusted in the
range between 0 and 99 units, with the maximum capacity of 70.0 and 93.9 ml h
-1
for 20
and 30 ml syringe respectively. Water vaporisation occurs as a result of thermal
radiation emitted by heating coil inside the steam generator. The steam generation
temperature was set at 473 K. The produced steam was then mixed with the
propane/helium mixture in a mixing chamber within the insulated steam generator unit.
The pre-heated, well-mixed reactant mixture was transferred to the reactor unit via an
insulated line equipped with heating elements operated at 523 K in order to prevent
steam condensation.
3.5.3 Fluidised bed reactor rig
The fluidised bed reactor was constructed from a cylindrical quartz glass with 20 mm
i.d. and 490 mm length. Fig. 3.12 illustrates the fluidised bed reactor set-up. The tube
was placed vertically in a Ceramic Engineering (Sydney, Australia) electrical furnace
103
Chapter 3: Experimental
(240 V, 10 A, 1800 W). A 3-mm o.d. stainless steel thermocouple was positioned
axially in the catalyst bed to measure the reaction temperature. Catalyst loading inside
the tube was supported by 3 mm thick sintered quartz (50 m holes), located 10 cm
from the bottom end, which also acts as a gas distributor. About 1 cm thick of quartz
wool strip was fitted at the reactor outlet to prevent any entrainment of catalyst particles
during fluidisation. Both ends of reactor tubes were sealed with Teflon fittings. The
pressure drop across the catalyst bed was measured by a U-tube water manometer (5
mm o.d.) which was mounted across the reactor. Sections of tubular reactor exposed at
both top and bottom ends of the furnace were covered with kaowool for insulation as to
minimise temperature variation. The line connecting the reactor outlet and the
condenser unit was also insulated and heated at 523 K to avoid condensation of
unreacted steam.
3.5.4 Product analysis
Prior to product analysis by gas chromatography, steam was removed in a glass
condenser and a drierite packing. The condenser (150 ml volume) was immersed in a 5 l
ice-water bath. Water trapped inside the condenser may be evacuated via a liquid
drainage system fitted at the base. Drying of product gas was further achieved by
subsequent water-absorbing drierite granules (8 mesh) packed inside a 15-cm length
glass tube (1 cm o.d.). The drierite particles were immediately replaced once saturated
with moisture which was indicated by colour change from blue to pink.
The dried product gas was then sent to a Shimadzu gas chromatograph (model 8A)
fitted with a thermal conductivity detector. The separation of He, H
2
, CO, CH
4
, CO
2
and
104
Chapter 3: Experimental
C
3
H
8
components was achieved by an Alltech Haysep DB column. The GC was
operated isothermally at 393 K using argon at 30 ml min
-1
as the carrier gas. The peak
area corresponding to each component was measured by a Shimadzu C-R8A integrator.
Since peak areas from GC analysis are often influenced by system or atmospheric
conditions, therefore prior to each experiment a 1% CH
4
(argon balance) standard gas
was injected for several times. The data obtained from standard gas injections were used
to correct variations in peak areas due to random fluctuations [Opoku-Gyamfi, 1999].
105
Chapter 3: Experimental
Manometer
Product
Out
Distributor Plate
Fluidised
Catalyst Bed
Thermowell
Kaowool
Thermocouple
Feed In
Fig. 3.12. Fluidised bed reactor.
106
Chapter 3: Experimental
References
Anderson, J.R. and Pratt, K.C. (1985). Introduction to Characterization and Testing of
Catalysts. Academic Press, Sydney.
Augustine, R.L. (1996). Heterogeneous Catalysis for the Synthetic Chemist. Marcel
Dekker, Inc., New York.
Auroux, A., Monaci, R., Rombi, E., Solinas, V., Sorrentino, A. and Santacesaria, E.
(2001). Acid sites investigation of simple and mixed oxides by TPD and
microcalorimetric techniques. Thermochim. Acta, 379, 227-231.
Bergeret, G. and Gallezot, P. (1997). Particle size and dispersion measurements, in: Ertl,
G., Knzinger, H. & Weitkamp, J. (Eds.), Handbook of Heterogeneous
Catalysis, vol. 2. VCH, Weinheim, pp. 439-464.
Brown, M.E. (2001), Introduction to Thermal Analysis: Techniques and Applications.
Kluwer Academic Publishers, The Netherlands.
Brunauer, S., Emmett, P.H. and Teller, E. (1938). Adsorption of gases in
multimolecular layers. J. Am. Chem. Soc., 60, 309-319.
Chen, H. (1995). A study of a Co-Mo bimetallic catalyst system for Fischer-Tropsch
synthesis. Ph.D. thesis, University of New South Wales, Sydney, Australia.
Crorkendorff, I. and Niemantsverdriet, J.W. (2003). Concepts of Modern Catalysis and
Kinetics. Wiley-VCH, Weinheim.
Cullity, B.D. (2001). Elements of X-ray Diffraction. Prentice-Hall International,
London.
Freel, J. (1972). Chemisorption on supported platinum: I. Evaluation of a pulse method.
J. Catal., 25, 139-148.
Geus, J.W. and van Veen, J.A.R. (1999). Preparation of supported catalysts, in: van
Santen, R.A., van Leewen, P.W.N.M., Moulijn, J.A. & Averill, B.A. (Eds.),
Catalysis: An Integrated Approach. Elsevier, Amsterdam, pp. 459-485.
Huang, Y.J. and Schwarz, J.A. (1987a). The effect of catalyst preparation on catalytic
activity: IV. The design of Ni/Al
2
O
3
catalysts prepared by incipient wetness.
Appl. Catal., 32, 59-70.
107
Chapter 3: Experimental
Huang, Y.J. and Schwarz, J.A. (1987b). The effect of catalyst preparation on catalytic
activity: II. The design of Ni/Al
2
O
3
catalysts prepared by wet impregnation.
Appl. Catal., 30, 255-263.
Iengo, P., Di Serio, M., Sorrentino, A., Solinas, V. and Santacesaria, E. (1998a).
Preparation and properties of new acid catalysts obtained by grafting alkoxides
and derivatives on the most common supports. Note I - grafting aluminium and
zirconium alkoxides and related sulphates on silica. Appl. Catal. A: Gen., 167,
85-101.
Iengo, P., Di Serio, M., Solinas, V., Gazzoli, D., Salvio, G. and Santacesaria, E.
(1998b). Preparation and properties of new acid catalysts obtained by grafting
alkoxides and derivatives on the most common supports. Part II: Grafting
zirconium and silicon alkoxides on -alumina. Appl. Catal. A: Gen., 170, 225-
244.
JCPDS (1983). Mineral Powder Diffraction File. JCPDS International Centre for
Diffraction Data, Swarthmore.
Kapteijn, F., Moulijn, J.A. and Tarfaoui, A. (1999). Catalyst characterization and
mimicking pretreatment procedures by temperature-programmed techniques, in:
van Santen, R.A., van Leeuwen, P.W.N.M., Moulijn, J.A. & Averill, B.A.
(Eds.), Catalysis: An Integrated Approach. Elsevier, Amsterdam, pp. 525-541.
Laosiripojana, N. and Assabumrungrat, S. (2005). Methane steam reforming over
Ni/Ce-ZrO
2
catalyst: Influences of Ce-ZrO
2
support on reactivity, resistance
toward carbon formation, and intrinsic reaction kinetics. Appl. Catal. A: Gen.,
290, 200-211.
Langmuir, I. (1916). The constitution of fundamental properties of solids and liquids:
Part I. Solids. J. Am. Chem. Soc., 38, 2221-2295.
Lee, S.Y. and Aris, R. (1983). Theoretical and catalytic aspects of catalyst
impregnation, in: Poncelet, G., Grange, P. & Jacobs, P.A. (Eds.), Preparation of
Catalysts III. Elsevier, Amsterdam-Oxford-New York.
108
Chapter 3: Experimental
Liebhiafsky, H.A., Peiffer, H.G., Winslow, E.H., and Zemany, P.D. (1972). X-ray
Electrons and Analytical Chemistry. Wiley-Interscience, John Wiley & Sons
Inc., New York, vol. 6, pp. 236.
Mattos, L.V., Rodino, E., Resasco, D.E., Possos, F.B. and Noronha, F.B. (2003). Partial
oxidation and CO
2
reforming of methane on Pt/Al
2
O
3
, Pt/ZrO
2
, and Pt/Ce-ZrO
2
catalysts. Fuel. Proc. Technol., 83, 147-161.
Opoku-Gyamfi, K., Tafreshi, Z.M. and Adesina, A.A. (1998). Activities of o-Al
2
O
3
-
supported bimetallic Pt-Ni and Co-Ni catalysts for CH
4
oxidation. React. Kinet.
Catal. Lett., 64, 229-238.
Opoku-Gyamfi, K. (1999). A novel thermally self-sustaining fluidised bed reactor for
hydrogen production from methane. Ph.D. thesis, University of New South
Wales, Sydney, Australia.
Roh, H.S., Jun, K.W. and Park, S.E. (2003). Methane-reforming reactions over Ni/Ce-
ZrO
2
/-Al
2
O
3
catalysts. Appl. Catal. A: Gen., 251, 275-283.
Rostrup-Nielsen, J.R. and Bak-Hansen, J.-H. (1993). CO
2
-reforming of methane over
transition metals. J. Catal., 144, 38-49.
Yaluris, G., Larson, R.B., Kobe, J.M., Gonzalez, M.R., Fogash, K.B. and Dumestic,
J.A. (1996). Selective poisoning and deactivation of acid sites on sulfated
zirconia catalysts for n-butane isomerization. J. Catal., 158, 336-342.
109
Chapter 4: Preliminary Work and System Characterisation
C
C
h
h
a
a
p
p
t
t
e
e
r
r
4
4
P
P
R
R
E
E
L
L
I
I
M
M
I
I
N
N
A
A
R
R
Y
Y
W
W
O
O
R
R
K
K
A
A
N
N
D
D
S
S
Y
Y
S
S
T
T
E
E
M
M
C
C
H
H
A
A
R
R
A
A
C
C
T
T
E
E
R
R
I
I
S
S
A
A
T
T
I
I
O
O
N
N
4.1 General considerations
In view of further justifying the main objective of this investigation in examining
deactivation during steam reforming, in this chapter we performed thermodynamic
assessments of propane steam reforming and associated reactions at various reaction
conditions. Additionally, thermodynamic calculations were carried out to predict
equilibrium composition which can then be used as an indicator when selecting
operating variables required for fulfilling the goal of this study.
While most steam reforming studies utilise methane as the hydrocarbon substrate,
methane steam reforming however requires high temperature at a level which favours
sintering. In this investigation, propane was therefore selected where reforming may be
carried out at lower temperatures, thus ensuring carbon deposition as the primary source
of deactivation.
Moreover, it is also important to ensure that kinetic parameters evaluated from
experimental data are not masked by heat and mass transfer resistances [Froment and
110
Chapter 4: Preliminary Work and System Characterisation
Bischoff, 1980]. Thus, prior to proper kinetic investigation, preliminary analyses were
carried out to ensure selection of appropriate operating conditions which would
preclude or minimise transport interference.
This chapter presents a range of preliminary studies undertaken to provide a
thermodynamic scrutiny on the effect of varying reaction parameters and to characterise
the experimental system. The system characterisation procedures include, blank runs to
isolate wall or support effects, calibration of gas chromatograph, determination of
operating minimum fluidisation velocity, evaluation of both heat and mass transport
effects and treatment of experimental raw data.
4.2 Thermodynamics
The thermodynamic feasibility of a reaction may be assessed from the value of standard
free energy change, AG, where AG < 0 indicates that the reaction is favourable at the
given condition. Sandler [1999] created a computer program which can be utilised to
provide an accurate value of AG. The resulting AG estimates for reactions (2.5)-(2.8)
may be adequately correlated via a linear expression with temperature as shown in
Table 4.1. This is also graphically illustrated in Fig. 4.1.
In light of the standard free energy changes, AG, limiting operating temperature for each
reaction was evaluated as hosted in Table 4.2.
111
Chapter 4: Preliminary Work and System Characterisation
Table 4.1. Thermodynamic characteristics of propane steam
reforming and related reactions.
Reaction
) (T G A
Propane steam reforming
2 2 8 3
7H 3CO O 3H H C + = +
AG
SR
(T) = 529,867 744T
Propane dehydrogenation
2 4 8 3
2H 2C(s) CH H C + + =
AG
DHY
(T) = 38,858 210T
Boudouard
2
CO C 2CO + = AG
BOU
(T) = -172,028 + 177T
Water-gas shift
2 2 2
H CO O H CO + = + AG
WGS
(T) = -36,974 + 34T
-200
-100
0
100
200
300
400 600 800 1000 1200 1400
Temperature (K)
A
G

(
k
J

m
o
l
-
1
)
Propane steam reforming
Propane dehydrogenation
Boudouard
Water-gas shift
Fig. 4.1. Thermodynamic characteristics of propane steam reforming and related
reactions.
112
Chapter 4: Preliminary Work and System Characterisation
Table 4.2. Limiting temperatures for propane steam reforming and related
reactions.
Reaction Temperature (K) Upper/lower Limit
Propane steam reforming
2 2 8 3
7H 3CO O 3H H C + = +
712 Lower
Propane dehydrogenation
2 4 8 3
2H 2C(s) CH H C + + =
185 Lower
Boudouard
2
CO C 2CO + = 972 Upper
Water-gas shift
2 2 2
H CO O H CO + = + 1087 Upper
Based on the thermodynamic features in Fig. 4.1 and Table 4.2, for propane steam
reforming reaction to be feasible (i.e. AG < 0) the minimum operating temperature must
be at least 712 K. Its operation therefore will always be accompanied by coke formation
via dehydrogenation which has a lower limiting temperature of 185 K. Higher operating
temperature will indeed promote both steam reforming and dehydrogenation reactions.
On the other hand, carbon may also be produced at low temperature via CO
disproportionation (Boudouard) with feasibility below 972 K.
Subsequently, equilibrium conversion at various conditions may be evaluated by first
determining equilibrium constants, K, which is defined as
113
Chapter 4: Preliminary Work and System Characterisation
RT
T G
e K
) ( A
=
(4.1)
Estimates of the K values can also be precisely calculated using the computer program
available in Sandler [1999]. The K values for reactions (2.5)-(2.8) are provided in Table
4.3.
Table 4.3. Equilibrium constants of propane steam reforming and related
reactions.
Reaction Temperature
(K) (2.5) (2.6) (2.7) (2.8)
500 3.70110
-17
8.32410
6
5.97110
8
1.33610
2
600 4.31710
-8
3.67710
7
5.64810
5
2.72910
1
700 1.53010
-1
1.10810
8
3.93010
3
9.00810
0
800 1.37510
04
2.58510
8
9.57110
1
4.01110
0
900 1.03410
08
5.04010
8
5.38410
0
2.17610
0
1000 1.34310
11
8.62410
8
5.45010
-1
1.35410
0
1100 4.80610
13
1.33910
9
8.45910
-2
9.28810
-1
1200 6.48610
15
1.93210
9
1.81010
-2
6.85110
-1
1300 4.11310
17
2.63410
9
4.95510
-3
5.33710
-1
1400 1.43610
19
3.43410
9
1.64710
-3
4.33510
-1
1500 3.10210
20
4.32010
9
6.39310
-4
3.64010
-1
As the presence of multiple reactions often adds a substantial complexity in determining
composition of an equilibrium mixture, in such a system a more direct procedure based
on minimisation of the total Gibbs free energy has been preferred [Perry and Green,
1997]. This approach predicts equilibrium composition for given T and P and for a
given initial feed where the required input is the choice of a set of species (refer to
114
Chapter 4: Preliminary Work and System Characterisation
Appendix B for further details). For propane steam reforming conducted in the range
600-900 K and at steam-to-carbon (S:C) feed ratio between 0.5-5.0, the observed
species in the product gas are H
2
, CO, CO
2
, CH
4
, H
2
O and C
3
H
8
. Under these
conditions, the presence of any alkenes and alkynes is essentially negligible even at S:C
ratio as low as 0.5.
Fig. 4.2 shows the equilibrium composition during propane steam reforming at 800 K
and S:C = 1 at varying pressure between 0.1-5 bar. Higher hydrocarbon has a greater
tendency to participate in steam reforming compared to its methane counterpart. Thus, it
is not surprising that propane is completely reacted at all pressure ranges at 800 K and
S:C = 1. However, the reforming products (CO and H
2
) decrease in concentration at
elevated pressure, whilst H
2
O become increasingly unused. Therefore, it is evident that
propane conversion via steam reforming is indeed favoured by lower pressure. The
water-gas shift reaction does not appear to have a significant effect; as evident from
stoichiometry, the numbers of moles produced and reacted are equal. Consequently, it
seems that the balance between CO and CO
2
in the system was dictated by the
Boudouard reaction in which high pressure drives the transformation of CO to CO
2
.
115
Chapter 4: Preliminary Work and System Characterisation
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 1 2 3 4 5
Pressure (atm)
E
q
u
i
l
i
b
r
i
u
m

c
o
m
p
o
s
i
t
i
o
n
C3H8
CH4
H2O
CO
CO2
H2
Fig. 4.2. Effect of pressure on equilibrium composition for propane steam
reforming. Reaction conditions: temperature, 800 K; S:C ratio, 1.
Fig. 4.3 depicts the effect of varying temperature (600-900 K) in propane steam
reforming under pressure and feed ratio at 1 bar and S:C = 1 respectively. Propane feed
is mostly consumed at 600 K, while above 700 K complete conversion can be achieved.
Thus, increasing temperature enhances steam reforming as evidenced by improved H
2
and CO yields. The fact that a significant amount of H
2
O - notably at lower temperature
range - has not been consumed, indicate that a large fraction of C
3
H
8
goes through
dehydrogenation which releases H
2
and CH
4
. The declining composition of CH
4
(a
product of dehydrogenation), however, suggests that steam reforming becomes
increasingly more competitive over dehydrogenation at high temperatures. Further, as
Boudouard and water-gas shift reactions (both produce CO
2
) are known to be
exothermic in nature, it is expected that CO
2
concentration would be higher at low-
temperature end.
116
Chapter 4: Preliminary Work and System Characterisation
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
600 650 700 750 800 850 900
Temperature (K)
E
q
u
i
l
i
b
r
i
u
m

c
o
m
p
o
s
i
t
i
o
n
C3H8
CH4
H2O
CO
CO2
H2
Fig. 4.3. Effect of temperature on equilibrium composition for propane steam
reforming. Reaction conditions: pressure, 1 bar; S:C ratio, 1.
The effect of S:C feed ratio on equilibrium composition is presented in Fig. 4.4. This
analysis was evaluated for propane steam reforming performed at 800 K and 1 bar.
Propane was fully used up for S:C > 1. Similarly, steam-rich environment also favours
steam reforming over dehydrogenation as the two reactions compete for propane
substrate; accordingly, CH
4
composition declines with increasing S:C ratio. Moreover,
it is interesting to note that the compositions of reforming products, H
2
and CO, are
practically constant beyond S:C > 2. The presence of excess H
2
O in addition favours
water-gas shift reaction, thus responsible for the disappearance of CO and generation of
CO
2
at elevated S:C ratio.
117
Chapter 4: Preliminary Work and System Characterisation
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 0.5 1 1.5 2 2.5 3
Steam-to-carbon feed ratio
E
q
u
i
l
i
b
r
i
u
m

c
o
m
p
o
s
i
t
i
o
n
C3H8
CH4
H2O
CO
CO2
H2
Fig. 4.4. Effect of steam-to-carbon ratio (S:C) on equilibrium composition for
propane steam reforming. Reaction conditions: temperature, 800 K; pressure, 1
bar.
4.3 Preliminary experiments
4.3.1 Blank runs
The endothermic nature of steam reforming leads to a necessity for high-temperature
operation. It was therefore necessary to assess the effect of -Al
2
O
3
support and reactor
materials on the reactions under the selected experimental conditions.
A blank run was conducted using 1 g of alumina powders (212-250 m) at 973 K and
S:C = 3. Reactor exit stream was regularly injected into the TCD-GC for composition
measurement. Under this condition, no product components were detected throughout
the 10-h run. In fact, the amount of propane exiting the reactor was essentially the same
as via the reactor by-pass line. Consequently, it was acceptable to assume that the
118
Chapter 4: Preliminary Work and System Characterisation
support and the reactor materials have negligible effect on steam reforming reactions
carried out under these operational conditions.
4.3.2 Product analysis
Products of propane steam reforming were quantified using a Shimadzu gas
chromatograph (model 8A) with an Alltech Haysep DB column. Under the reaction
conditions employed in this study (i.e. 773-873 K and 101.3 kPa), the final product
stream was composed mainly H
2
, CO, CH
4
, CO
2
and C
3
H
8
in He diluent gas, while H
2
O
was removed via passage through an ice-cooled condenser and a drierite packing.
Traces of C
2
H
2
, C
2
H
4
and C
2
H
6
were rarely observed. The GC operating conditions are
listed in Table 4.4.
4.3.3 GC calibration
GC calibration was performed for major components, i.e. He, H
2
, CO, CH
4
, CO
2
and
C
3
H
8
. Varying quantities of these gases were mixed with argon gas to yield mixtures at
a range of concentration. For each concentration, at least 3 sample injections were
measured where variation in peak areas must be less than 1%. As GC performance may
fluctuate with respect to system and ambient conditions, it is necessary to normalise the
measured peak-area against peak area of a standard gas (1% CH
4
in argon). The
normalised area, A
i,norm
, may thus be defined as,
ref
i
norm i
A
A
A =
, (4.2)
where A
i,norm
= normalised area of component i,
A
i
= absolute area of component i,
119
Chapter 4: Preliminary Work and System Characterisation
A
ref
= reference area of component CH
4
in standard gas mixture.
Subsequently, the mole fraction, X
i
, was evaluated from,
norm i i i
A k X
,
=
(4.3)
where k
i
is the response factor of component i.
Table 4.4. GC operating conditions.
Parameter Setting
Ar (carrier gas) 30 ml min
-1
Column temperature 393 K
Detector temperature 423 K
Injection temperature 423 K
Detector current 60 mA
Attenuation 1
Polarity -ve
An example of chromatogram output is shown in Fig. 4.5. Typical retention times and
response factors, k
i
, for the various components are hosted in Table 4.5.
120
Chapter 4: Preliminary Work and System Characterisation
Fig. 4.5. Typical chromatogram.
Table 4.5. Retention times and response factors for various components.
Component Retention time (min) Response factor (k
i
)
He 2.78 0.0046
H
2
3.00 0.0037
CO 3.49 0.0212
CH
4
4.48 0.0100
CO
2
6.35 0.0390
C
3
H
8
36.98 0.0162
4.3.4 Optimum fluidisation flow rate
When an increasing rate of fluid is passed upwards through a bed of particles, the
upward drag force increases until it eventually overcomes the apparent weight of
particles in the bed. At this point, the particles are being lifted by the fluid causing an
increase in particle separation and so, bed fluidisation occurs [Rhodes, 1998]. The
121
Chapter 4: Preliminary Work and System Characterisation
transition point at which the bed behaviour shifts from packed-bed to fluidised-bed
corresponds to minimum fluidisation velocity. While there is a need to maintain gas
flow above the minimum fluidisation velocity at all times, it is also necessary to avoid
excessively high fluid velocities which may be responsible for entrainment of bed
particles. Consequently, the total gas flow rate must be carefully determined to ensure
good gas-solid contact and to eliminate any transport effects in the reaction system. The
following section outlines the procedure for the establishment of the optimum
fluidisation regime.
4.3.4.1 Experimental
In this investigation, the fluidisation behaviour may be conveniently characterised by
considering variables such as gas velocity and particle size. Nevertheless, it is generally
accepted that particles in the range of 180-295 m are sufficiently small to avoid
internal mass transport effects [Opoku-Gyamfi, 1999]. Given that particle density,
s
, is
approximately 4000 kg m
-3
, the particles in this size range are categorised as Group B
powder according to Geldarts classification [Geldart, 1973]. This class of powder is
known to be relatively easy to fluidise without major risks, such as generation of cracks
and solid plugs, and formation of deep spouting beds. Therefore, the catalyst particles
prepared for the fluidised bed experiments were sieved to sizes within this range.
The minimum fluidisation velocity was experimentally determined with a bed loading
of 4 g (Al
2
O
3
powders, 212-250 m). To take into consideration the effect of gaseous
thermal expansion, reactor furnace was heated to a typical reaction temperature at 873
K. Subsequently, flow rate of fluidisation gas (argon) was increased gradually between
122
Chapter 4: Preliminary Work and System Characterisation
0-600 ml min
-1
, while the pressure drop across the reactor was being monitored by a
water manometer. The observed hydrodynamic profile of the fluidised bed as a function
of Reynolds number, Re, is presented in Fig. 4.6. It is evident that the best operating
fluidisation regime for the reactor under experimental conditions was obtained between
200-450 ml min
-1
(Re = 0.041-0.092) with incipient fluidisation velocity occurred at 150
ml min
-1
(Re = 0.031).
0
0.5
1
1.5
2
2.5
3
3.5
0.00 0.02 0.04 0.06 0.08 0.10 0.12
Re
d
P
/
L

(
m
m

H
2
O
/
c
m
)
Fig. 4.6. Hydrodynamic behaviour of fluidised bed during fluidisation.
4.3.4.2 Theoretical
The minimum fluidisation velocity, U
mf
, may be evaluated from a correlation by Wen
and Yu [1966] described as,
( ) | | 25 25 0651 0 25 25
5 0
2
. Ar . .
d
U
.
p g
mf
+ =

(4.4)
123
Chapter 4: Preliminary Work and System Characterisation
where Ar is the Archimedes number, given by:
( )
2
3


p g s g
gd
Ar

= (4.5)
where d
p
= average particle diameter (m),

g
= density of fluidising gas (kg m
-3
),

s
= particle density (kg m
-3
),
= viscosity of fluidising gas (Pa.s),
g = acceleration of gravity (m s
-2
).
Using parameters provided in Table 4.6, the minimum fluidisation velocity obtained
from Wen and Yu correlation (cf. Eq. 4.4) is 226.6 cm min
-1
(243.1 ml min
-1
). This is
considerably higher than the experimental value of 139.9 cm min
-1
(150 ml min
-1
). The
discrepancy was perhaps largely due to the various assumptions used in estimating
some parameters in the theoretical approach.
124
Chapter 4: Preliminary Work and System Characterisation
Table 4.6. Physical properties of catalyst and reaction system at 873 K.
Parameter Value Source

s
4000 kg m
-3
Satyamoorthy and Raja Rao [1992]

b
2000 kg m
-3
(1-c)
s

g
4.04 10
-1
kg m
-3
Perry and Green [1997]
d
p
2.31 10
-4
m experiment
T 873 K experimental condition
7.12 10
-5
Pa.s Perry and Green [1997]
g 9.8 m s
-2
constant
-AH
rxn(SR)
-498.6 kJ mol
-1
theoretical computation
C
vs
3.06 10
6
J m
-3
K
-1
Perry and Green [1997]
C
vg
2.19 10
2
J m
-3
K
-1
Perry and Green [1997]
C
pg
5.42 10
2
J m
-3
K
-1
Perry and Green [1997]

p
36 W m
-1
K
-1
Perry and Green [1997]

g
3.97 10
-2
W m
-1
K
-1
Perry and Green [1997]
D
AB
2.61 10
-5
m
2
s
-1
estimated for C
3
H
8
-Ar system
D
eff
2.61 10
-6
m
2
s
-1
0.1 D
AB
c 0.5 assumption
n 1 assumption
V 300 ml min
-1
(0.0159 m s
-1
) experimental condition
A 3.14 cm
2
reactor tube property
b
A
C
2.44 10
-3
kmol m
-3
theoretical computation
C
He
1.20-11.10 10
-3
kmol m
-3
experimental condition
U
mf
2.332 10
-2
m s
-1
experiment
P 101.32 kPa experimental condition
R
g
8.314 J mol
-1
K
-1
gas constant
125
Chapter 4: Preliminary Work and System Characterisation
4.4 Transport resistance considerations
In general, heterogenous catalytic reactions take place sequentially via the following
steps [Kraemers and Westerterp, 1963];
1. Mass transfer (diffusion) of the reactants from the bulk fluid to the external
surface of the catalyst,
2. Transport of reactants by diffusion process through the pores into the particle.
3. Adsorption of reactants on the internal catalyst surface,
4. Chemical reaction between the adsorbed reactants,
5. Desorption of the products from the catalyst surface,
6. Transport of products by diffusion process through the pores of the particle,
7. Mass transfer of products from the external surface to the bulk of the fluid.
Steps 3, 4 and 5 represent the true chemical kinetics. They are thus called reaction steps,
while steps 1, 2, 6 and 7 are known as the diffusion steps. When the diffusion steps are
faster than the reaction steps, the reaction steps are therefore the limiting stage and
overall reaction rate is representative of the true chemical reaction. On the other hand,
when the diffusion steps are slower compared to the reaction steps, the diffusions steps
are the limiting stage and hence, rate data are likely to be overshadowed by the effect of
mass transfer. By the same token, resistance to heat transfer from the bulk fluid to the
catalyst surface and from the entrance of the pores to the inner catalyst surface may
jeopardise kinetic rate measurements.
126
Chapter 4: Preliminary Work and System Characterisation
In a fluidised bed system, heat and mass transfer rates between gases and particles are
high due to enhanced hydrodynamic and transport properties [Kunii and Levenspiel,
1969]. Even so, the selected operating conditions were checked against relevant
diagnostic criteria to determine whether the observed reaction rate represent the true
kinetic behaviour.
For the sake of simplicity, the properties of argon were used for this analysis and the
system temperature was taken as 873 K. Typical kinetic data for propane steam
reforming, i.e. r
A
= 1.94 10
-6
mol g cat.
-1
s
-1
[Mailet et al., 1996] and E
A
= 65 kJ mol
-1
[Figueiredo, 1975] were considered for these calculations. All other parameters utilised
in the computation are listed in Table 4.6.
4.4.1 External mass transfer
External mass transfer exists due to resistance in the path between the bulk phase and
the catalyst surface (Step 1). Mears [1971] criterion requires that for a system where
external resistance may be neglected, it should satisfy,
b
A c
p b A
C k
n d r
< 3
(4.6)
where -r
A
= rate of reaction (kmol kg cat.
-1
s
-1
),

b
= bulk density of catalyst (kg m
-3
),
d
p
= particle diameter (m),
n = reaction order,
k
c
= mass transfer coefficient (m s
-1
),
127
Chapter 4: Preliminary Work and System Characterisation
b
A
C = bulk gas-phase concentration of component A (kmol m
-3
).
For a fluidised or fixed bed with Reynolds number, Re, between 0.01 to 15,000, mass
transfer coefficient, k
c
, may be estimated from [Dwivedi and Upadhyay, 1977]:
36 . 0 82 . 0
365 . 0 765 . 0
Re Re
J
D
+ = c (4.7)
where c = voidage,
J
D
= mass transfer factor whose correlation with k
c
is defined in Eq. 4.8.
2
3
( )
c
D
k Sc
J
U
= (4.8)
where U = superficial gas velocity (m s
-1
),
Sc = Schmidt number (dimensionless).
Substituting the physical properties of the catalyst and the reaction system as given in
Table 4.6, k
c
was evaluated as 0.226 m s
-1
(Appendix C.1). Subsequently, inserting k
c
into Eq. 4.6 yielded the term on the left-hand side (LHS) as 1.63 10
-3
, confirming that
external mass transfer resistance was practically absent under the adopted reaction
conditions. This conclusion still holds even in situations where reaction order, n, is
much larger than 1. It is known that n, in general, rarely exceeds > 3 in most reactions.
4.4.2 Pore diffusion limitation
The pore diffusion (internal) mass transfer corresponds to the resistance encountered
during transport of reactants within the catalyst structure between the pore mouth and
128
Chapter 4: Preliminary Work and System Characterisation
the adsorption sites (Step 2). The extent of the pore diffusion limitation may be
neglected provided that the following Weisz-Prater criterion [Fogler, 1999] is fulfilled:
s
A eff
p b A
C D
d r
2

< 4
(4.9)
where D
eff
= effective diffusivity (m
2
s
-1
),
s
A
C = concentration of component A on catalyst surface (kmol m
-3
).
Since the external mass transfer is negligible, the surface concentration, , is
therefore essentially the same as the bulk fluid gas concentration, . Consequently,
substitution of relevant values from Table 4.6 gave the LHS of Eq. 4.9 as 3.25 10
s
A
C
b
A
C
-2
,
suggesting that pore diffusional resistance is insignificant under the chosen reaction
conditions.
4.4.3. External heat transfer
The external heat transfer resistance resulted in a substantial difference between the gas
phase and the catalyst surface. As a consequence, measurement of rate data at a
particular temperature may be substantially different to what exists at the catalyst
surface. For a condition where interphase heat transfer inhibition does not affect the true
chemical kinetics, Mears [1971] proposed that:
g b
A p b A rxn
R hT
E d r H
2
) ( A
< 0.3
(4.10)
where -AH
rxn
= heat or reaction (J mol
-1
),
129
Chapter 4: Preliminary Work and System Characterisation
h = gas-solid heat transfer coefficient (W m
-2
K
-1
),
T
b
= bulk-phase temperature (K).
The heat transfer coefficient, h, was estimated from a gas fluidised-bed correlation
suggested by Sathiyamorthy and Raja Rao [1992], given as:
3 0
5 0
63 0 42 0
0046 0
.
.
vg
vs . .
max
Pr
C
C
Ar Fr . Nu
|
|
.
|

\
|
=

(4.11)
where Fr = Froude number (dimensionless),
Ar = Archimedes number (dimensionless),
Pr = Prandtl number (dimensionless),
C
vs
= solid volumetric specific heat capacity (J m
-3
K
-1
),
C
vg
= gas volumetric specific heat capacity (J m
-3
K
-1
),

g
= thermal conductivity of gas (W m
-1
K
-1
),
Nu
max
= Nusselt number whose relation with h
max
is correlated in Eq. 4.12.
max
g
p max
Nu
d h
=

(4.12)
Using the appropriate values from Table 4.6 into Eqs. 4.11 and 4.12, h
max
was evaluated
as 1,682.7 W m
-2
K
-1
(Appendix C.2). Subsequently, the LHS of Eq. 4.10 yields 2.72
10
-3
, signifying that the effect of external heat transfer may be neglected.
130
Chapter 4: Preliminary Work and System Characterisation
4.4.4 Intraparticle heat transfer
Anderson [1963] suggests that the following to be satisfied if intrapellet temperature
gradient is absent,
g s p
A p b A rxn
R T
E d r H
2
2
) (

A
< 3
(4.13)
where
p
= thermal conductivity of catalyst particle (W m
-1
K
-1
),
T
s
= catalyst surface temperature (K).
As previously established, the external fluid-catalyst temperature gradient is
insignificant and hence, surface temperature, T
s
, is essentially the same as bulk
temperature, T
b
. Inserting the relevant values, the LHS of Eq. 4.13 was evaluated as
2.94 10
-5
, indicating the absence of intraparticle heat transfer under the adopted
conditions.
4.5 Experimental data treatment
4.5.1 Conversion
Conversion, X, may be described as the ratio of C
3
H
8
reacted to that fed into the reactor.
It is mathematically defined as;
f f A
p p A f f A
F y
F y F y
X
,
, ,

=
(4.14)
where y
A,f
= mole fraction of C
3
H
8
in feed stream,
y
A,p
= mole fraction of C
3
H
8
in product stream,
131
Chapter 4: Preliminary Work and System Characterisation
F
f
= total molar flow rate in feed stream (mol s
-1
),
F
p
= total molar flow rate in product stream (mol s
-1
).
4.5.2 Reaction rate
In view of the isothermality across fluidised bed systems, the design equation for a well-
mixed reactor was used in the determination of reaction rate, -r
A
, thus [Levenspiel,
1999],
W
X F y
r
f f A
A
,
= (4.15)
where -r
A
= rate of C
3
H
8
conversion (mol g cat.
-1
s
-1
),
W = total weight of catalyst (g).
4.5.3 Selectivity
Assuming propane steam reforming process takes place in accompany of several side
reactions as described in Eqs. 2.5-2.8, the selectivities, S
i
, of major components may
therefore be defined by,
) ( 7
, ,
8 3 8 3
2
2
out H C in H C
H
H
F F
F
S

=
(4.16)
) ( 3
, ,
8 3 8 3
out H C in H C
CO
CO
F F
F
S

=
(4.17)
) ( 3
, ,
8 3 8 3
4
4
out H C in H C
CH
CH
F F
F
S

=
(4.18)
132
Chapter 4: Preliminary Work and System Characterisation
) ( 3
, ,
8 3 8 3
2
2
out H C in H C
CO
CO
F F
F
S

=
(4.19)
where F
i
= molar flow rate of species i (mol s
-1
).
4.6 Range of experimental variables
Industrial steam reforming is typically performed using excess steam-to-carbon ratio
(S:C > 3) to effectively minimise carbon lay-down. In a substantial part of this project,
anti-coking attributes of various prepared catalysts were compared and deactivation
kinetic parameters were collected under conditions favouring carbon deposition. In this
case, low steam-to-carbon ratio between 0.8-1.6 was selected in order to deliberately
induce coking. As a result, change in catalytic behaviour due to carbon-induced
deactivation may be better observed. On the other hand, reaction studies at constant
activity level were also required in the steady-state kinetic analyses, whence steam-to-
carbon ratio employed was in the range 2.4-4.8 to ensure minimum carbon lay-down.
As propane, rather than methane, was used as hydrocarbon feed in this investigation,
reforming reaction may therefore take place at lower temperatures. Thermodynamically,
steam reforming of propane would proceed at temperature above 718 K. On the other
hand, it was also necessary to establish an upper temperature limit as sintering of
alumina takes place at > 1100 K [Moulijn et al., 2001]. Additionally, in order to
minimise deactivation via thermal sintering, temperature needs to be kept reasonably
low. This would ensure that only carbon deposition remains the significant source of
catalyst deactivation. In view of these competing constraints, the experimental
temperature range was selected as 773-873 K.
133
Chapter 4: Preliminary Work and System Characterisation
References
Anderson, J.B. (1963). A criterion for isothermal behaviour of catalyst pellet. Chem.
Eng. Sci., 18, 147-148.
Dwivedi, P.N. and Upadhyay, S.N. (1977). Particle-fluid mass transfer in fixed and
fluidized beds. Ind. Eng. Proc. Des. Dev., 16, 157-165.
Figueiredo, J.L. (1975). Carbon formation on steam reforming catalysts. Ph.D. thesis,
Imperial College, University of London, London.
Fogler, H.S. (1999). Elements of Chemical Reaction Engineering. Prentice Hall
International, Inc., New Jersey.
Froment, G.F. and Bischoff, K.B. (1979). Chemical Reactor Analysis and Design. John
Wiley, New York.
Geldart, D. (1973). Types of gas fluidization. Powder Technol., 7, 285-292.
Kraemers, H.A and Westerterp, K.R. (1963). Elements of Chemical Reactor Design and
Operation. Chapman & Hall, London.
Kunii, D. and Levenspiel, O. (1969). Fluidization Engineering. John Wiley & Sons,
New York.
Levenspiel, O. (1999). Chemical Reaction Engineering, 3
rd
Edn. Wiley, New York.
Mailet, T., Barbier Jr., J. and Duprez, D. (1996). Reactivity of steam in exhaust gas
catalysis: III. Steam and oxygen/steam conversions of propane on a Pd/Al
2
O
3
catalyst. Appl. Cat. B: Environ., 9, 251-266.
Mears, D.E. (1971). Test for transport limitations in experimental reactors. Ind. Eng.
Chem. Proc. Des. Dev., 10, 541-547.
Moulijn, J.A., van Diepen, A.E. and Kapteijn, F. (2001). Catalyst deactivation: is it
predictable? What to do? Appl. Catal. A: Gen., 212, 3-16.
Opoku-Gyamfi, K. (1999). A novel thermally self-sustaining fluidised bed reactor for
hydrogen production from methane. Ph.D. thesis, University of New South
Wales, Sydney, Australia.
134
Chapter 4: Preliminary Work and System Characterisation
Perry, R.H. and Green, D.W. (1997). Perrys Chemical Engineers Handbook, 7
th
Edn.
McGraw Hill, Inc., New York.
Rhodes, M. (1998). Introduction to Particle Technology. John Wiley & Sons, England.
Sandler, S.I. (1999). Chemical and Engineering Thermodynamics. John Wiley & Sons,
New York.
Satyamoorthy, D. and Raja Rao, M. (1992). Prediction of maximum heat transfer
coefficient in gas fluidized bed. Int. J. Heat Mass Transfer, 35, 1027-1034.
Wen, C.Y. and Yu, Y.H. (1966). A Generalized Method for Predicting Minimum
Fluidization Velocity. AIChE. J., 12, 610-612.
135
Chapter 5: Bimetallic Catalyst Design
C
C
h
h
a
a
p
p
t
t
e
e
r
r
5
5
B
B
I
I
M
M
E
E
T
T
A
A
L
L
L
L
I
I
C
C
C
C
A
A
T
T
A
A
L
L
Y
Y
S
S
T
T
D
D
E
E
S
S
I
I
G
G
N
N
5.1 General considerations
The ultimate performance of steam reforming supported metal catalysts is determined
by the interaction of various preparation and activation procedures. The nature of metal
adsorption profile on the support is dependent on the preparation conditions, while the
textural and metal crystallite properties are largely influenced by the activation strategy.
The key preparation and activation variables adjudged most important from the
literatures [Heise and Schwarz, 1985; Mile et al., 1990; Che and Bennett, 1989] were
examined following a two-level factorial design. The experimental variables selected
were; pH-level (preparation), and calcination temperature, heating rate and duration
(activation).
This section presents a comprehensive study on the design of a bimetallic Co-Ni
catalyst. It involves theoretical discussions, characterisation and activity tests on the
various stages of a catalyst life-cycle, from preparation, activation, reaction to post-
mortem analysis. The relationships between preparation conditions and
physicochemical attributes, steam reforming activity and resilience to deactivation of
the finished catalysts were investigated. Furthermore, possible synergistic effects in
136
Chapter 5: Bimetallic Catalyst Design
bimetallic system were identified via comparison with monometallic cobalt and nickel
catalysts.
5.2 Catalyst preparation
As described previously, the catalysts employed throughout this study were prepared by
impregnation. In this technique, a solution containing pre-determined amount of metal
salts was initially added to the support. Within the solution, metal salts dissociate into
metal and nitrate ions; whereupon, these ions migrate into the support to find suitable
adsorption sites. Evidently, the ability of ion to move about in an aqueous system is
controlled by solution acidity which is in turn a critical factor in determining the
quantity of metal adsorbed and depth of penetration into the support pores [Heise and
Schwarz, 1985]. The effect of solution pH on the charge on alumina surface may be
assessed via electrophoretic mobility data. For -alumina suspensions, a typical plot of
surface potential versus initial pH of the solution is shown in Fig. 5.1. At low-pH end,
the surface exhibits a strong positive charge, however as pH increases it slowly loses its
positively-charged potential followed by a rapid -potential drop at pH > 7.
Subsequently, the surface potential becomes predominantly neutral as it reaches a
transition point of zero charge (often referred to as the isoelectric point) which takes
place at pH 8.2 for -alumina. Following further decrease in acidity, the potential
becomes increasingly negatively-charged until pH reaches about 11.
137
Chapter 5: Bimetallic Catalyst Design
Fig. 5.1. potential as a function of the pH of the solution [Heise and Schwarz,
1985].
The influence of pH may be reflected in different types of adsorption profile at the
catalyst pellets [Geus and van Veen, 1999]. When positively-charged metallic (Co
2+
or
Ni
2+
) ions approach catalyst particle with a predominantly negatively-charged surface,
these ions would be quickly adsorbed on the surface leading to metal deposition
primarily on the particle exterior, thus it is often referred as an egg-shell profile. Since
metal sites can be easily accessed with reactants, this type of distribution may find its
uses in systems under limitation of severe diffusion (e.g. industrial reactors). In contrast,
catalyst with a positive-charge surface allows Co
2+
or Ni
2+
to travel further into the
138
Chapter 5: Bimetallic Catalyst Design
interior giving rise to homogenous metallic deposition. As a result, an egg-yolk profile
with a feature of uniform metal distribution is obtained, where it is often preferred for
kinetically-controlled applications [Adesina, 1996].
Olsbye et al. [1997] suggests that 2 different reaction mechanisms take place at the
solid-liquid interface under impregnation at low- and high-pH levels. These reactions
may cause pH change during impregnation as described by:
(i) protonation of the surface
+ +
+
2
OH Al H OH Al at low pH (5.1)
O H O Al OH OH Al
2
+ +

at high pH (5.2)
(ii) dissolution of alumina
O 3H 2Al 6H O Al
2
3
3 2
+ +
+ +
at low pH (5.3)

+ +
4 2 3 2
2Al(OH) O 3H 2OH O Al at high pH (5.4)
Based on the foregoing discussion, acidity level of impregnating solution obviously
plays a critical role in controlling the metal distribution profile of a steam reforming
catalyst. In view of exploring this key effect, two levels of impregnation pH were
utilised during the preparation stage: 2 (low) and 8 (high). The pH-level was capped
below 10 to prevent rapid causticisation of the alumina support. The solution acidity
throughout impregnation was controlled by addition of HNO
3
and NH
4
OH drops for pH
139
Chapter 5: Bimetallic Catalyst Design
2 and 8 respectively. During stirring under low pH, the slurry revealed a pink-coloured
solution. In comparison, light-blue mixture was observed for the high-pH impregnation.
Meanwhile, the impregnating solution of the monometallic Co and Ni catalysts at pH 2
appeared dark-pink and green respectively. For identification, the catalyst samples
prepared under pH 2 and 8 were designated with the letters L and H correspondingly.
5.3 Catalyst activation
Following adsorption of metallic ions on the support surface, the catalyst must be
activated in two-stage process, namely calcination and reduction. During calcination,
the adsorbed metallic ions are transformed to surface oxides. In the subsequent
reduction, the oxides formed are converted to metallic species possessing active sites for
reaction. Thermal gravimetric experiments can be implemented to assist in selecting the
range of operating variables appropriate for calcination [Marturano et al., 1999a] and
reduction [Marturano et al., 1999b].
5.3.1 Calcination
The calcination conditions determine the types of metallic phases and alloys formed on
the support and the nature of bonding established between the precursor oxide and the
support [Mile et al., 1990; Che and Bennett, 1989]. Besides, calcination also leads to the
decomposition of impregnated salts, removal of impurities and sintering of the formed
oxide compounds [Che and Bennett, 1989]. The decomposition of the nitrate precursors
was observed by the emission of brown gas accompanied with strong acidic smell via
the furnace exhaust in the range 473-673 K during ramping.
140
Chapter 5: Bimetallic Catalyst Design
In order to examine the possible intermetallic oxide formation arising from calcination,
three controlling variables (i.e. duration, ramping rate and temperature) were selected. A
preliminary calcination run carried out in a TGA instrument was initially performed
using uncalcined low-pH specimen (where active components are deposited further
from the catalyst exterior compared to the high-pH). This experiment was performed
under the supposedly least-favoured condition for facilitating optimum nitrate
decomposition (i.e. fast heating rate - 20 K min
-1
and low temperature - 873 K). The
temperature was held constant at 873 K to allow extended monitoring of sample weight.
The weight change profile is presented in Fig. 5.2.
65
70
75
80
85
90
95
100
105
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
Time (h)
W
e
i
g
h
t

(
m
g
)
273
373
473
573
673
773
873
973
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 5.2. Transient weight profile during calcination to 873 K (with heating rate of
20 K min
-1
) for catalyst impregnated at pH 2.
Even so, under the adopted procedure, constant weight was observed after holding at
873 K for less than 0.25 h indicating removal of all nitrate precursors. This is also
141
Chapter 5: Bimetallic Catalyst Design
substantiated by a total weight drop of 29.63% in comparison to the theoretical 25.53%
predicted for complete nitrate decomposition (refer to Appendix D.1). The slightly
higher actual % weight drop was attributed to the removal of physisorbed and interstitial
moisture. Therefore, the temperature of 873 K would be sufficient as the lower limit for
calcination, since it would also ensure phase stability during the subsequent propane
steam reforming tests performed in the range 773-873 K. The upper limit was selected
as 973 K to minimise significant loss in support area due to thermal phase changes in
the support. Similarly, the faster heating rate of 20 K min
-1
may then be used for
examining the high formation rate of aluminate species. Although regarded as inactive
for reaction, metal-support interactions exhibit important features as a highly stable,
non-interacting support for metallic crystallites [Bartholomew, 1975; Bartholomew and
Farrauto, 1976]. In contrast, the establishment of slower heating rate at 5 K min
-1
would
permit the assessment of solid phase transformation reaction under possible internal
restructuring [Hardiman et al., 2004]. The lower limit of 1 h for duration was
implemented to ensure complete nitrate decomposition. On the other hand, to minimise
sintering due to prolonged heating, the thermal treatment was restricted to a minimum
period of 5 h.
Subsequently, the effect of various catalyst preparation and activation factors (involving
impregnating pH level, calcination time, heating rate and temperature) were investigated
based on the 2
4
-factorial design strategy as summarised in Table 5.1, where the
designation for the two sets of bimetallic catalysts is also provided. This two-level
factorial technique (also known as Yates statistical analysis) can uncover major trends
with a minimum number of runs in the factor space which can also provide valuable
142
Chapter 5: Bimetallic Catalyst Design
information and promising directions for future efforts in catalyst optimisation [Box et
al., 1978]. Background of the Yates algorithm is well covered in Montgomery [1997].
Table 5.1. 2
4
-factorial design plan for bimetallic catalyst synthesis.
Sample code
Low pH (2) High pH (8)
Holding time
(h)
Heating rate
(K min
-1
)
Temperature
(K)
L1 H1 1 5 873
L2 H2 5 5 873
L3 H3 1 20 873
L4 H4 5 20 873
L5 H5 1 5 973
L6 H6 5 5 973
L7 H7 1 20 973
L8 H8 5 20 973
Apart from assessing the preparative variables for bimetallic systems, the interplay
between individual Co and Ni metals were investigated using monometallic catalysts.
Monometallic Co and Ni catalysts were prepared with equivalent metal loading using
low-impregnation pH (to yield more homogenous distribution). For calcination process,
a program with longer duration, slower heating rate and higher temperature (as in L6)
was employed. This ensures optimum thermal treatment yielding distinct characteristics
and it would therefore allow better comparison between mono- and bi-metallic catalysts.
The calcination condition and catalyst designation for monometallic samples are
summarised in Table 5.2.
143
Chapter 5: Bimetallic Catalyst Design
Table 5.2. Preparation condition for monometallic catalysts.
Sample pH level Holding time (h) Heating rate (K min
-1
) Temperature (K)
L6Co 2 5 5 973
L6Ni 2 5 5 973
After calcination, all bimetallic specimens and the monometallic cobalt catalyst showed
colour change to black, however the monometallic Ni-catalyst appeared green.
5.3.2 Reduction
It is generally known that nickel-based steam reforming catalyst is more active under its
reduced state [Rostrup-Nielsen, 1984; Bartholomew and Farrauto, 1976]. As a result,
catalyst specimen must be reduced under hydrogen flow prior to steam reforming
experiment, where metal oxides, alloys and some aluminates are transformed to metallic
(Co
0
and Ni
0
) species.
Based on previous experience in our laboratory, complete reduction of 2 g of bimetallic
Co-Ni was achieved via H
2
treatment at 200 ml min
-1
for 4 h [Opoku-Gyamfi, 1999].
However, as smaller catalyst loading (1 g) was used for the current study, the H
2
-
reduction process was therefore carried out with a shorter duration (2 h), while other
variables were unchanged from previous study. This was done primarily to minimise the
possibility of sintering, as well as to reduce experimental time. Additionally, this
reduction scheme was subjected to further assessment via a TGA weight analysis. It is
expected that higher acidity impregnation, faster calcination heating rate, higher
144
Chapter 5: Bimetallic Catalyst Design
calcination temperature and longer calcination time would favour the formation of
aluminate compounds making it more difficult for the catalyst to be reduced [Mile et al.,
1990; Bartholomew and Farrauto, 1976; Li and Chen, 1995; Molina and Poncelet, 1998;
Hardiman et al., 2004 & 2005a]. Therefore, the calcined Sample H8 was used in the
reducibility study under the adopted procedure where the corresponding weight profile
is depicted in Fig. 5.3. As shown, catalyst weight stabilised after about 40 mins of
reductive treatment at 973 K. This suggests that the reduction duration of 2 h would be
sufficient to facilitate complete phase transformation.
70
75
80
85
90
95
100
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (h)
W
e
i
g
h
t

(
m
g
)
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Fig. 5.3. Transient weight profile during reduction of calcined Sample H8.
5.4 Catalyst characterisation
Principally, characterisation of industrial catalysts is carried out to evaluate intrinsic
properties that allow discrimination of efficient from less efficient catalysts and to relate
these properties with catalytic performance, hence it provides a direction for the
145
Chapter 5: Bimetallic Catalyst Design
production of an active, selective, stable and mechanically robust catalyst
[Niemantsverdriet, 1999].
In this thesis, catalyst characterisation has been grouped into three sections in view of
investigating physicochemical attributes, morphology and transient weight responses
under temperature programming.
5.4.1 Physicochemical attributes
Physicochemical properties involve the measurement of physical (i.e. support area,
metallic dispersion, surface area and active particle size) and chemical (i.e. acid site
strength and concentration) identities.
5.4.1.1 BET surface area
Table 5.3 displays BET areas for all 16 bimetallic catalysts and the corresponding
Yates statistical analysis for evaluating the magnitude and direction associated with the
influence of preparation variables and their interactions. The significance of each effect
was statistically discriminated based on calculated F-values. Accordingly, a particular
effect is deemed statistically significant when the relevant calculated F-value is greater
than the standard value obtained from F-table (F
8,8
= 3.44 at 95% confidence level [Box
et al., 1978]). The statistically important factors are shown in bold in Table 5.3. Further,
the sign of the variance value gives an indication on the direction that an effect has on
the response (dependent) variable.
146
Chapter 5: Bimetallic Catalyst Design
As quick inspection of data in Table 5.3 revealed, impregnating pH exhibited very
strong effect on surface area. Lower BET values in pH 2 catalysts were attributed to the
alumina dissolution reaction occurring under highly acidic environment (as denoted in
Eq. 5.3). On the other hand, the basic impregnation carried out at pH 8 is closer to the
isoelectric point (IEP) of alumina of 8.2 where nearly zero-charge potential existed on
the surface (cf. Fig. 5.1). The small difference in the pH level between the alumina
surface and the surrounding solution resulted in only minor charge transfer between
these two media and hence, reducing the potential for alumina causticisation (via Eq.
5.4). Indeed, the high-pH catalysts show comparable BET area to that of pure -Al
2
O
3
reported as 181 m
2
g
-1
[Opoku-Gyamfi, 1999].
Besides pH, other statistically relevant effects are calcination temperature, T, heating
rate, r, time, t, 2-factor interactions - Th, rh, rT, tr & tT, 3-factor interactions - trh &
rTh, and 4-factor interaction - trTh. The variances of calcination temperature, heating
rate and time show negative signs implying that higher surface area would be achieved
with a decrease in each of these variables. Interestingly, it was observed that the 2-factor
interaction, Th, possesses a positive impact revealing a mutual improvement arising
from parallel interplay between these two factors, where (+) = (+)(+) or (-)(-).
Evidently, as proposed before, low temperature and high pH level are beneficial in
attaining superior BET area. In a bigger picture, this proposition also implies that the
effect of primary factors cannot be analysed without considering the associated multiple
interactions between individual variables under study [Eisenacher and Adesina, 2000].
147
Chapter 5: Bimetallic Catalyst Design
The BET areas for the monometallic Co- (Sample L6Co) and Ni- (Sample L6Ni)
catalysts are summarised in Table 5.4. The BET values of 134.68 and 137.11 m
2
g
-1
for
cobalt and nickel catalysts respectively are higher than their counterpart bimetallic Co-
Ni catalyst (Sample L6) with 108.62 m
2
g
-1
prepared using the same procedure. The
formation of non-porous alloy (e.g. NiCo
2
O
4
) as a result of interaction between different
metallic species - as commonly occurs in multimetallic systems - is primarily
responsible for the reduced surface areas. Therefore, it appears that the variation in BET
area among bimetallic specimens under different calcination condition may be attributed
to disparity in the resulting metal oxide particle distribution, dispersion and
reagglomeration rates, while structural changes would not be prominent during
calcination as all supports underwent pre-treatment at 1073 K [Hardiman et al., 2004].
148
Chapter 5: Bimetallic Catalyst Design
Table 5.3. BET areas of bimetallic catalysts and corresponding Yates analysis.
Sample S
BET
(m
2
g
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 120.40 16 137.40 Average
L2 118.07 8 -1.91 t 99.07
L3 119.82 8 -2.76 r 142.97
L4 117.70 8 -0.60 tr 31.11
L5 109.47 8 -6.37 T 329.78
L6 108.62 8 0.32 tT 16.73
L7 108.79 8 -0.67 rT 34.94
L8 106.56 8 -0.06 trT 3.24
H1 164.56 8 47.45 h 2456.87
H2 163.49 8 -0.03 th 1.59
H3 162.04 8 -1.84 rh 95.21
H4 158.61 8 -0.31 trh 15.97
H5 163.33 8 4.27 Th 221.03
H6 162.33 8 -0.02 tTh 1.00
H7 158.46 8 -0.23 rTh 11.77
H8 156.19 8 0.33 trTh 17.34
Table 5.4. BET areas of monometallic catalysts.
Sample BET area, S
BET
(m
2
g
-1
)
L6Co 134.68
L6Ni 137.11
149
Chapter 5: Bimetallic Catalyst Design
5.4.1.2 Chemisorptive properties
Chemisorptive properties, namely metal dispersion, PD, metal surface area, S
m
, and
active particle size, APS, were estimated from H
2
-chemisorption experiments. The metal
dispersion of bimetallic catalysts and the associated Yates analysis are provided in
Table 5.5.
Clearly, the calcination time, t, temperature, T, heating rate, r, impregnating pH, h, 2-
factor interactions - tT, th, rh & rT, 3-factor interactions - trh & trT, and 4-factor
interaction - trTh are statistically important at 95% confidence interval in determining
metal dispersion, PD. The variance analyses, in addition, suggest that single variables (t,
T, r and h) brought about the strongest impact; however, only increase in temperature
would improve dispersion while longer calcination duration, faster heating rate and
higher pH impregnating solution yielded poorer dispersion. Indeed, Chen et al. [1991]
and Mieth and Schwarz [1989] pointed out that acidic environment in impregnation
facilitates mobility and penetration of metal nitrate ions into porous alumina giving rise
to superior dispersion. Additionally, the negatively-signed 2-factor interaction, tT, (the
most prominent among other interaction factors) further underlines that mutual
enhancement in dispersion can be obtained by lowering one variable (time) and also
increasing the other (temperature), i.e. (-) = (-)(+).
Table 5.6 displays the metal dispersion for Co- and Ni- monometallic catalysts with
5.10 and 6.79% respectively. Although the dispersion of bimetallic catalyst L6 (5.62%)
is in between these two values, it is not a simple linear combination of the individual
150
Chapter 5: Bimetallic Catalyst Design
components. Hence, it appears that chemical or textural synergism played a substantial
role in the bimetallic system.
Table 5.7 hosts the metal surface areas and relevant Yates analysis. Like dispersion, the
significant variables in controlling metal surface area, S
m
, (at 95% confidence level) are
calcination time, t, temperature, T, heating rate, r, impregnating pH, h, 2-factor
interactions - tT, th, rh & rT, 3-factor interactions - trh & trT, and 4-factor interaction -
trTh. Furthermore, as suggested by variance estimates, metal surface area can be
enhanced by using lower pH during impregnation and adopting shorter time, higher
temperature and slower heating rate in the subsequent calcination.
As shown in Table 5.8, the small variation in metal surface area between the L6Co and
L6Ni (7.60 and 8.79 m
2
g
-1
respectively) catalysts and the L6 (7.52 m
2
g
-1
) catalyst was
probably a reflection of the identical metal loading and thermal conditioning scheme
which was employed for all three samples [Sinfelt, 2002].
The active particle size and corresponding statistical analysis are presented in Table 5.9.
Accordingly, all parameters, except the 4-factor interaction, trTh, seem to contribute
significantly to particle size. Among these, the most influential factors are primary
variables - t, h, T & r, followed by 2-factor interactions - th & rT, and 3-factor
interaction, trh. Smaller particle size is favoured by a decrease in impregnation pH
level, calcination time and heating rate. On the other hand, lower temperature does not
support reduction in particle size.
151
Chapter 5: Bimetallic Catalyst Design
These propositions are further substantiated by the interaction terms, th, rT, and trh. The
positively-signed th suggests parallel interplay between temperature and pH level where
in this case the magnitudes of both factors should be decreased in a bid to produce
smaller metal crystallites, i.e. (+) = (+)(+). In contrast, the negative sign in rT indicates
that heating rate and temperature must be adjusted in opposite directions to bring about
mutual enhancement, i.e. (-) = (+)(-). In agreement with metallic dispersion and surface
area analyses, it is expected that lower heating rate and higher temperature during
thermal treatment would yield widely-dispersed, active particles with smaller size.
Thus, it seems that higher calcination temperature supports distribution and dispersion
of metal oxide particles, however prolonged conditioning at such circumstance could
pose a risk of losing metallic sites due to particle agglomeration. Interestingly, the 3-
factor interaction, trh, with a positive sign signalled the same indication in which the
crystallite size can be effectively reduced by lowering all of impregnation pH level,
calcination time and heating rate, i.e. (+) = (+)(+)(+), which can also be written as (-) =
(-)(-)(-).
The active particle sizes of the Co- and Ni-catalysts are measured as 17.22 and 15.23
nm (cf. Table 5.10). The particle size of bimetallic Sample L6 (17.94 nm) is greater
compared to both monometallic systems which could implicate on the presence of
bimetallic clusters in the Co-Ni system [Sinfelt, 1973; Heracleous et al., 2005].
Consequently, the presence of smaller particles in corresponding monometallic cobalt
and nickel catalysts leads to greater coverage as reflected by the higher metal surface
areas.
152
Chapter 5: Bimetallic Catalyst Design
Table 5.5. Metal dispersions of bimetallic catalysts and corresponding Yates
analysis.
Sample PD (%)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 6.19 16 4.97 Average
L2 6.06 8 -1.55 t 26.13
L3 3.95 8 -1.24 r 20.93
L4 3.75 8 0.15 tr 2.58
L5 8.08 8 1.24 T 21.02
L6 5.62 8 -0.74 tT 12.49
L7 5.55 8 0.22 rT 3.65
L8 4.96 8 0.28 trT 4.72
H1 4.64 8 -1.10 h 18.60
H2 3.41 8 -0.70 th 11.88
H3 4.23 8 0.70 rh 11.84
H4 2.56 8 -0.30 trh 5.08
H5 6.84 8 0.18 Th 3.03
H6 3.87 8 -0.06 tTh 1.00
H7 6.47 8 -0.12 rTh 2.09
H8 3.35 8 -0.21 trTh 3.50
Table 5.6. Metal dispersions of monometallic catalysts.
Sample Metal dispersion, PD (%)
L6Co 5.10
L6Ni 6.79
153
Chapter 5: Bimetallic Catalyst Design
Table 5.7. Metal surface areas of bimetallic catalysts and corresponding Yates
analysis.
Sample S
m
(m
2
g
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 8.28 16 6.65 Average
L2 8.11 8 -2.06 t 25.91
L3 5.28 8 -1.65 r 20.76
L4 5.02 8 0.20 tr 2.56
L5 10.81 8 1.66 T 20.87
L6 7.52 8 -0.99 tT 12.40
L7 7.41 8 0.29 rT 3.65
L8 6.63 8 0.37 trT 4.69
H1 6.20 8 -1.47 h 18.48
H2 4.56 8 -0.94 th 11.77
H3 5.65 8 0.94 rh 11.75
H4 3.42 8 -0.40 trh 5.04
H5 9.14 8 0.24 Th 3.01
H6 5.17 8 -0.08 tTh 1.00
H7 8.65 8 -0.16 rTh 2.05
H8 4.48 8 -0.28 trTh 3.47
Table 5.8. Metal surface areas of monometallic
catalysts.
Sample Metal surface area, S
m
(m
2
g
-1
)
L6Co 7.60
L6Ni 8.79
154
Chapter 5: Bimetallic Catalyst Design
Table 5.9. Active particle sizes of bimetallic catalysts and corresponding Yates
analysis.
Sample APS (nm)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 16.29 16 22.21 Average
L2 16.64 8 7.32 t 286.99
L3 25.54 8 5.55 r 217.49
L4 26.88 8 1.07 tr 41.98
L5 12.48 8 -5.56 T 217.91
L6 17.94 8 1.04 tT 40.93
L7 18.19 8 -2.31 rT 90.61
L8 20.32 8 -1.10 trT 43.31
H1 21.74 8 5.86 h 229.79
H2 29.59 8 5.00 th 195.95
H3 23.85 8 -1.35 rh 52.98
H4 39.41 8 1.65 trh 64.88
H5 14.76 8 -1.45 Th 56.99
H6 26.09 8 -0.43 tTh 17.01
H7 15.59 8 0.54 rTh 21.20
H8 30.11 8 -0.03 trTh 1.00
Table 5.10. Active particle sizes of monometallic
catalysts.
Sample Active particle size, APS (nm)
L6Co 17.22
L6Ni 15.23
155
Chapter 5: Bimetallic Catalyst Design
Furthermore, it may be observed that the H
2
-chemisorptive parameters, i.e. dispersion,
metal surface area and particle size, all consistently point to same conclusion. For all
bimetallic specimens investigated, a reduction in particle size was followed by a
corresponding increase in dispersion and metal surface area, and vice versa.
5.4.1.3 Acidity
In a supported bimetallic catalyst, alloy formation between two oxides and metal-
support interaction usually results in acidic interfaces. The attributes of these acidic sites
may be obtained from NH
3
-TPD data in terms of acidic site strength (measured by heat
of NH
3
desorption for, -AH
des
) and concentration (indicated by net amount of NH
3
adsorbed), V
d
. Table 5.11 provides the -AH
des
-values and corresponding Yates analysis
for all bimetallic catalysts.
At 95% confidence interval, the strength of acid sites were found to be influenced by
impregnating pH, h, calcination time, t, temperature, T, heating rate, r, 2-factor
interactions - rh, Th, rT, th & tr, 3-factor interaction - trT, and 4-factor interaction -
trTh. From the variance analyses, higher energy was required to overcome the bonds
formed between the acid sites and the adsorbed NH
3
molecules for the catalysts
prepared under more acidic impregnating solution, longer calcination time, faster
heating rate and lower temperature. These patterns are supported by two most important
variables 2-factor interactions, rh (negative-sign), i.e. (-) = (-)(+), & Th (positive-sign),
i.e. (+) = (-)(-), and the 4-factor interaction, trTh (positive-sign), i.e. (+) = (+)(+)(-)(-).
Table 5.12 denotes that the -AH
des
-values of the monometallic Co and Ni catalysts were
evidently lower which confirm poorer acidity in the absence of alloy formation.
156
Chapter 5: Bimetallic Catalyst Design
Table 5.11. Heats of desorption of bimetallic catalysts and corresponding Yates
analysis.
Sample
-AH
des
(kJ mol
-1
K
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 58.67 16 45.08 Average
L2 63.56 8 15.55 t 15.88
L3 58.08 8 6.46 r 6.60
L4 94.19 8 3.75 tr 3.83
L5 17.12 8 -10.15 T 10.37
L6 40.19 8 -2.06 tT 2.11
L7 49.92 8 6.12 rT 6.25
L8 65.84 8 -4.37 trT 4.47
H1 31.79 8 -21.74 h 22.21
H2 45.89 8 -4.45 th 4.54
H3 16.82 8 -15.66 rh 15.99
H4 32.19 8 -2.27 trh 2.31
H5 36.19 8 15.21 Th 15.53
H6 41.33 8 -1.57 tTh 1.60
H7 29.80 8 -0.98 rTh 1.00
H8 39.61 8 5.22 trTh 5.33
Table 5.12. Heats of desorption of monometallic
catalysts.
Sample Heat of desorption, -AH
des
(kJ mol
-1
K
-1
)
L6Co 27.66
L6Ni 15.66
157
Chapter 5: Bimetallic Catalyst Design
The NH
3
desorption profiles for the low (L) and high (H) pH catalysts are displayed in
Figs. 5.4(a) and (b) respectively. The desorption peaks of Catalysts L appeared in the
range 600-650 K which are approximately 50 K higher than those of type H catalysts
with peaks at 550-600 K. This provides further indication that the low-pH specimens
possessed more superior acid strength than high-pH specimens as also demonstrated
previously by the Yates analysis in Table 5.11. Similarly, as depicted in Fig. 5.4(c), the
L6Co and L6Ni catalysts showed TPD peaks at 555 and 570 K respectively (lower than
590 K for L6 catalyst). Even so, since the temperature peaks were all less than 673 K, it
appears that the acid sites on all specimens were predominantly medium or weak Lewis
acid centres. Meanwhile, signal bumps detected at 1023 K were an electronic artefact
caused by temperature overshot.
As the concentration of acid site is proportional to the amount of NH
3
desorbed, the
total number of acid sites may be estimated from integration of desorption peaks
[Spiewak et al., 1997]. The net amount of NH
3
desorbed for bimetallic catalysts, as well
as the relevant statistical analysis, are summarised in Table 5.13.
158
Chapter 5: Bimetallic Catalyst Design
273 373 473 573 673 773 873 973 1073 1173 1273
Temperature (K)
T
C
D

s
i
g
n
a
l
(L1)
973
(L2)
(L3)
(L4)
(L5)
(L6)
(L7)
(L8)
Fig. 5.4(a). NH
3
-TPD profiles for low-pH (L) catalysts.
159
Chapter 5: Bimetallic Catalyst Design
273 373 473 573 673 773 873 973 1073 1173 1273
Temperature (K)
T
C
D

s
i
g
n
a
l
(H1)
973
(H2)
(H3)
(H4)
(H5)
(H6)
(H7)
(H8)
Fig. 5.4(b). NH
3
-TPD profiles for high-pH (H) catalysts.
160
Chapter 5: Bimetallic Catalyst Design
373 473 573 673 773 873 973
Temperature (K)
T
C
D

s
i
g
n
a
l
(L6)
(L6Co)
(L6Ni)
Fig. 5.4(c). NH
3
-TPD profiles for Catalysts L6, L6Co and L6Ni.
Accordingly, the variables registering significance impact on the amount of NH
3
adsorbed (at 95% confidence level) are primary factors - T, h, t & r, 2-factor interactions
- rT, tr, Th & th, and 3-factor interactions - trT, rTh & trh. Additionally, it appears that
higher-pH impregnation, shorter calcination time, faster heating rate and lower
temperature increased the extent of NH
3
desorption at the surface of a bimetallic
catalyst. Among other influential variables, 2-factor interaction, tr (negative-sign), i.e. (-
) = (-)(+), and 3-factor interactions, trT (positive-sign), i.e. (+) = (-)(+)(-), & trh
(negative-sign), i.e. (-) = (-)(+)(+), indicated similar trends.
Although bimetallic systems exhibited stronger acid strength (as reflected by high -
AH
des
-values), the surfaces of monometallic catalysts shown in Table 5.14 are however
populated by more acid sites leading to greater amount of NH
3
adsorption than the
bimetallic catalysts.
161
Chapter 5: Bimetallic Catalyst Design
Table 5.13. Amounts of NH
3
desorbed in bimetallic catalysts and corresponding
Yates analysis.
Sample
V
d
(mol g
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 1.74 16 1.79 Average
L2 1.87 8 -0.12 t 10.67
L3 1.99 8 0.12 r 10.04
L4 1.80 8 -0.08 tr 6.87
L5 1.68 8 -0.20 T 17.07
L6 1.41 8 -0.02 tT 1.74
L7 1.62 8 0.10 rT 8.59
L8 1.65 8 0.14 trT 12.46
H1 1.90 8 0.14 h 12.17
H2 2.01 8 -0.05 th 4.35
H3 2.13 8 0.03 rh 2.20
H4 1.67 8 -0.08 trh 6.54
H5 1.70 8 0.06 Th 5.59
H6 1.56 8 0.03 tTh 2.46
H7 2.06 8 0.10 rTh 8.52
H8 1.87 8 -0.01 trTh 1.00
Table 5.14. Amounts of NH
3
desorbed in monometallic
catalysts.
Sample Amount of NH
3
desorbed, V
d
(mol g
-1
)
L6Co 2.669
L6Ni 2.846
162
Chapter 5: Bimetallic Catalyst Design
In view of the low heat of desorption values (-AH
des
<< 125 kJ mol
-1
) [Yaluris et al.,
1996] and the moderate range of desorption peaks (550-650 K) as according to
Santacesarias acid strength classification [Auroux et al., 2001; Iengo et al., 1998a;
Iengo et al., 1998b], the dominant acid type in the bimetallic Co-Ni and their
corresponding monometallic systems prepared for this study may be generally described
as medium Lewis acid sites.
5.4.1.4 Model building and optimisation
Further to qualitative analysis, polynomial models may be developed for descriptive and
optimisation purposes. For a 2
4
-factorial design, the response variable, Y
i
, may be
described in terms of the primary factors and factor-interaction effects as,
h t T t r t h T r t i
X X a X X a X X a X a X a X a X a a Y
7 6 5 4 3 2 1 0
+ + + + + + + =
h r t T r t h T h r T r
X X X a X X X a X X a X X a X X a
12 11 10 9 8
+ + + + +
h T r t h T r h T t
X X X X a X X X a X X X a
15 14 13
+ + +
(5.5)
where X
t
, X
r
, X
T
and X
h
are the re-scaled variables defined as:
L U
L
t
t t
t t
X

=
L U
L
r
r r
r r
X

=
L U
L
T
T T
T T
X

=
L
h
U L
h h
X
h h

(5.6)
The independent variables for calcination time, t, heating rate, r, temperature, T, and
impregnation pH level, h, were appropriately scaled in order to minimise numerical
errors in subsequent multilinear regression that may occur due to direct use of original
163
Chapter 5: Bimetallic Catalyst Design
variables with different orders of magnitude [Nguyen et al., 2003]. Additionally,
subscripts U and L refer to the upper and lower values of each independent variable
respectively.
Consequently, the data obtained from each measure of the physicochemical properties
were used to evaluate the coefficients a
0
to a
15
in Eq. 5.5 via multilinear regression tool
in Microsoft Office Excel 2003. Nevertheless, in order to simplify the multilinear
expression, regression analysis may be carried out by neglecting statistically-
insignificant terms (factors and interactions) as previously indicated by F-value
comparison (assuming 95% confidence level). Table 5.15 summarises the coefficients
a
0
to a
15
and the corresponding correlation parameters (R
2
-values) for each response
variable.
Subsequently, constrained optimisation was carried out to yield the optimum
preparative conditions. The resulting best factor levels for various catalyst
physicochemical attributes are presented in Table 5.16. These were obtained as X
t
= 0 (t
= 1 h), X
r
= 0 (r = 5 K min
-1
), X
T
= 0 (T = 873 K) and X
h
= 1 (h = 8) which yields a
predicted S
BET
of 164.80 m
2
g
-1
(identical to Sample H1). By the same token, highest
heat of desorption, -AH
des
, of 94.19 kJ mol
-1
K
-1
(equivalent to Sample L4) - a reflection
of optimum acidic site strength - is expected when adopting X
t
= 1 (t = 5 h), X
r
= 1 (r =
20 K min
-1
), X
T
= 0 (T = 873 K) and X
h
= 0 (h = 2). On the other hand, smaller
crystallite size is required for attaining a catalyst with superior dispersion, in which case
X
t
= 0 (t = 1 h), X
r
= 0 (r = 5 K min
-1
), X
T
= 1 (T = 973 K) and X
h
= 0 (h = 2) may be
implemented to give the minimum value of 12.47 nm (as in Sample L5).
164
Chapter 5: Bimetallic Catalyst Design
Table 5.15. Coefficients of the regression polynomials for each response variable.
Polynomial
coefficients
S
BET
(m
2
g
-1
)
PD
(%)
S
m
(m
2
g
-1
)
APS
(nm)
-AH
des
(kJ mol
-1
K
-1
)
V
d
(mol g
-1
)
a
0
120.01 6.06 8.10 16.30 52.40 1.84
a
1
-1.55 -0.17 -0.22 0.32 16.35 -0.07
a
2
-0.04 -2.22 -2.96 9.23 9.13 0.16
a
3
-10.43 2.08 2.78 -3.83 -31.38 -0.26
a
4
44.79 -1.36 -1.82 5.43 -16.64 0.13
a
5
-0.86 1.04 13.94 -0.13
a
6
0.48 -2.11 -2.82 5.16
a
7
-0.88 -1.17 7.55 -9.10 0.05
a
8
-0.90 -0.20 -0.27 -3.49 18.68 -0.12
a
9
-2.72 1.93 2.58 -7.09 -31.52
a
10
8.99 -3.12 30.22 -0.07
a
11
1.51 2.02 -4.42 -13.29 0.25
a
12
-1.02 -0.80 -1.08 6.62 -0.30
a
13
-1.74
a
14
-1.25 2.16 0.40
a
15
0.67 -0.49 -0.66 0.82
R
2
0.999 0.990 0.990 0.999 0.974 0.903
165
Chapter 5: Bimetallic Catalyst Design
Table 5.16. Values of best factor levels for the bimetallic Co-Ni catalysts.
Factor
S
BET
(m
2
g
-1
)
PD
(%)
S
m
(m
2
g
-1
)
APS
(nm)
-AH
des
(kJ mol
-1
K
-1
)
V
d
(mol g
-1
)
t 1 1 1 1 5 1
r 5 5 5 5 20 20
T 873 973 973 973 873 873
h 8 2 2 2 2 8
Optimum
value
164.80 8.14 10.88 12.47 94.19 2.13
5.4.2 Structural (bulk and surface) studies
Relevant morphological characteristics of prepared catalysts were studied via X-ray
diffraction (XRD), transmission electron microscopy (TEM) and scanning electron
microscopy (SEM). The XRD analysis revealed the types of phases (metallic, oxides,
alloys, or aluminates) that metal components formed on the catalyst surface. TEM
examination provides 2-dimensional contrasting image featuring metal deposition sites.
On the other hand, surface structure may be assessed using 3-dimensional micro-sized
images taken from SEM observation.
5.4.2.1 X-ray diffractograms
XRD patterns were measured to examine the resulting bulk-phase modifications due to
solid reactions that occurred during activation phase. Measurements were therefore
performed following calcination as well as the subsequent reduction on four samples
(L6, H6, L6Co and L6Ni). The XRD patterns of calcined (but unreduced) bimetallic L6
166
Chapter 5: Bimetallic Catalyst Design
and H6 samples in Fig. 5.5 showed large peaks for CoAl
2
O
4
(2u = 36.7) and NiAl
2
O
4
(37 and 44.8). Spinel-type NiCo
2
O
4
alloy (31 and 36.6), Co
3
O
4
(31.2 and 36.8)
and NiO (43.2) are also present, albeit in modest amounts as indicated by smaller
peaks. From the same Fig. 5.5, the monometallic cobalt system (L6Co) revealed that
cobalt by its self has greater tendency to form Co
3
O
4
than the more stable aluminates. In
comparison, NiAl
2
O
4
peaks of the monometallic nickel specimen (L6Ni) however
appeared in similar intensities as in the two bimetallic catalysts.
10 20 30 40 50 60
2-theta angle (deg)
I
n
t
e
n
s
i
t
y
NiAl
2
O
4
, NiCo
2
O
4
NiO
NiAl
2
O
4
NiCo
2
O
4
Co
3
O
4
-Al
2
O
3
Co
3
O
4
CoAl
2
O
4
, Co
3
O
4
-Al
2
O
3
NiO
NiAl
2
O
4
-Al
2
O
3
L6 (Co-Ni, pH 2)
L6Ni (Ni, pH 2)
NiAl
2
O
4
CoAl
2
O
4
, Co
3
O
4
H6 (Co-Ni, pH 8)
L6Co (Co, pH 2)
Co
3
O
4
NiCo
2
O
4
CoAl
2
O
4
, Co
3
O
4
NiAl
2
O
4
, NiCo
2
O
4
NiO
NiAl
2
O
4
-Al
2
O
3
Fig. 5.5. XRD patterns for freshly-calcined Catalysts L6, H6, L6Co and L6Ni.
Following reduction, the NiCo
2
O
4
, Co
3
O
4
and NiO were mostly converted to metallic
Co (44.2 and 51.5) and Ni (44.5 and 51.8) phases as displayed in Fig. 5.6. Even so,
both metal-support interactions CoAl
2
O
4
and NiAl
2
O
4
were more resistant to H
2
-
reduction at 873 K as implicated by smaller decrease in peak intensities for these
species. Further, the extent of reduction was higher in the low-pH catalyst (L6) than the
167
Chapter 5: Bimetallic Catalyst Design
high-pH catalyst (H6). Not surprisingly, the previous H
2
-chemisorption analysis indeed
demonstrated that sample L6 yielded higher H
2
uptake (which in turn gave higher
dispersion and metal surface area) compared to the more basic H6 catalyst. Moreover,
upon H
2
-treatment it appeared that almost all Co
3
O
4
and even CoAl
2
O
4
species in the
L6Co catalyst were transformed to form Co particles. A trace of Co
3
O
4
observed in the
reduced state of L6Co was probably due to re-oxidation caused by atmospheric air. On
the other hand, NiAl
2
O
4
seems to possess higher stability than CoAl
2
O
4
. For the L6Ni
specimen, there was no substantial change detected for NiAl
2
O
4
peak whereas NiO was
completely converted to the metallic Ni phase.
10 20 30 40 50 60
2-theta angle (deg)
I
n
t
e
n
s
i
t
y
NiAl
2
O
4
Co
3
O
4
Co
3
O
4
NiAl
2
O
4
L6 (Co-Ni, pH 2)
NiAl
2
O
4
Ni
Co
Ni
Co
Co
Co
Ni
Ni
NiAl
2
O
4
CoAl
2
O
4
CoAl
2
O
4
H6 (Co-Ni, pH 8)
L6Co (Co, pH 2)
L6Ni (Ni, pH 2)
NiAl
2
O
4
CoAl
2
O
4
Ni
Co
NiAl
2
O
4 Ni
Co
Fig. 5.6. XRD patterns for freshly-reduced Catalysts L6, H6, L6Co and L6Ni.
5.4.2.2 Transmission electron microscopy
The TEM analysis is particularly useful to study the metal deposition site, which is
dependent on the mobility of the corresponding metallic ions (as largely determined by
168
Chapter 5: Bimetallic Catalyst Design
solution pH) during impregnation. Thus, for comparison of relevant features, the TEM
micrographs were obtained for reduced Catalysts L6 and H6 (cf. Figs. 5.7(a) and (b)
respectively).
In the acid L6 catalyst, the metal species was deposited primarily - as shown by darker
section - within the particle core. This phenomenon is commonly referred as eggyolk
profile. In contrast, metal appeared to be deposited closer to the particle exterior in the
more basic H6 catalyst. The similarity in potential of both positively-charged metal ions
and support surface at low pH allowed ions to travel further into the support interior
giving rise to the profile shown in L6 specimen, while increasing pH resulted in rapid
metal adsorption at the surface, as seen in H6 specimen, since the surface loses its
positively-charged potential [Geus and van Veen, 1999].
Fig. 5.7(a). TEM micrograph for reduced Catalyst L6.
169
Chapter 5: Bimetallic Catalyst Design
Fig. 5.7(b). TEM micrograph for reduced Catalyst H6.
5.4.2.3 Scanning electron microscopy
The structural features observed via SEM revealed homogenous distributions of
alumina particles and metal crystallites. These observations further suggest that the
surfaces of reduced catalysts possessed high-level dispersion at both micro- and nano-
levels. Closer inspection of reduced L6 and L6Co (as displayed in Figs. 5.8(a) and (b)
respectively) showed that their surfaces appeared rougher and bulky indicating the
presence of Co and Ni atoms, which were resulted from the transformation of metal
aluminates [Hardiman et al., 2004]. However, as substantiated by previous XRD
analysis, the highly-structured and -ordered NiAl
2
O
4
aluminate species in L6Ni were
largely unaffected by H
2
-reduction at 873 K, hence it is expected that the surface of this
specimen remained smooth (cf. Fig. 5.8(c)).
170
Chapter 5: Bimetallic Catalyst Design
Fig. 5.8(a). SEM micrograph for reduced Catalyst L6.
Fig. 5.8(b). SEM micrograph for reduced Catalyst L6Co.
171
Chapter 5: Bimetallic Catalyst Design
Fig. 5.8(c). SEM micrograph for reduced Catalyst L6Ni.
5.4.3 Thermal analyses
Temperature-programmed calcination, reduction and oxidation runs may provide
information on the stability and reactivity of the catalysts under reforming conditions.
Additionally, the transient weight data obtained may also be used to extract solid-state
kinetic parameters which uniquely characterise the solid behaviour in subsequent
catalytic reactions.
5.4.3.1 Calcination
The decreasing solid weight during calcination may be attributed mostly to
decomposition of precursor salts, although moisture evaporation also contributes to
weight loss at temperatures below 393 K. Fig. 5.9(a) compares the weight derivative
profiles of the acid L6 catalyst and the basic H6 catalyst. These curves are characterised
by a peak at about 492 K signalling decomposition of the main metal nitrate, which
172
Chapter 5: Bimetallic Catalyst Design
gave rise to the formation of metal oxides (NiO, NiCo
2
O
4
and Co
3
O
4
). The peak shape
(width and intensity) reflects the reaction rate which is in turn influenced by the nature
of active site distribution (i.e. shape, number and distribution) on the catalyst surface
[Hanic et al., 1985]. The broader and shorter calcination peak for the catalyst
impregnated at pH 2 may be an indication of the more distributed metal deposition site
at the catalyst pellet in comparison to the basic catalyst (which shows a sharp and
narrow peak). This, indeed, corroborates previous results obtained from H
2
-
chemisorption and TEM image. Since the procurement intrinsic kinetics requires
elimination or minimisation of mass transport effects, a catalyst with uniform metal
distribution throughout the support matrix is most preferred [Adesina, 1996].
Consequently, the thermogravimentric investigation was focussed on the low-pH (L)
catalyst series.
423 473 523 573 623 673 723
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
(L6)
(H6)
Fig. 5.9(a). Calcination weight derivative profiles for acid L6 and basic H6
catalysts.
173
Chapter 5: Bimetallic Catalyst Design
The thermal spectra of the L samples treated under varying calcination conditions (as
outlined in Table 5.1) are depicted in Fig. 5.9(b) from which it can be clearly seen that
metal nitrate decomposition rates were strongly affected by heating rate. Calcination at
higher heating rate resulted in much sharper and narrower peaks - an indication of the
more rapid reaction rate due to the faster increase in temperature.
Fig. 5.9(c). plots the weight derivative profiles for bimetallic L6 and its monometallic
constituents, L6Co and L6Ni catalysts. The peak observed at around 373 K in all
systems was attributed to the removal of hydration water. The calcination spectrum of
Catalyst L6Co is characterised by a single sharp main peak at 415 K implying that
decomposition of precursors originating from cobalt salts was a rapid one-step process.
However, the Catalyst L6Ni comprised of a very broad profile consisting of multi peaks
at 363, 396, 433, 491 and 558 K, where the first peak was assigned to the moisture
removal and the others corresponded to sequential decomposition of nitrate groups. In a
thermogravimetric study by Marturano et al. [1999a], they reported the presence of
three peaks at 405, 470 and 574 K in the calcination of -Al
2
O
3
supported Ni catalyst,
which were attributed to the removal of hydration water and by nitrate groups of
precursor. The more decomposition peaks observed in the current study may be
implicated to additional transformations involved resulting from higher degree and
more complex interaction between the less stable -support (as compared to -Al
2
O
3
)
with the metal precursors. The bimetallic L6 showed a narrower profile compared to
L6Ni, but still broader and marked by more decomposition steps than L6Co. Thus, it is
174
Chapter 5: Bimetallic Catalyst Design
not surprising that the magnitudes of metal dispersion and metal surface area (measured
by H
2
-chemisorption runs) follow the order: L6Ni > L6 > L6Co.
0 1 2 3 4 5 6 7
Time (h)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
(L1)
(L2)
(L3)
(L4)
(L5)
(L6)
(L7)
(L8)
Fig. 5.9(b). Calcination weight derivative profiles for low-pH (L) catalysts.
175
Chapter 5: Bimetallic Catalyst Design
0
0.4
0.8
1.2
1.6
2
273 373 473 573 673 773 873 973
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
(L6)
(L6Ni)
(L6Co)
Fig. 5.9(c). Calcination weight derivative profiles for Catalysts L6, L6Co and L6Ni.
Furthermore, using transient weight conversion data, kinetic rate parameters for the
metal nitrate decomposition during calcination may be evaluated. Therefore, it is was
necessary to estimate the solid conversion, , defined as,
f i
i
w w
w w

= o
(5.7)
where w = instantaneous weight,
w
i
= initial weight,
w
f
= final weight.
The solid-state reaction data may be described using one of Avrami-Erofeev (A), Prout-
Tompkins (B), geometrical (R), diffusion (D) or order of reaction (F) models,
depending on the type of transient conversion curve. A typical plot of weight-based
176
Chapter 5: Bimetallic Catalyst Design
conversion, , versus time, t, presented in Fig. 5.10 reveals an S-shaped profile, which
suggests that the metal nitrate decomposition to the oxide phase proceeds via an initial
induction period followed by reaction-controlled oxide evolution before a final
relaxation (intraphase restructuring) [Hardiman et al., 2004].
0
0.2
0.4
0.6
0.8
1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (h)
o
Fig. 5.10. A typical weight-based conversion profile during the calcination run.
As proposed by Brown [2001] and Kapteijn et al. [1999], the resulting sigmoid curve
(cf. Fig. 5.10) indicates that the kinetic parameter, k
s
, may be more suitably evaluated
via Avrami-Erofeev (A) model given by,
| | t k
s
n
=
1
) 1 ( ln o 2 s n s 4 (5.8)
where k
s
= kinetic coefficient (h
-1
),
t = time (h),
n = order in the Avrami-Erofeev model.
177
Chapter 5: Bimetallic Catalyst Design
The order n essentially gives a reflection on the shape of the conversion curve, which in
turn also depends on the characteristic of the thermal peak. As slow decomposition is
indicated by a broad and short peak, the corresponding conversion profile would be
marked by a slow initial increase, followed by a constant rise and finally a slow
relaxation phase. Such a process may be better described by an Avrami-Erofeev model
with a lower order n. On the other hand, fast decomposition results in a narrow and
sharp peak, where the corresponding conversion profile would reveal a faster initial
increase, followed by a steeper linear rise and then a rapid stabilisation. Such a process
may be better described by an Avrami-Erofeev model with a higher order n.
The calcination kinetic coefficients, k
s,cal
, estimated for the various orders (n = 2, 3 and
4) in the Avrami-Erofeev model are displayed in Table 5.17, where the corresponding
R
2
-estimates (shown in parenthesis) provides a comparison on the degree of fitness for
each order. Although the qualitative trend of k
s,cal
-values across column between all the
catalysts were independent of the Avrami-Erofeev order, n, it is however evident that n
= 3 gave a better fit for the overall system based on statistical R
2
- values.
From Table 5.17, the k
s,cal
-value for Catalyst H6 is almost three times higher in
magnitude compared to the acid L6 specimen. This is indeed an implication of the
sharper and narrower peak feature corresponding to nitrate decomposition as depicted in
the spectrum of Catalyst H6 (cf. Fig. 5.9(a)).
178
Chapter 5: Bimetallic Catalyst Design
In light of the statistical consideration, the k
s,cal
-values estimated from 3
rd
-order Avrami-
Erofeev model were subjected to Yates statistical assessments detailed in Table 5.18.
For the 2
3
experimental design, the variance ratio of a statistically-significant effect is
F
4,4
> 6.39 at 95% confidence interval. Accordingly, the Yates technique revealed that
the important variables affecting solid-state kinetic constants are heating rate, r,
temperature, T, and the 3-factor interaction - trT. The variance estimates indicated that
increase in both r and T resulted in higher k
s,cal
-value. For maximum k
s,cal
, the
negatively-signed trT proposed that one of the factors must be lowered, i.e. (-) = (-
)(+)(+), or all factors are to be decreased, i.e. (-) = (-)(-)(-). Consistent with previous
suggestion that higher k
s,cal
-value is possible by increasing both r and T, it would appear
that calcination time, t, may need to be reduced for mutual interaction to be more
effective. Even though the variance of primary variable t exhibits a positive small value,
the effect of this factor is, however, less important as reflected by the low F-value.
By taking into account the relevant variables, the regression model for the calcination
kinetic coefficients may thus be proposed as:
T r t T r k
X X X a X a X a a Y
cal s
3 2 1 0
,
+ + + =
(5.9)
with a
0
= 3.94, a
1
= 4.52, a
2
= 0.27 and a
3
= -0.33. Based on constrained optimisation, it
is evident that the maximum k
s,cal
-value may be achieved under X
t
= 0 (t = 1 h), X
r
= 1 (r
= 20 K min
-1
) and X
T
= 1 (T = 873 K) with k
s,cal
= 8.73 h
-1
. This also implies that to
some extent calcination performed at shorter time may result in more rapid removal of
nitrate precursors (higher k
s,cal
-value).
179
Chapter 5: Bimetallic Catalyst Design
Table 5.17. Calcination kinetic coefficients, k
s,cal
, and R
2
-values for the various
orders in the Avrami-Erofeev model.
n in the Avrami-Erofeev model
Sample
2 3 4
L1 5.74 (0.970) 4.04 (0.986) 3.16 (0.981)
L2 5.40 (0.971) 3.81 (0.986) 2.97 (0.982)
L3 11.61 (0.967) 8.24 (0.988) 6.46 (0.987)
L4 12.22 (0.965) 8.69 (0.987) 6.82 (0.987)
L5 5.66 (0.971) 3.99 (0.987) 3.12 (0.983)
L6 6.31 (0.969) 4.45 (0.985) 3.47 (0.981)
L7 12.23 (0.965) 8.71 (0.987) 6.83 (0.987)
L8 11.81 (0.967) 8.40 (0.988) 6.58 (0.987)
H6 15.59 (0.423) 12.22 (0.508) 10.21 (0.520)
The numbers in parenthesis are the R
2
-coefficients.
180
Chapter 5: Bimetallic Catalyst Design
Table 5.18. Yates analysis for calcination kinetic constants (n = 3) of low-pH (L)
catalysts.
Sample
k
s,cal
(h
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 4.04 8 6.29 Average
L2 3.81 4 0.09 t 4.94
L3 8.24 4 4.44 r 245.23
L4 8.68 4 -0.02 tr 1.12
L5 3.99 4 0.19 T 10.42
L6 4.45 4 -0.02 tT 1.00
L7 8.71 4 -0.10 rT 5.76
L8 8.40 4 -0.36 trT 20.11
181
Chapter 5: Bimetallic Catalyst Design
5.4.3.2 Reduction-oxidation
The temperature-programmed reduction (TPR) and oxidation (TPO) experiments were
performed in series by ramping to 973 K at 5 K min
-1
and holding for 1 h followed by
cooling to ambient temperature at 5 K min
-1
. Prior to each run, the catalysts were pre-
treated in inert N
2
flow at 423 K for 1 h to remove interstitial water and possible
impurities. Fig. 5.11 illustrates the TPR-TPO cycle of the low (L6) and high pH (H6)
catalysts as well as the monometallic cobalt (L6Co) and nickel (L6Ni) catalysts. The
TPR-TPO was repeated in a second cycle for the two solids with different pH condition
(L6 and H6) in order to compare the phase stability of these catalysts which have
different metal distribution profile.
88
90
92
94
96
98
100
0 2 4 6 8 10 12 14 16 18 20
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
w
t

%
)
273
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
(L6)
(H6)
TPR TPR TPO TPO
(L6Co)
(L6Ni)
Fig. 5.11. Weight profiles of TPR-TPO-TPR-TPO runs for Catalysts L6, H6, L6Co
and L6Ni.
182
Chapter 5: Bimetallic Catalyst Design
Table 5.19 displays the associated percent weight changes and extents of conversion (in
parenthesis) at each stage of the TPR-TPO runs. It seems that the oxides formed at
impregnation pH of 2 were completely reduced regardless of catalyst composition
compared to 82.4% reducibility in the high-pH catalyst (H6). In the following TPO, it
was also revealed that the H6 catalyst exhibited the lowest extent of re-oxidation among
the four catalysts. Nevertheless, both cobalt and nickel catalysts (L6Co and L6Ni)
showed higher degree re-oxidation than Catalyst L6. The second TPR-TPO cycle of L6
and H6 further indicated that under alternating reduction and oxidation conditions, the
low-pH system preserved higher proportion of oxidic phases than the high-pH system.
Since the nature of steam reforming reaction resembles a combination of reducing and
oxidising environments, Catalyst L6 would be a more stable catalyst during prolonged
steam reforming [Hardiman et al., 2005a].
Table 5.19. Percent weight changes and extents of conversion.
Percent of weight change (%)
L6 H6 L6Co L6Ni
TPR I 5.690 (1.000) 4.590 (0.824) 7.881 (1.000) 7.444 (1.000)
TPO I 3.649 (0.655) 2.700 (0.485) 5.507 (0.816) 4.262 (0.824)
TPR II 3.911 (0.702) 2.947 (0.529) - -
TPO II 3.874 (0.696) 2.857 (0.513) - -
The value in parenthesis indicate the extent of conversion
Derivative spectra were then derived from transient weight profile to provide important
information on the type of solid phases during each stage of the thermal treatment. Fig.
183
Chapter 5: Bimetallic Catalyst Design
5.12 plots the weight derivative spectra during reduction (TPR I) of acidic L6 and basic
H6 catalysts which depicts three primary peaks at 493, 655 and 968 K. The lowest
temperature peak at 493 K corresponds to the reduction of Ni
3+
(Ni
2
O
4
) to Ni
2+
or Ni,
where this compound was undetected via XRD measurement due to its amorphous
nature [Mile et al., 1990]. The second peak at 655 K is characterised by a large broader
feature which may be separated into three sub-peaks as signals of NiCo
2
O
4
, Co
3
O
4
and
NiO phases in order of increasing temperature [Haenen et al., 1986; Jacobs et al., 2002;
Li and Chen, 1995], while the highest temperature at 968 K represents the metal
aluminate phases consisting of CoAl
2
O
4
and NiAl
2
O
4
[Roh et al., 2002]. Subsequent re-
oxidation (TPO I) resulted in the reinstatement of the previously-reduced phases as
shown by a single, large and thermally stable peak at about 495 K. In the second H
2
-
reduction (TPR II), the low-temperature Ni
2
O
3
species appeared as a shoulder peak
albeit its peak temperature position is similar as in TPR I (493 K). Moreover, it may be
observed that the main oxide peak ascribed to the major metal oxides (NiCo
2
O
4
, Co
3
O
4
and NiO) shifted to the right to 716 K from its original position at 655 K in TPR I. The
observed changes in shape and position between peaks of TPR I and TPR II reflected
the change in composition of the constituent metal oxides upon re-oxidation of the Co
0
and Ni
0
species back to NiO, Co
3
O
4
and NiCo
2
O
4
during TPR II [Hardiman et al.,
2005a]. The second TPR cycle also accommodates additional reduction of the
remaining metal aluminate phases as featured by a small hump at 973 K in both
catalysts. Upon the second re-oxidation (TPO II), the oxidation profile appears to
parallel the attributes in TPO I demonstrating the reproducibility of identical composite
oxide in both states.
184
Chapter 5: Bimetallic Catalyst Design
373 473 573 673 773 873 973 1073
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
973
TPO I
TPO II
(L6)
(L6)
(H6)
(H6)
TPR I
TPR II
(L6)
(H6)
(H6)
(L6)
Fig. 5.12. TPR-TPO-TPR-TPO weight derivative profiles for acid L6 and basic H6
catalysts.
185
Chapter 5: Bimetallic Catalyst Design
For acidic bimetallic catalysts, the TPR analyses gave weight changes in the range 6-
6.5% for all eight samples indicating full transformation under the reducing conditions
employed. Appendix D.2 displays calculation of the expected weight drop in
thermogravimetric run for complete reduction of the corresponding catalyst with
specified metal loading. The weight derivative profiles depicted in Fig. 5.13 highlight
that the size and position of individual peaks in a TPR spectrum varied depending on
the calcination conditions. At lower calcination temperature, the presence of
metastable spinel-type species NiCo
2
O
4
was more distinct (cf. curves L1-L4).
However, NiCo
2
O
4
peaks observed in the higher calcination temperature (cf. curves L5-
L8) appeared in smaller intensity since most of this less-stable compound was probably
transformed into Co
3
O
4
, NiO, CoAl
2
O
4
or NiAl
2
O
4
species. By the same token, stronger
intensity for aluminate peaks are evident in the Samples L5-L8 consistent with the fact
that the formation of CoAl
2
O
4
and NiAl
2
O
4
phases via metal-support interaction is
promoted by higher calcination temperature.
The position of the TPR peaks as a function of the temperature is summarised in Table
5.20. As may be seen from the qualitative trend in Fig. 5.13, the peaks assigned to
NiCo
2
O
4
, Co
3
O
4
and NiO respectively are reasonably separated from each other in
catalysts with lower calcination temperature (i.e. L1, L2, L3 & L4). On the other hand,
for the catalysts calcined at higher temperature, these peaks were lumped together as a
single broad peak in the 560-730 K range. This is a reflection on the formation of
increasingly complex metal oxide structure favoured by increased calcination
temperature. Similar observation was reported by Arnoldy and Moulijn [1985] for
CoO/Al
2
O
3
catalysts calcined at various temperatures between 380 and 1290 K. They
186
Chapter 5: Bimetallic Catalyst Design
observed that in samples with lower calcination temperatures, five broad peaks
indicative of various forms of cobalt oxide were evident. However, these distinct peaks
became integrated and sharper for catalysts calcined at higher temperatures.
Furthermore, at calcination temperatures greater than 1025 K, a single sharp peak
obtained at around 1200 K suggests a reaction between the metal and alumina support.
In principle, the TPO spectra displayed in Fig. 5.14 implicate similar features for all
eight catalysts. This is shown by a large peak with wide profile between 373 and 673 K.
Nevertheless, the asymmetrical nature of these curves suggests that they may actually
comprise of multiple peaks from re-oxidation of previously reduced phases. Above 673
K, the weight change profiles remained practically constant implicating the superior
thermal stability of oxide structures during re-oxidation. Table 5.21 shows the positions
of the TPO peaks as a function of temperature. It reveals that oxidation of previously-
reduced phases occurred at lower temperatures in catalysts calcined with faster heating
rate and lower temperature.
187
Chapter 5: Bimetallic Catalyst Design
273 373 473 573 673 773 873 973 1073 1173 1273
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
(L1)
973
(L2)
(L3)
(L4)
(L5)
(L6)
(L7)
(L8)
Fig. 5.13. TPR weight derivative profiles for low-pH (L) catalysts.
188
Chapter 5: Bimetallic Catalyst Design
273 373 473 573 673 773 873 973 1073 1173 1273
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
973
(L1)
(L2)
(L3)
(L4)
(L5)
(L6)
(L7)
(L8)
Fig. 5.14. TPO weight derivative profiles for low-pH (L) catalysts.
189
Chapter 5: Bimetallic Catalyst Design
Table 5.20. Positions of the TPR peaks as a function of the temperature.
Peak temperature (K)
Sample
Peak I Peak II Peak III Peak IV Peak V
L1 449.4 561.2 689.2 809.7 942.6
L2 451.2 569.9 698.0 851.7 946.0
L3 493.7 554.4 699.5 777.6 935.0
L4 479.6 557.8 711.8 854.2 954.5
L5 483.4 564.3 614.5 699.9 963.3
L6 483.9 582.2 641.3 730.6 965.4
L7 480.6 571.7 642.1 723.5 960.3
L8 482.0 570.8 641.3 716.0 965.8
Table 5.21. Positions of the TPO peaks as
a function of the temperature.
Sample Peak temperature (K)
L1 470.5
L2 470.9
L3 454.0
L4 462.1
L5 475.0
L6 495.9
L7 470.9
L8 469.9
190
Chapter 5: Bimetallic Catalyst Design
The TPR spectra of monometallic L6Co and L6Ni specimens are displayed in Fig. 5.15.
The reduction profile of L6Co is characterised by a large sharp peak at 830 K and a
much smaller hump at 975 K indicating the reduction of Co
3
O
4
and CoAl
2
O
4
. In the
L6Ni, metal-support interaction seems to be more evident with higher aluminate
(NiAl
2
O
4
) peak at 972 K, while NiO species is present as a modest peak at 832 K. The
reduction of Co
2+
and Ni
2+
to Co
0
and Ni
0
was reported to take place at approximately
673 K [Gonzlez et al., 2005]. However, these authors showed that reduction of metallic
species on supported metal catalysts tend to be more difficult because of metal-support
interaction as indicated by peak shifts to higher temperature.
Furthermore, metal oxide species in the bimetallic L6 were more easily reduced than
either of monometallic L6Co or L6Ni systems. This bimetallic synergism was explained
by Gonzlez et al. [2005] as a result of activated reduction process, in which the
presence of atomic hydrogen (H
-
) arising from the reduction of first metallic species
played a role in assisting the reduction of second metallic species. Interestingly, the
reduction temperatures of aluminate phases for all three catalysts are located in a very
similar range.
The TPO spectra in Fig. 5.16 show that oxidation peak temperature increased in the
order of L6Co (482 K), L6 (495 K) and L6Ni (535 K). In addition, the intensity and size
of peaks appeared to decrease in the same order. These trends correspond to the effect
of different cobalt loading in these three catalysts as it is widely known that cobalt has a
better reactivity than nickel during oxidation. Nevertheless, as previously shown in
191
Chapter 5: Bimetallic Catalyst Design
Table 5.19, the extents of oxidation for monometallic Catalysts L6Co and L6Ni (81.6
and 82.4%) are higher compared to bimetallic Catalyst L6 (65.5%).
0
0.1
0.2
0.3
0.4
0.5
373 473 573 673 773 873 973 1073 1173
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
Co/Al
297333
O
3
973
(L6)
(L6Ni)
(L6Co)
Fig. 5.15. TPR weight derivative profiles for Catalysts L6, L6Co and L6Ni.
-0.03
0
0.03
0.06
0.09
0.12
0.15
373 473 573 673 773 873 973 1073 1173
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
Co/Al
297333
O
3
973
(L6Co)
(L6)
(L6Ni)
Fig. 5.16. TPO weight derivative profiles for Catalysts L6, L6Co and L6Ni.
192
Chapter 5: Bimetallic Catalyst Design
Kinetic analysis was also performed on the reduction and oxidation processes to provide
further insights on the nature of oxide phases formed and reduced phases present in each
catalyst. Similar technique was implemented to evaluate relevant parameters as
previously used for calcination. Thus, conversions were initially plotted with respect to
time revealing sigmoid shapes (as typically shown in Fig. 5.17) which suggest that
kinetic parameters may be best obtained using an Avrami-Erofeev model.
0
0.2
0.4
0.6
0.8
1
0 0.3 0.6 0.9 1.2 1.5
Time (h)
o
Fig. 5.17. A typical weight-based conversion profile during the reduction run.
Table 5.22 summarises the reduction kinetic constants, k
s,red
, for the various order (n =
2, 3 and 4) in the Avrami-Erofeev model. The relevant statistical R
2
-values are shown in
parenthesis. It may be clearly seen that 3
rd
-order Avrami-Erofeev provided the best fit in
light of statistical perspective. Further, while the reduction rate constant of H6 appeared
to be comparable to L6, the ratio of the reduction constant to that of calcination
193
Chapter 5: Bimetallic Catalyst Design
(k
s,red
/k
s,cal
) of the metal oxide peak for L6 is about 0.40 compared to 0.16 for H6. This
shows that the acidic catalyst was more easily reduced, as also corroborated by the fact
that L6 achieved complete reduction while H6 exhibited lower degree of conversion
(82.4%) under the same reducing conditions (cf. Table 5.19).
Accordingly, reduction kinetic constants, k
s,red
, extracted using Avrami-Erofeev with n
= 3 were employed in the Yates analysis (cf. Table 5.23) to contrast the effect of each
factor and their interactions. It was, however, found that none of these variables was
statistically significant (where all F-values are smaller than 6.39 for 95% confidence
interval). Even so, the signs of variance estimates imply that increase in calcination time
and temperature and decrease in heating rate would yield faster reduction rate.
Therefore, the catalyst with highest k
s,red
-value may be produced using the calcination
conditions similar to those adopted in Catalyst L6. The regression model relating the
reduction kinetic coefficients to preparation conditions may be written as:
T r t T t T r t k
X X X a X X a X a X a X a a Y
red s
5 4 3 2 1 0
,
+ + + + + =
(5.10)
with a
0
= 1.42, a
1
= 0.01, a
2
= -0.08, a
3
= 0.08, a
4
= 0.26 and a
5
= -0.19.
194
Chapter 5: Bimetallic Catalyst Design
Table 5.22. Catalyst reduction kinetic coefficients, k
s,red
, and R
2
-values for the
various orders in the Avrami-Erofeev model.
n in the Avrami-Erofeev model
Sample
2 3 4
L1 2.15 (0.963) 1.49 (0.969) 1.16 (0.957)
L2 2.03 (0.966) 1.42 (0.973) 1.10 (0.962)
L3 1.79 (0.949) 1.26 (0.966) 0.98 (0.961)
L4 1.89 (0.937) 1.34 (0.963) 1.04 (0.962)
L5 2.03 (0.966) 1.42 (0.982) 1.10 (0.976)
L6 2.49 (0.961) 1.76 (0.983) 1.37 (0.980)
L7 2.09 (0.933) 1.49 (0.965) 1.17 (0.968)
L8 2.09 (0.955) 1.48 (0.981) 1.16 (0.981)
H6 2.74 (0.952) 1.94 (0.977) 1.52 (0.977)
The numbers in parenthesis are the R
2
-coefficients.
195
Chapter 5: Bimetallic Catalyst Design
Table 5.23. Yates analysis for reduction kinetic constants (n = 3) of low-pH (L)
catalysts.
Sample
k
s,red
(h
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 1.49 8 1.46 Average
L2 1.42 4 0.08 t 3.08
L3 1.26 4 -0.13 r 4.77
L4 1.34 4 -0.05 tr 1.74
L5 1.42 4 0.16 T 5.85
L6 1.76 4 0.08 tT 3.07
L7 1.49 4 0.03 rT 1.00
L8 1.48 4 -0.12 trT 4.60
196
Chapter 5: Bimetallic Catalyst Design
The conversion-time curve of the oxidation displayed in Fig. 5.18 also gave an S-shaped
profile indicative of Avrami-Erofeev model. Comparison of R
2
-values for each Avrami-
Erofeev order, n, presented in Table 5.24 suggest that n = 3 would be most suitable
model for estimating oxidation kinetic parameters.
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2 2.5
Time (h)
o
3
Fig. 5.18. A typical weight-based conversion profile during the oxidation run.
The Yates analysis performed for the parameters extracted via Avrami-Eroofeev with
an order, n, of 3 (detailed in Table 5.25) indicated that none of the calcination variables
are statistically significant in controlling the oxidation kinetic rate, k
s,ox
. Similar to the
findings for TPR, it seems that longer calcination time and higher temperature, as well
as slower heating rate, would lead to higher value of k
s,ox
. This corresponds to the
calcination procedure of Catalyst L6 which evidently exhibits the highest k
s,ox
-value
among all specimens. By same token, regression model of the oxidation kinetic
coefficients may also be derived based on the 3 most relevant variables, as
197
Chapter 5: Bimetallic Catalyst Design
T t T r k
X X a X a X a a Y
ox s
3 2 1 0
,
+ + + =
(5.11)
with a
0
= 0.88, a
1
= -0.08, a
2
= 0.10 and a
3
= 0.14.
198
Chapter 5: Bimetallic Catalyst Design
Table 5.24. Catalyst oxidation kinetic coefficients, k
s,ox
, and R
2
-values for the
various orders in the Avrami-Erofeev model.
n in the Avrami-Erofeev model
Sample
2 3 4
L1 1.32 (0.985) 0.87 (0.986) 0.66 (0.970)
L2 1.31 (0.987) 0.85 (0.989) 0.64 (0.978)
L3 1.28 (0.986) 0.84 (0.986) 0.64 (0.971)
L4 1.22 (0.988) 0.80 (0.984) 0.60 (0.968)
L5 1.43 (0.988) 0.96 (0.999) 0.73 (0.978)
L6 1.72 (0.990) 1.17 (0.994) 0.90 (0.983)
L7 1.37 (0.984) 0.92 (0.988) 0.70 (0.987)
L8 1.46 (0.989) 0.98 (0.991) 0.75 (0.978)
H6 2.71 (0.955) 1.866 (0.990) 1.453 (0.991)
The numbers in parenthesis are the R
2
-coefficients.
199
Chapter 5: Bimetallic Catalyst Design
Table 5.25. Yates analysis for oxidation kinetic constants (n = 3) of low-pH (L)
catalysts.
Sample
k
s,ox
(h
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 0.87 8 0.92 Average
L2 0.85 4 0.05 t 1.90
L3 0.84 4 -0.08 r 2.81
L4 0.80 4 -0.04 tr 1.56
L5 0.96 4 0.17 T 5.95
L6 1.17 4 0.08 tT 3.05
L7 0.92 4 -0.04 rT 1.33
L8 0.98 4 -0.03 trT 1.00
200
Chapter 5: Bimetallic Catalyst Design
5.5 Catalyst activity evaluation
In order to assess coking resilience of the prepared catalysts, steam reforming activity
tests were examined under low steam-to-carbon (S:C) ratio of 1 which gives rise to
simultaneous carbon deposition. The reaction run was performed at 773 K for 6 h. This
temperature ensured minimum contribution of sintering on activity behaviour, while it
was also sufficiently high to carry out steam reforming of propane. In addition, the total
gas flow rate to the fluidised bed reactor was 400 ml min
-1
(instead of 300 ml min
-1
as
specified in Table 4.6) to permit more reactant gases (propane and steam) entering the
reactor, thus allowing better observation of conversion behaviour. This flow is however
still within the recommended fluidisation regime between 200-450 ml min
-1
displayed
in Fig. 4.6 for the fluidised bed system.
Based on transient conversion data, the deactivation feature and reforming performance
of each catalyst would be quantitatively assessed. Levenspiel [1999] has recently
developed an integrated deactivation-reaction model that can be used to simultaneously
extract deactivation and reforming kinetic parameters from conversion-time profile. As
would be detailed in a subsequent chapter, Hardiman et al. [2005b] have provided a
generalisation of this model with respect to reforming order, deactivation form and
reactor type. Furthermore, it was suggested that propane steam reforming accompanied
by carbon-forming reactions in a fluidised bed reformer may be described as a second-
order reaction with hyperbolic deactivation in a CSTR system, which is given by:
201
Chapter 5: Bimetallic Catalyst Design
t
k
k
k
C C
C
W d W
A A
A
1
1
2
0
|
.
|

\
|
' '
'
+
' '
=
|
|
.
|

\
|

t
|
t
|
(5.12)
where = feed propane concentration (mol L
0
A
C
-1
),
C
A
= instantaneous propane concentration (mol L
-1
),
|
W
= Wagner modulus,
k' = steam reforming kinetic constant (L
2
mol
-1
s
-1
g cat.
-1
),
k
d
' = deactivation rate constant (s
-1
),
t' = gas residence time (g cat. s L
-1
),
t = reaction run time (s).
In this section, deactivation, k
d
', and reforming, k', kinetic parameters evaluated from
reaction data from catalysts prepared under different impregnation pH and calcination
variables were contrasted. Similarly, synergistic properties of bimetallic catalyst during
deactivation-influenced steam reforming were explored using monometallic cobalt and
nickel catalysts as a basis for comparison.
5.5.1 Effect of impregnation pH
The propane conversion profiles for the two specimens (Catalysts L6 and H6) with
different impregnation pH are shown in Fig. 5.19. Evidently, the high-pH catalyst (H6)
demonstrated higher average conversion within the first 3 h, however beyond this point
it rapidly dropped to about half of the final conversion level attained by the low-pH
catalyst (L6) during the 6-h run. On the other hand, while the low-pH catalyst
202
Chapter 5: Bimetallic Catalyst Design
experienced a slight drop in conversion within the first 2 h, but constant conversion was
attained thereafter.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 1 2 3 4 5
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
6
(L6)
(H6)
Fig. 5.19. Propane conversion profiles for acid L6 and basic H6 catalysts.
Thus, the conversion-time data in Fig. 5.19 were fitted into Eq. 5.12 to yield the
coefficients in Table 5.26. Catalyst H6 is characterised by a higher k
d
'-value (32.6 10
-
5
) compared to Catalyst L6 (2.4 10
-5
), where the reaction coefficient, k', for the basic
catalyst (2.84) was also found higher than in the acid catalyst (0.98). These reforming
attributes appear to be associated with the metal distribution profile and active particle
size. It may be recalled that the high-pH catalyst H6 has metal deposition primarily on
the particle exterior (cf. TEM image in Fig. 5.7(b)) and thus, higher propane conversion
would be expected in most of time duration during the initial period (first 4 h reaction
run time) than in the low-pH catalyst L6 with more internally distributed metal sites.
Given the location of the metal sites in the basic catalyst, carbon lay-down (arising from
propane dehydrogenation at low S:C ratio) around the pore mouths would quickly cover
203
Chapter 5: Bimetallic Catalyst Design
active sites as well as blocking access of reactants to inner pores, resulting in substantial
activity loss. In comparison, Catalyst L6 with more uniform metal dispersion (cf. Fig.
5.7(a)) reduces the risk of localised carbon accumulation, particularly in the region near
the pore mouths, giving better accessibility to internal active sites than Catalyst H6,
hence more stable conversion was achieved. In addition, higher carbon formation in the
basic catalyst may also be due to the bigger crystallites associated with this catalyst (cf.
Table 5.9), since larger particle size favours the transformation of surface CH
x
species
to naphthalenic carbon polymers responsible for site blockage and loss in reforming
activity [Dahl et al., 1999; Martnez and Lpez, 2005; Hardiman et al., 2005a].
In general, k' is a reflection of propane conversion due to both reforming and
dehydrogenation reactions. A lower value of k' may be due to lower dehydrogenation
rate yielding less carbon deposition and hence, smaller deactivation coefficient, k
d
'.
Accordingly, the ratio of reaction to deactivation rate constant, k'/k
d
', is a measure of the
carbon resilience of the catalyst and may be used for comparison. Clearly, Catalyst L6
has a higher k'/k
d
' ratio (40,937.7) than Catalyst H6 (8,693.5) conforming that the acid
catalyst possessed superior anti-coking attributes.
Table 5.26. Deactivation and reforming kinetic parameters for acid L6 and basic
H6 catalysts.
Catalyst k
d
' 10
5
(s
-1
) k' (L
2
mol
-1
s
-1
g cat.
-1
) k'/k
d
' (L
2
mol
-1
g cat.
-1
)
L6 2.4 0.98 40,937.7
H6 32.6 2.84 8,693.5
204
Chapter 5: Bimetallic Catalyst Design
5.5.2 Effect of calcination variables
Investigation on the various calcination variables on deactivation and kinetic rate
constants were centred on the low (L) pH specimens with uniform metal distribution.
The Yates assessment presented in Table 5.27 for all the eight bimetallic catalysts
suggests that, at 95% confidence interval, the deactivation coefficient, k
d
', is statistically
influenced by calcination temperature, T, time, t, heating rate, r, 2-factor interactions -
rT & tr, and 3-factor interaction - trT. The effects t and T with negative signs signify
that increase in these two variables would lead to lower k
d
'-value, while the positively-
signed r would bring about the opposite impact. As a result, catalyst with minimum k
d
'-
value may be achieved using preparation variables adopted in L6.
A study by Song et al. [2005] on the coking behaviour of HZSM-5 catalyst suggested
that the intensity of carbon deposition is linearly dependent on the amount of acid sites.
This is consistent with the current study since the effects of calcination variables on the
amount of acid sites (cf. Table 5.13) revealed a similar qualitative trend as the
deactivation rate coefficient (cf. Table 5.27). To some extent, crystallite size may have
also played a role on the deactivation behaviour of the eight catalysts. As carbon-
induced deactivation is favoured with bigger metal clusters, it is not surprising that the
particle size analysis in Table 5.9 shows a similar dependency on r and T. However,
catalysts calcined at shorter duration (low t) with smaller crystallites also yielded higher
k
d
'-values suggesting that acid site density probably has a more dominant contribution.
Consequently, the deactivation rate coefficient may be expressed as a function of
preparation factors via:
205
Chapter 5: Bimetallic Catalyst Design
T r t T r r t T r t k
X X X a X X a X X a X a X a X a a Y
d
6 5 4 3 2 1 0
+ + + + + + =
' (5.13)
with a
0
= 0.000287, a
1
= -0.000036, a
2
= 0.000102, a
3
= -0.00017, a
4
= -0.000234, a
5
=
0.000033 and a
6
= 0.00025.
The Yates analysis performed for the reaction rate coefficient, k', is shown in Table
5.28. It is evident that, at 95% confidence interval, the variables responsible for
controlling k' are calcination temperature, T, heating rate, r (marginally), and the
interaction effects - rT and trT. The sign of the variance estimates suggests that high k'
is favoured by lower calcination temperature, slower heating rate and longer calcination
time. Thus, preparation procedure of Catalyst L2 would offer the highest k'-value.
Taking account of only statistically-relevant terms, the regression model for the reaction
kinetic coefficient is:
T r t T r T r k
X X X a X X a X a X a a Y
4 3 2 1 0
+ + + + =
'
(5.14)
with a
0
= 112.36, a
1
= -77.79, a
2
= -97.75, a
3
= 75.80 and a
4
= 6.26.
A similar statistical analysis for the carbon resilience, k'/k
d
', is displayed in Table 5.29.
As may be seen, the signs of variance estimates suggest that lower calcination
temperature, slower heating rate and longer calcination time (similar to the calcination
variables of Catalyst L2) are favourable for better anti-coking catalysts. However, based
on the statistical F-value, only primary variables - T & r, and 2-factor interaction - rT,
exhibited significant roles (at 95% confidence interval) on the value of k'/k
d
'.
206
Chapter 5: Bimetallic Catalyst Design
Accordingly, the regression model for the anti-coking dependency on catalyst
preparation variables may be derived as:
T r T r k k
X X a X a X a a Y
d
3 2 1 0 /
+ + + =
' ' (5.15)
with a
0
= 403,644.6, a
1
= -272,423.0, a
2
= -301,357.0 and a
3
= 236,142.1.
207
Chapter 5: Bimetallic Catalyst Design
Table 5.27. Values for k
d
' of low-pH (L) catalysts and corresponding Yates
analysis.
Sample
k
d
' 10
5
(s
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 23.1 8 0.000215 Average
L2 30.7 4 -0.000091 t 14.08
L3 38.8 4 0.000064 r 9.91
L4 11.8 4 -0.000054 tr 8.42
L5 17.3 4 -0.000091 T 14.19
L6 2.4 4 0.000006 tT 1.00
L7 25.1 4 0.000079 rT 12.28
L8 23.1 4 0.000119 trT 18.44
208
Chapter 5: Bimetallic Catalyst Design
Table 5.28. Values for k' of low-pH (L) catalysts and corresponding Yates
analysis.
Sample
k'
(L
2
mol
-1
s
-1
g cat.
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 69.12 8 44.33 Average
L2 155.61 4 6.27 t 1.00
L3 54.79 4 -38.32 r 6.11
L4 14.37 4 -23.35 tr 3.72
L5 28.24 4 -58.29 T 9.30
L6 0.98 4 -16.76 tT 2.67
L7 12.63 4 39.47 rT 6.30
L8 18.89 4 40.11 trT 6.40
209
Chapter 5: Bimetallic Catalyst Design
Table 5.29. Values for k'/k
d
' of low-pH (L) catalysts and corresponding Yates
analysis.
Sample
k'/k
d
'
(L
2
mol
-1
g cat.
-1
)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 299,646.4 8 175,790.2 Average
L2 507,642.9 4 24,237.0 t 1.32
L3 141,087.5 4 -154,352.0 r 8.38
L4 121,355.7 4 -18,411.2 tr 1.00
L5 163,637.9 4 -183,285.8 T 9.96
L6 40,937.7 4 -69,895.3 tT 3.80
L7 50,315.1 4 118,071.1 rT 6.41
L8 81,698.6 4 95,453.0 trT 5.18
210
Chapter 5: Bimetallic Catalyst Design
5.5.3 Comparison with monometallic Co and Ni catalysts
The time-on-stream activity behaviour of the monometallic L6Co and L6Ni catalysts
are presented in Fig. 5.20. The L6Co system with an initially high conversion at 70%
relaxed to around 40% following 2 h run and remained stable at this level for the
remaining experimental duration. On the other hand, the conversion of L6Ni reached a
maximum at about 60% after ca. 1 h on stream, followed by a gradual activity loss to
30% at the completion of the run. The conversion levels of both monometallic catalysts
were always higher than the bimetallic catalyst throughout the experiment.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 1 2 3 4 5 6
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
(L6)
(L6Co)
(L6Ni)
Fig. 5.20. Propane conversion profiles for Catalysts L6, L6Co and L6Ni.
Using the reactor performance model in Eq. 5.12, the deactivation and kinetic constants
as well as their ratio were obtained as shown in Table 5.30. Consistent with the
conversion trends in Fig. 5.20, the highly stable L6 catalyst gave lower k
d
' estimate (2.4
10
-5
) than the two monometallic catalysts (3.2 10
-5
and 42.8 10
-5
for L6Co and
211
Chapter 5: Bimetallic Catalyst Design
L6Ni respectively). Interestingly, the reaction rate parameter, k', derived from propane
consumption showed a similar trend where k'-values are higher in L6Co (2.98) and
L6Ni (8.00) compared to L6 (0.98). Indeed, Choudhary et al. [1997, 1998a & 1998b]
have also reported that while the bimetallic reforming Co-Ni gave enhanced stability
due to less carbon deposition, it was accompanied by lower overall conversion.
Although the addition of cobalt has improved the performance of bimetallic L6, it is
interesting to note that the monometallic cobalt (L6Co) catalyst indeed exhibited the
highest k'/k
d
' ratio. The k'/k
d
' ratio decreases in the order of: L6Co (94,282.7) > L6
(40,937.7) > L6Ni (18,668.6), i.e. in the same order of decreasing cobalt content. This
suggests that cobalt may in fact offer a greater advantage over nickel during steam
reforming with low steam feed. This was predominantly attributed to bifunctional
capability of the cobalt species in providing an active site for steam reforming as well as
a site of carbon gasification [van Doorn et al., 1986]. Moreover, cobalt may also form a
surface phase which acts as a barrier to prevent diffusion of carbon through the metal
lattice - a critical step in the formation of coke filament responsible for deactivation
[Tang, 1999].
Table 5.30. Deactivation and reforming kinetic parameters for Catalysts L6, L6Co
and L6Ni.
Catalyst k
d
' 10
5
(s
-1
) k' (L
2
mol
-1
s
-1
g cat.
-1
) k'/k
d
' (L
2
mol
-1
g cat.
-1
)
L6 2.4 0.98 40,937.7
L6Co 3.2 2.98 94,282.7
L6Ni 42.8 8.00 18,668.6
212
Chapter 5: Bimetallic Catalyst Design
The extent of synergy in bimetallic catalyst may be defined as the ratio of the specific
rate of the bimetallic catalyst to the weighted sum of the individual specific rates of the
monometallic constituents from which it was synthesised [Gwaunza and Adesina,
1997], and thus in this case
Ni Co
Ni Co
s s
s
75 . 0 25 . 0 +
=

_
(5.16)
Thus, the synergistic parameter greater than unity (_ > 1, ca. 1.09) lends credence to the
fact that synergistic interplay between the two components gave rise to a bimetallic
system which preserved the high-activity characteristic of nickel as well as maintaining
the anti-coking attribute of cobalt.
5.6 Post-mortem characterisation
5.6.1 Total carbon
Following steam reforming runs, the coked catalysts were subjected to solid sample
total carbon analysis. Table 5.31 presents carbon content and the associated Yates
analysis for spent bimetallic catalysts. In agreement with the higher deactivation rate
coefficient in Catalyst H6, the carbon analysis also confirmed that the basic catalyst has
a coke content of 55.85% while the acid catalyst possessed 44.50%. At 95% confidence
level, calcination time, t, 2-factor interactions - rT, and 3-factor interaction - trT,
exhibited significant effects on the amount of carbonaceous deposits. The variance
estimates further indicated that carbon lay-down may be minimised via judicious
213
Chapter 5: Bimetallic Catalyst Design
combination of longer calcination time, slower heating rate and higher temperature.
This may be further substantiated from the data in Table 5.31 which contained
information on Catalyst L6 prepared under the recommended conditions. Furthermore,
the carbon content in the specimens coked under the reforming conditions may be
related to preparation variables via the linear multivariate expression:
T r t T r t coke
X X X a X X a X a a Y
3 2 1 0 %
+ + + =
(5.17)
with a
0
= 55.78, a
1
= -5.14, a
2
= 2.10 and a
3
= 3.19.
Interestingly, the deactivation rate coefficients, k
d
', evaluated previously paralleled the
trends of coke yields. From Fig. 5.21, the deactivation rate constants in the bimetallic
systems prepared at pH 2 were shown to exhibit linear dependency on the coke content
deposited under the same reforming conditions. While Catalyst H6 has a higher carbon
deposit than Catalyst L6, it is also noted that the basic sample is characterised by a
higher deactivation rate compared to the acid catalyst with equivalent coke deposition.
This was shown in Fig. 5.21 where the data point belonging to H6 was higher than the
correlation line of L-type catalysts.
Table 5.32 summarises the carbon content of monometallic systems. The deactivation
constants of catalysts composing of different metal compositions are, however, not
related to carbon content with a simple linear law. For instance, while incorporation of
cobalt into bimetallic system reduced coke deposition by 15.7%, it was attended by a
corresponding 94.4% decrease in deactivation rate coefficient.
214
Chapter 5: Bimetallic Catalyst Design
Table 5.31. Carbon content of low-pH (L) catalysts and corresponding Yates
analysis.
Sample
Carbon content
(%)
Degrees of
freedom
Variance
estimates
Effect ID
Calculated
F-values
L1 54.76 8 54.14 Average
L2 59.02 4 -4.34 t 6.43
L3 60.00 4 2.84 r 4.21
L4 48.41 4 -2.43 tr 3.60
L5 52.58 4 -2.83 T 4.19
L6 44.50 4 -0.68 tT 1.00
L7 57.88 4 5.53 rT 8.19
L8 55.93 4 5.50 trT 8.14
215
Chapter 5: Bimetallic Catalyst Design
y =200.9x - 87.223
R
2
=0.9369
0
5
10
15
20
25
30
35
40
45
0.40 0.45 0.50 0.55 0.60 0.65
Carbon content
k
d
'

x

1
0
5

(
s
-
1
)
H6
Fig. 5.21. Correlation between k
d
' and carbon content in bimetallic catalysts.
Table 5.32. Carbon contents of monometallic catalysts.
Sample Carbon content (%)
L6Co 10.13
L6Ni 52.78
5.6.2 X-ray diffractions
The effect of carbon deposition on metallic phases of the catalysts prepared with
different metal composition was investigated using XRD measurements as presented in
Fig. 5.22. The carbonaceous deposit in each sample is characterised by a carbon peak at
26. The intensity of this peak reflects the coke quantity and hence, it corroborates the
trend previously revealed via total carbon analysis, in which carbon content decreases in
the order: L6Ni > L6 > L6Co. In addition, the surface metallic phases appeared to
possess good stability during reforming as comparison between freshly-reduced (cf. Fig.
216
Chapter 5: Bimetallic Catalyst Design
5.6) and coked specimens (cf. Fig. 5.22) showed no distinct changes in XRD patterns.
This further highlights the regeneration potential of these catalysts to regain initial
metallic states.
10 20 30 40 50 60
2-theta angle (deg)
I
n
t
e
n
s
i
t
y
NiAl
2
O
4
Co
3
O
4
Co
3
O
4
NiAl
2
O
4
L6 (Co-Ni, pH 2)
L6Co (Co, pH 2)
L6Ni (Ni, pH 2)
NiAl
2
O
4
Ni
Co
Ni
Co
Co
Co
Ni
Ni
NiAl
2
O
4 Carbon
Carbon
CoAl
2
O
4
CoAl
2
O
4
Carbon
Fig. 5.22. XRD patterns for used Catalysts L6, L6Co and L6Ni.
5.6.3 Carbon burn-off runs
The nature and reactivity attributes of carbonaceous deposits over the catalysts with
different metallic composition were examined by temperature-programmed oxidation
(TPO). The curves displayed in Fig. 5.23 indicated final weight losses of 45.71, 12.60
and 66.83% for L6, L6CO and L6Ni respectively in agreement with the previous
sequence in total carbon analysis. As expected, the weight change profiles in the
thermal analyses were also attributed to removal of other impurities and moisture as
well as possible oxidation of metallic species to oxidic phases, thus resulting in larger
net weight drops than the total carbon percents. The L6Co catalyst with lower coke
217
Chapter 5: Bimetallic Catalyst Design
deposition showed modest carbon removal rate, while more rapid weight drops were
observed with increased carbon contents (as in L6 and L6Ni).
20
40
60
80
100
120
0 0.5 1 1.5 2 2.5 3
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
%
)
273
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
(L6Co)
(L6)
(L6Ni)
Fig. 5.23. Transient weight profiles during oxidative carbon burn-off runs of spent
Catalysts L6, L6Co and L6Ni.
Additional information may be obtained from the conversion-time profiles in Fig. 5.24.
Coke removal for all three used catalysts with different coke contents took place in a
similar range of time-on-stream between 0.4-1.4 h (493-793 K) and were characterised
by 3 phases, namely initial induction, reaction-controlled gasification and final
relaxation. It is evident that gasification period (as indicated by linear slopes) tended to
occur at higher temperatures for the catalysts with higher coke residues. This was
particularly an implication of the longer induction step and shorter final relaxation phase
in the more heavily-coked, nickel-containing specimens. Even so, similar slopes of the
conversion curves seem apparent for all three systems.
218
Chapter 5: Bimetallic Catalyst Design
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (h)
o
(L6Ni)
(L6Co)
(L6)
Fig. 5.24. Weight-based conversion profiles during oxidative carbon burn-off runs
of spent Catalysts L6, L6Co and L6Ni.
Consequently, as shown in the weight derivative profiles in Fig. 5.25, the peak
temperatures for each sample are placed at different positions. Single peaks
corresponding to carbonaceous deposits are evident in L6Co and L6 at 680 and 771 K
respectively, while the L6Ni exhibits two peaks at 735 and 798 K. In agreement, a study
by van Doorn et al. [1986] found the TPO peak of carbon gasification at lower
temperature in the case of Co/Al
2
O
3
compared to Ni/Al
2
O
3
. These investigators ascribed
this observation to the characteristic of cobalt as a better catalyst for the C-O
2
reaction.
In addition, the presence of a single carbon peak in cobalt-containing L6Co and L6
specimens than the twin peaks (possessing roughly equal-sized) in L6Ni suggests the
formation of carbonaceous deposits with simpler chemical nature as may be ascribed to
the addition of cobalt.
219
Chapter 5: Bimetallic Catalyst Design
0
0.5
1
1.5
2
2.5
3
373 473 573 673 773 873 973 1073 1173
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t

%
/
m
i
n
Co/Al
297333
O
3
973
(L6Co)
(L6)
(L6Ni)
Fig. 5.25. TPO weight derivative profiles for spent Catalysts L6, L6Co and L6Ni.
As demonstrated previously, the carbon burn-off conversion features of L6 and L6Ni
are characterised by an initiation step that is more time-consuming than final relaxation
phase. Accordingly, the order in the Avrami-Erofeev model (cf. Eq. 5.8) for evaluating
the kinetic coefficient of oxidative carbon burn-off, , is found between 3 to 4
[Brown, 2001]. In light of the R
2
,O reg
k
2
-values, n = 4 is preferred for L6 and L6Ni, while the
L6Co showed a good fit with n = 3 (cf. Table 5.33).
The plots of versus carbon content in Fig. 5.26 indicate similar general features
regardless of model order. The kinetic rate constants during oxidative carbon removal in
particular increased with higher coke accumulation, even though their correlation is
better described by a polynomial than linear expression. As a result, it appears that
2
,O reg
k
220
Chapter 5: Bimetallic Catalyst Design
composition of metallic components brought about a substantial synergism in
determining the nature of coke residues which in turn affects gasification kinetics.
Table 5.33. Kinetic coefficients, , and R
2
,O reg
k
2
-values for oxidative carbon burn-off
runs based on the various orders in the Avrami-Erofeev model.
n in the Avrami-Erofeev model
Sample
2 3 4
L6 2.517 (0.748) 1.878 (0.870) 1.557 (0.928)
L6Co 2.410 (0.966) 1.720 (0.993) 1.369 (0.993)
L6Ni 2.749 (0.811) 2.030 (0.914) 1.669 (0.957)
The numbers in parenthesis are the R
2
-coefficients.
y =1.6514e
0.3497x
R
2
=0.9101
y =2.3316e
0.2553x
R
2
=0.7416
y =1.3049e
0.4382x
R
2
=0.9718
1.0
1.2
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Carbon content
k
r
e
g
,
O
2
(
h
-
1
)
L6
L6Co
L6Ni
2
nd
order
Avrami-Erofeev
3
rd
order
Avrami-Erofeev
4
th
order
Avrami-Erofeev
Fig. 5.26. Correlation between and carbon content.
2
,O reg
k
221
Chapter 5: Bimetallic Catalyst Design
However, the thermogravimetric data of the coke-burn off process may also be used to
obtain surface energetic information of the carbonaceous deposits in terms of pre-
exponential factor, , and activation energy, . In particular, the kinetics of
coke removal reaction is generally analysed in light of mono-layer coke formation on an
acidic solid surface. Subsequently, the reaction may be assumed to follow first-order
with respect to coke; and, hence as demonstrated by Liu et al. [2001 & 2002] who
suggested that the carbon burn-off oxidation kinetics may be described by:
2
,O reg
A
2
,O reg
E
|
.
|

\
|
|
|
.
|

\
|

|
|
.
|

\
|

|
|
.
|

\
|
=
|
.
|

\
|

T R
E
E
T R
E
R A
T
C
g
O reg
O reg
g
O reg
g O reg 1
2
1 ln
ln
ln
2
2 2
2
,
, ,
,
2
|
(5.18)
where C = amount of coke (wt. %),
T = instantaneous temperature (K),
2
,O reg
A = pre-exponential factor (K
2
min
-1
),
R
g
= gas constant,
| = heating rate (K min
-1
),
2
,O reg
E = activation energy of oxidative coke removal (kJ mol
-1
).
The value of is commonly calculated from the slope of the straight line, i.e.
[ /R
2
,O reg
E
2
,O reg
E
g
], of the plot between [ln (-ln C/T
2
)] versus [1/T]. However, it would mean
that the determination of the Arrhenius parameter, , from the intercept [ln
(
2
,O reg
A
2
,O reg
A R
g
/| )(1-(2R
2
,O reg
E
g
T/ ))] is somewhat compromised since it is itself
dependent on temperature, T [Hardiman et al., 2006]. As the Arrhenius parameter,
, is a reflection of the turn-over frequency (TOF) for a first-order reaction, it
2
,O reg
E
2
,O reg
A
222
Chapter 5: Bimetallic Catalyst Design
provides valuable information on the density (turn-over number, TON = TOF
-1
) of the
reacting site. Indeed, Hardiman et al. [2006] recently showed that estimates for the
values of and may be computed via an iterative multilinear regression
technique. Adopting this approach, Eq. 5.18 has been re-written as,
2
,O reg
E
2
,O reg
A
|
.
|

\
|
|
|
.
|

\
|

|
|
.
|

\
|
+
|
|
.
|

\
|
=
|
.
|

\
|

T R
E
E
T R
E
R A
T
C
g
O reg
O reg
g
O reg
g O reg 1
2
1 ln ln
ln
ln
2
2 2
2
,
, ,
,
2
|
(5.19)
and noting that /R
2
,O reg
E
g
T >> 1, then by Taylor approximation;
2 2
, ,
2 2
1 ln
O reg
g
O reg
g
E
T R
E
T R
=
|
|
.
|

\
|

(5.20)
which simplifies Eq. 5.19 to
|
.
|

\
|
|
|
.
|

\
|

|
|
.
|

\
|
=
|
.
|

\
|

T R
E
E
T R
E
R A
T
C
g
O reg
O reg
g
O reg
g O reg 1
2
ln
ln
ln
2
2 2
2
,
, ,
,
2
|
(5.21)
Using transient carbon removal data obtained from thermogravimetric response,
multilinear regression of Eq. 5.21 would yield initial estimates for both and
. These initial estimates may subsequently be substituted back into Eq. 5.19 to
secure a more accurate set of and parameters, which can then be re-
inserted to the same equation and so on until convergence has been reached (typically
2
,O reg
E
2
,O reg
A
2
,O reg
E
2
,O reg
A
223
Chapter 5: Bimetallic Catalyst Design
after 4-5 iterations to achieve a tolerance factor smaller than 1%) to yield final values as
provided in Table 5.34.
Table 5.34. Pre-exponential factors, , and activation energies, , for
oxidative carbon burn-off runs.
2
,O reg
A
2
,O reg
E
Catalyst
Carbon content
(%)
Pre-exponential factor,
2
,O reg
A (K
2
min
-1
)
Activation energy,
2
,O reg
E (kJ mol
-1
)
L6 44.50 68.35 58.6
L6Co 10.13 0.16 9.3
L6Ni 52.78 446.23 67.6
As shown in Fig. 5.27, the coke removal activation energy reveals a linear relationship
to the coke content for all samples despite the different metallic constituents in each
catalyst system. In fact, the -value of the L6 specimen is in a fairly good
agreement with the range (54-58.5 kJ mol
2
,O reg
E
-1
) for the same bimetallic catalysts coked for
10 h duration under S:C ratio = 0.8-1.6 and reforming temperature between 773-873 K
[Hardiman, 2006]. These suggest that coke reactivity is strongly dependent on metallic
composition, while the effect of different carbon content (as a result of varying
operating variables) is not so prominent. The L6Co specimen exhibited a low -
value implying that the oxygen molecules may play another role, in addition to
gasifying carbonaceous residues, in which they reacted with reduced Co metallic sites to
yield surface oxides. In contrast, the correlation plot in Fig. 5.28 indicates that density
of the reacting sites involved during coke removal and carbon quantity were not related
2
,O reg
E
224
Chapter 5: Bimetallic Catalyst Design
by a linear identity. The cobalt-containing systems yielded lower -estimate and
hence, higher TON and density of reacting sites.
2
,O reg
A
y =5.8468e
4.8568x
R
2
=0.9858
0
10
20
30
40
50
60
70
80
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Carbon content
E
r
e
g
,
O
2

(
k
J

m
o
l
-
1
)
L6
L6Co
L6Ni
Fig. 5.27. Correlation between and carbon content.
2
,O reg
E
y =0.0241e
18.316x
R
2
=0.9981
0
50
100
150
200
250
300
350
400
450
500
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Carbon content
A
r
e
g
,
O
2

(
K
2

m
i
n
-
1
)
L6
L6Co
L6Ni
Fig. 5.28. Correlation between and carbon content.
2
,O reg
A
225
Chapter 5: Bimetallic Catalyst Design
Nomenclature
a
0
-a
15
polynomial coefficients in Eq. 5.5
2
,O reg
A pre-exponential factor (K
2
min
-1
)
APS active particle size (nm)
C amount of coke (wt. %)
C
A
concentration of propane (mol L
-1
)
0
A
C initial concentration of propane (mol L
-1
)
2
,O reg
E activation energy for oxidative coke removal (kJ mol
-1
)
F statistical F-value
h pH level
-AH
des
heat of desorption (kJ mol
-1
K
-1
)
k' second-order steam reforming kinetic constant (L
2
mol
-1
s
-1
g cat.
-1
)
k
s,cal
calcination kinetic coefficient (h
-1
)
k
s,red
reduction kinetic coefficient (h
-1
)
k
s,ox
oxidation kinetic coefficient (h
-1
)
k
d
' deactivation rate coefficient in hyperbolic decay law (s
-1
)
2
,O reg
k kinetic rate coefficient for oxidative carbon burn-off (h
-1
)
k
s
kinetic coefficient for processes in thermogravimetric analysis (h
-1
)
n order in the Avrami-Erofeev model
PD percent metal dispersion
r heating rate (K min
-1
)
R
2
statistical R-squared value
226
Chapter 5: Bimetallic Catalyst Design
R
g
ideal gas constant
s specific rate
S
BET
BET surface area (m
2
g
-1
)
S
m
metallic surface area (m
2
g
-1
metal)
t time (h)
T temperature (K)
V
d
amount of NH
3
desorbed in TPD runs (mol g
-1
)
w instantaneous weight
w
f
final weight
w
i
initial weight
X
h
re-scaled variable for impregnation pH-level
X
r
re-scaled variable for calcination heating rate
X
t
re-scaled variable for calcination time
X
T
re-scaled variable for calcination temperature
Y
i
response variable of property i
Greek letters
o conversion in thermogravimetric analysis
| heating rate (K min
-1
)
_ extent of bimetallic synergy
|
W
Wagner modulus
t' gas residence time (g cat. s L
-1
)
227
Chapter 5: Bimetallic Catalyst Design
References
Adesina, A.A. (1996). Hydrocarbon synthesis via Fischer-Tropsch reaction: travails and
triumphs. Appl. Catal. A: Gen., 138, 345-367.
Arnoldy, P. and Moulijn, J.A. (1985). Temperature-programmed reduction of
CoO/Al
2
O
3
catalysts. J. Catal., 93, 38-54.
Auroux, A., Monaci, R., Rombi, E., Solinas, V., Sorrento, A. and Santacesaria, E.
(2001). Acid sites investigation of simple and mixed oxides by TPD and
microcalorimetric techniques. Thermochim. Acta, 379, 227-231.
Bartholomew, C.H. (1975). Reduction of nitric oxide by monolithic-supported
palladium-nickel and palladium-ruthenium alloys. Ind. Eng. Chem. Prod. Res.
Develop., 14, 29-33.
Bartholomew, C.H. and Farrauto, R.J. (1976). Chemistry of nickel-alumina catalysts. J.
Catal., 45, 41-53.
Box, G.E.P., Hunter, W.G. and Hunter, J.S. (1978). Statistics for Experimenters. Wiley,
New York.
Brown, M.E. (2001), Introduction to Thermal Analysis: Techniques and Applications.
Kluwer Academic Publishers, The Netherlands.
Che, M. and Bennett, C.O. (1989). The influence of particle size on the catalytic
properties of supported metals. Adv. Catal., 36, 55-172.
Chen, S.L., Zhang, H.L., Hu, J., Contescu, C. and Schwarz, J.A. (1991). Effect of
alumina supports on the properties of supported nickel catalysts. Appl. Catal.,
73, 289-312.
Choudhary, V.R., Rane, V.H. and Rajput, A.M. (1997). Beneficial effects of cobalt
addition to Ni-catalysts for oxidative conversion of methane to syngas. Appl.
Catal. A: Gen., 162, 235-238.
Choudhary, V.R., Rajput, A.M., Prabhakar, B. and Mamman, A.S. (1998a). Partial
oxidation of methane to CO and H
2
over nickel and/or cobalt containing ZrO
2
,
ThO
2
, UO
2
, TiO
2
and SiO
2
catalysts. Fuel, 77, 1803-1807.
228
Chapter 5: Bimetallic Catalyst Design
Choudhary, V.R. and Mamman, A.S. (1998b). Simultaneous oxidative conversion and
CO
2
or steam reforming of methane to syngas over CoO-NiO-MgO catalyst. J.
Chem. Technol. Biotechnol., 73, 345-350.
Dahl, I.M., Wendelbo, R., Andersen, A., Akporiaye, D., Mostad, H. and Fuglerud, T.
(1999). The effect of crystallite size on the activity and selectivity of the reaction
of ethanol and 2-propanol over SAPO-34. Microporous Mesoporous Mater., 29,
159-171.
Eisenacher, K. and Adesina, A.A. (2000). A statistical evaluation of preparation
conditions on the performance of Ce-promoted Co-Mo Fischer-Tropsch catalyst.
Korean J. Chem. Eng., 17, 71-75.
Geus, J.W. and van Veen, J.A.R. (1999). Preparation of supported catalysts, in: van
Santen, R.A., van Leewen, P.W.N.M., Moulijn, J.A. & Averill, B.A. (Eds.),
Catalysis: An Integrated Approach. Elsevier, Amsterdam, pp. 459-485.
Gonzlez, O., Lujano, J., Pietri, E. and Goldwasser, M.R. (2005). New Co-Ni catalyst
systems used for methane dry reforming based on supported catalysts over an
INT-MM1 mesoporous material and a perovskite-like oxide precursor
LaCo
0.4
Ni
0.6
O
3
. Catal. Today, 107-108, 436-443.
Gwaunza, M. and Adesina, A.A. (1997). The performance of a Ru-Mo sulfide catalyst
for H
2
S decomposition. React. Kinet. Catal. Lett., 62, 55-62.
Hanic, F., Horvath, I., Plesch, G. and Galikova, L. (1985). Study of copper-chromium
oxide catalyst: I. Thermal decomposition of copper(III) chromate, CuCrO
4
. J.
Solid State Chem., 59, 190-200.
Hardiman, K.M., Cooper, C.G. and Adesina, A.A. (2004). Multivariate analysis of the
role of preparation conditions on the intrinsic properties of a Co-Ni/Al
2
O
3
steam
reforming catalyst. Ind. Eng. Chem. Res., 43, 6006-6013.
Hardiman, K.M., Hsu, C.-H., Ying, T.T. and Adesina, A.A. (2005a). The influence of
impregnating pH on the postnatal and steam reforming characteristics of a Co-
Ni/Al
2
O
3
catalyst. J. Mol. Catal. A: Chem., 239, 41-48.
229
Chapter 5: Bimetallic Catalyst Design
Hardiman, K.M., Trujillo, F.J. and Adesina, A.A. (2005b). Deactivation-influenced
propane steam reforming: reactor analysis and parameter estimation. Chem. Eng.
Proc., 44, 987-992.
Hardiman, K.M., Cooper, C.G., Adesina, A.A. and Lange, R. (2006). Post-mortem
characterization of coke-induced deactivated alumina-supported Co-Ni catalysts.
Chem. Eng. Sci., 61, 2565-2573.
Haenen, J., Visscher, W. and Barendrecht, E. (1986). Characterization of NiCo
2
O
4
electrodes for O
2
evolution: part II. Non-electrochemical characterization of
NiCo
2
O
4
electrodes. J. Electroanal. Chem., 208, 297-321.
Heise, M.S. and Schwarz, J.A. (1985). Preparation of metal distributions within catalyst
supports. I. Effect of ph on catalytic metal profiles. J. Colloid Interface Sci., 107,
237-243.
Heracleous, E., Lee, A.F., Wilson, K. and Lemonidou, A.A. (2005). Investigation of Ni-
based alumina-supported catalysts for the oxidative dehydrogenation of ethane
to ethylene: structural characterization and reactivity studies. J. Catal., 231, 159-
171.
Iengo, P., Di Serio, M., Sorrentino, A., Solinas, V. and Santacesaria, E. (1998a).
Preparation and properties of new acid catalysts obtained by grafting alkoxides
and derivatives on the most common supports. Note I - grafting aluminium and
zirconium alkoxides and related sulphates on silica. Appl. Catal. A: Gen., 167,
85-101.
Iengo, P., Di Serio, M., Solinas, V., Gazzoli, D., Salvio, G. and Santacesaria, E.
(1998b). Preparation and properties of new acid catalysts obtained by grafting
alkoxides and derivatives on the most common supports. Part II: Grafting
zirconium and silicon alkoxides on -alumina. Appl. Catal. A: Gen., 170, 225-
244.
Jacobs, G., Das, T.K., Zhang, Y., Li, J., Racoillet, G. and Davis, B.H. (2002). Fischer
Tropsch synthesis: support, loading, and promoter effects on the reducibility of
cobalt catalysts. Appl. Catal. A: Gen., 233, 263-281.
230
Chapter 5: Bimetallic Catalyst Design
Kapteijn, F., Moulijn, J.A. and Tarfaoui, A. (1999). Catalyst characterization and
mimicking pretreatment procedures by temperature-programmed techniques, in:
van Santen, R.A., van Leeuwen, P.W.N.M., Moulijn, J.A. & Averill, B.A.
(Eds.), Catalysis: An Integrated Approach. Elsevier, Amsterdam, pp. 525-541.
Levenspiel, O. (1999). Commentaries: Chemical reaction engineering. Ind. Eng. Chem.
Res., 38, 4140-4143.
Li, C. and Chen, Y.-W. (1995). Temperature-programmed-reduction studies of nickel
oxide/alumina catalysts: effects of the preparation method. Thermochim. Acta,
256, 457-465.
Liu, H., Li, T., Tian, B. and Xu, Y. (2001). Study of the carbonaceous deposits formed
on a Mo/HZSM-5 catalyst in methane dehydro-aromatization by using TG and
temperature-programmed techniques. Appl. Catal. A: Gen., 213, 103-112.
Liu, H., Su, L., Wang, H., Shen, W., Bao, X. and Xu, Y. (2002). The chemical nature of
carbonaceous deposits and their roles in methane dehydro-aromatization on
Mo/MCM-22 catalysts. Appl. Catal. A: Gen., 236, 263-280.
Martnez, A. and Lpez, C. (2005). The influence of ZSM-5 zeolite composition and
crystal size on the in situ conversion of Fischer-Tropsch products over hybrid
catalysts. Appl. Catal. A: Gen., 294, 251-259.
Marturano, M.A., Aglietti, E.F. and Ferretti, O.A. (1999a). Nature of Ni-Al developed
phases during thermal activation in relation to the preparation techniques. Part I:
calcination. Thermochim. Acta, 336, 47-54.
Marturano, M.A., Aglietti, E.F. and Ferretti, O.A. (1999b). Nature of Ni-Al developed
phases during thermal activation in relation to the preparation techniques. Part
II: reduction and catalytic properties. Thermochim. Acta, 336, 55-60.
Mieth, J.A. and Schwarz, J.A. (1989). Effects of alumina dissolution and metal ion
buffering on the dispersion of alumina supported nickel and ruthenium catalysts.
Appl. Catal., 55, 137-149
Mile, B., Stirling, D., Zammitt, M.A., Lovell, A. and Webb, M.J. (1990). TPR studies of
the effects of preparation conditions on supported nickel catalysts. J. Mol.
Catal., 62, 179-198.
231
Chapter 5: Bimetallic Catalyst Design
Molina, R. and Poncelet, G. (1998). -alumina-supported nickel catalysts prepared from
nickel acetylacetonate: a TPR study. J. Catal., 173, 257-267.
Montgomery, D.C. (1997). Design and Analysis of Experiments. John Wiley & Sons,
Inc., New York.
Nguyen, T.H., Yue, E.M.T., Lee, Y.J., Khodakov, A., Adesina, A.A. and Brungs, M.P.
(2003). Synthesis of bimetallic Mo-W carbide from its sulphide precursor via
propane carburization: statistical correlation of the physicochemical properties
with preparation conditions. Catal. Comm., 4, 353-359.
Niemantsverdriet, J.W. (1999). Catalyst characterization with spectroscopic techniques,
in: van Santen, R.A., van Leeuwen, P.W.N.M., Moulijn, J.A. & Averill, B.A.
(Eds.), Catalysis: An Integrated Approach. Elsevier, Amsterdam, pp. 489-524.
Olsbye, U., Wendelbo, R. and Akporiaye, D. (1997). Study of Pt/alumina catalysts
preparation. Appl. Catal. A: Gen., 152, 127-141.
Opoku-Gyamfi, K. (1999). A novel thermally self-sustaining fluidised bed reactor for
hydrogen production from methane. Ph.D. thesis, University of New South
Wales, Sydney, Australia.
Roh, H.-S., Jun, K.-W., Dong, W.-S., Chang, J.-S., Park, S.-E. and Joe, Y.-I. (2002).
Highly active and stable Ni/Ce-ZrO
2
catalyst for H
2
production from methane. J.
Mol. Catal. A: Chem., 181, 137-142.
Rostrup-Nielsen, J.R. (1984). Catalytic steam reforming, in: Anderson, J.R. & Boudart,
M. (Eds.), Catalysis: Science and Technology. Springer-Verlag, Berlin, pp. 1-
117.
Sinfelt, J.H. (1973). Supported bimetallic cluster catalysts. J. Catal., 29, 308-315.
Sinfelt, J.H. (2002). Role of surface science in catalysis. Surf. Sci., 500, 923-946.
Song, Y., Li, H., Guo, Z., Zhu, X., Liu, S., Niu, X. and Xu, L. (2005). Effect of
variations in acid properties of HZSM-5 on the coking behavior and reaction
stability in butane aromatization. Appl. Catal. A: Gen., 292, 162-170.
232
Chapter 5: Bimetallic Catalyst Design
Spiewak, B.E., Cortright, R.D. and Dumesic, J.A. (1997). Thermochemical
characterization, in: Ertl, G., Knzinger, H. & Weitkamp, J. (Eds.), Handbook of
Heterogeneous Catalysis, vol. 2. VCH, Weinheim, pp. 698-706.
Tang, S., Lin, J. and Tan, K.L. (1999). Partial oxidation of methane to synthesis gas
over o-Al
2
O
3
-supported bimetallic PtCo catalysts. Catal. Lett., 59, 129-135.
Van Doorn, J., Verheul, R.C.S., Singoredjo, L. and Moulijn, J.A. (1986).
Characterization of carbon deposits on alumina supported cobalt and nickel by
temperature programmed gasification with O
2
, CO
2
and H
2
. Fuel, 65, 1383-
1387.
Yaluris, G., Larson, R.B., Kobe, J.M., Gonzalez, M.R., Fogash, K.B. and Dumestic,
J.A. (1996). Selective poisoning and deactivation of acid sites on sulfated
zirconia catalysts for n-butane isomerization. J. Catal., 158, 336-342.
233
Chapter 6: Reaction Studies
C
C
h
h
a
a
p
p
t
t
e
e
r
r
6
6
R
R
E
E
A
A
C
C
T
T
I
I
O
O
N
N
S
S
T
T
U
U
D
D
I
I
E
E
S
S
6.1 General considerations
The ultimate performance of a catalytic steam reformer is often limited substantially by
carbon deposition resulting in lower conversion and cessation of flow across the reactor
[Farranto and Bartholomew, 1997]. In many cases, operational interruptions are
frequently necessary as required for catalyst replacement or regeneration process.
Although industries commonly utilise excess steam-to-carbon feed ratio (S:C > 3) in an
attempt to delay coking rate, this practice obviously translates to additional cost-
associated issue to the already intensive energy requirement of the steam reforming
process [Aasberg-Petersen, 2001].
Whilst the previous chapter has considered the design of a bimetallic Co-Ni reforming
catalyst with anti-coking attributes, it was pointed out that catalyst mortality via
hydrocarbon dehydrogenation and/or CO disproportionation is thermodynamically
inevitable. Consequently, in a bid to minimise down-time due to carbon deposition,
some researchers developed reactor operational strategies for admitting online catalyst
decay [Ogunye and Ray, 1971a; Kao and Bankoff, 1975; Crowe, 1976; Szwast and
Sieniutycz, 2001]. Thus, the inclusion of reactor engineering scrutiny in combination
234
Chapter 6: Reaction Studies
with catalyst optimisation would provide a more effective technique in enhancing
overall performance of deactivation-influenced steam reformer [Moulijn et al., 2001;
Birtill, 2003]. Subsequently, more comprehensive steam reforming analyses were
undertaken using Catalyst L6 which represented worst-case carbon resilience attribute
(k'/k
d
') among bimetallic specimens.
In this chapter, reaction and deactivation kinetic estimates were collected at varying
experimental variables during propane steam reforming undergoing concurrent coke-
induced deactivation. These parameters were then used to formulate an on-line reactor
operational policy for maintaining stable exit conversion. Furthermore, the quantity,
property and reactivity of carbonaceous residues as a function of steam reforming
history were thoroughly examined in order to uncover essential information valuable in
the overall optimisation strategy [Birtill, 2003].
6.2 Steady-state kinetics
Initially, steady-state kinetics of propane steam reforming were investigated via power-
law and mechanistic models. The mechanistic analysis would provide fundamental
insights into how the hydrocarbon steam reforming under minimum deactivation occurs.
This approach took into consideration Langmuir-Hinshelwood (LH) and Eley-Rideal
(ER) kinetic expressions with single- and dual-site adsorption-desorption schemes. The
reforming rates in the steady-state runs were determined upon attaining relatively-stable
conversion levels. For the runs at low S:C ratio, the catalyst was expected to lose its
stability rapidly over time. Hence, for these systems the reaction rate used in the steady-
235
Chapter 6: Reaction Studies
state analysis was determined at the reaction time prior to which the conversion starts to
decrease due to carbon-induced deactivation.
6.2.1 Power-law
Figs. 6.1(a) and (b) display the effect of varying steam partial pressure on the rate of
propane steam reforming at temperatures between 773-873 K for low (0.8-1.6) and high
(2.4-4.8) S:C regimes respectively. As expected, steam reforming rate is enhanced at
higher temperature. The presence of excess water was also found to favour steam
reforming reaction.
In particular, propane steam reforming rate data may be described by a power-law
model given by:
b
B
a
A
T R E
SR
p p e k r
g ' '
=
/
0
1
(6.1)
where -r
SR
= rate of propane steam reforming (mol s
-1
g cat.
-1
),
1
0
k
= frequency factor (mol s g cat. kPa ),
-1 -1 -(a'+b')
E = activation energy (kJ mol
-1
),
R
g
= ideal gas constant,
T = temperature (K),
p
A
= partial pressure of propane (kPa),
p
B
= partial pressure of steam (kPa), B
a' = reaction order with respect to propane,
b' = reaction order with respect to steam.
236
Chapter 6: Reaction Studies
0
4
8
12
16
0 20 40 60 80
p
B
(steam), kPa
-
r
S
R
,

m
o
l

s
-
1

g

c
a
t
.
-
1

x

1
0
6
100
773 K
823 K
873 K
T (K) p
A
(propane), kPa
773 15.67
823 16.69
873 17.70
Fig. 6.1(a). Effect of steam partial pressure on steam reforming rate as a function
of temperature at low S:C range (0.8-1.6).
0
2
4
6
8
0 20 40 60 80
p
B
(steam), kPa
-
r
S
R
,

m
o
l

s
-
1

g

c
a
t
.
-
1

x

1
0
6
100
773 K
823 K
873 K
T (K) p
A
(propane), kPa
773 5.84
823 6.22
873 6.59
Fig. 6.1(b). Effect of steam partial pressure on steam reforming rate as a function
of temperature at high S:C range (2.4-4.8).
237
Chapter 6: Reaction Studies
Consequently, linearisation of Eq. 6.1 provides
B A
g
SR
p b p a
T R
E
k r ln ln ) ln( ) ln(
1
0
' + ' + =
(6.2)
from where Arrhenius attributes in terms of activation energy, E, and frequency factor,
1
0
k
, and reaction orders, a' and b', with respect to propane and steam respectively were
evaluated by multiple linear regression of the reaction-composition-temperature data in
Figs. 6.1(a) and (b).
The obtained power-law parameters are summarised in Table 6.1. The E-value for
propane steam reforming was evaluated as 63.2 kJ mol
-1
, which is fairly consistent with
the generalised trend of activation energy given in Praharso et al. [2004] as a function of
carbon number in feed substrate (cf. Fig. 6.2). The positive dependencies of propane
(1.20) and steam (0.82) on the reforming rate imply that the presence of each species
brought about enhancement effect to reforming reaction.
The adequacy of power-law model for extracting essential information from propane
reforming data was appraised by a rigorous error analysis. Fig. 6.3 shows a parity plot
suggesting a fairly good agreement between predicted rate and observed rate with an R
2
-
value of 0.909. Additionally, as the residual plot depicted in Fig. 6.4 evinces no obvious
pattern, hence it can be assumed that power-law model is sufficient [Montgomery,
1997].
238
Chapter 6: Reaction Studies
Table 6.1. Power-law parameters.
Parameter Unit Value
1
0
k mol s
-1
g cat.
-1
kPa
-2.02
5.5410
-5
E kJ mol
-1
63.2
a' (propane order) 1.20
b' (steam order) 0.82
Fig. 6.2. Activation energy for steam reforming of various hydrocarbons (adapted
from Praharso et al. [2004]).
239
Chapter 6: Reaction Studies
0
2
4
6
8
10
12
14
0 2 4 6 8 10 12 14
Observed rate, mol s
-1
g cat.
-1
x 10
6
P
r
e
d
i
c
t
e
d

r
a
t
e
,

m
o
l

s
-
1

g

c
a
t
.
-
1

x

1
0
6
R
2
=0.909
Fig. 6.3. Comparison between predicted rate and observed rate (power law).
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0 2 4 6 8 10
Predicted rate, mol s
-1
g cat.
-1
x 10
6
R
e
s
i
d
u
a
l
,

m
o
l

s
-
1

g

c
a
t
.
-
1

x

1
0
6
12
Fig. 6.4. Residual plot for rate of propane steam reforming (power law).
240
Chapter 6: Reaction Studies
6.2.2 Langmuir-Hinshelwood and Eley-Rideal
The kinetic rate expressions obtained using mechanistic approach were derived based
on the fundamental steps that were suggested to take place during reaction. These steps
typically involve adsorption of reactants, desorption of products and surface reaction.
As outlined in Hardiman et al. [2006a], steam reforming of hydrocarbons may occur via
the following mechanism:
1 1 1
S H C S H C 2S H C
2 1 2 1
+ +
m m m y m x y x
2 2 2 2
S H S OH 2S O H + +
2 1 2 1
S H S O H C S OH S H C
2 1 2 1
+ +
m m m m
1 1 1 1 1 1
S H C S CHO S S O H C
2 1 2 1
+ +
m m m m
2 1 2 1
S H S CO S S CHO + +
2 1 2 1
S H S COO S OH S CO + + (in excess H
2
O)
1 1
S CO S CO + |
1 2 1
S CO S COO + |
2 2 2 2
2S H S H S H + | +
where m
1
, m
2
, x and y are integers and S
1
& S
2
are 2 types of sites. By quasi-steady-state
approximation and incorporating concept of most abundant reactive intermediate, a
range of expressions corresponding to several reaction mechanisms were developed as
summarised in Table 6.2. Model derivations are detailed further in Appendix E.
241
Chapter 6: Reaction Studies
Table 6.2. Langmuir-Hinshelwood (LH) and Eley-Rideal (ER) kinetic models.
No. Model (M) Remarks
1
( )( )
B B A A
B A rxn
p K p K
p p k
+ + 1 1
Dual-site adsorption of propane (A) and steam (B)
2
( ) ( )
B B A A
B A rxn
p K p K
p p k
+ + + 1 1
Dual-site adsorption with non-dissociative water
adsorption
3
( )
2
1
B A rxn
p K p K
p p k
+ +
B B A A
Adsorption of both propane (A) and steam (B) on
the same site with bimolecular rate determining
step
4
A A
B A rxn
p K
p p k
+ 1
Eley-Rideal model
In what follows, nonlinear regressions of the data into the various models given in Table
6.2 were performed to provide estimates for k
rxn
, K
A
and K
B
, as hosted in Table 6.3.
Clearly, all estimates are positive values where all R -coefficients are greater than 0.9. It
is apparent that statistical consideration alone was not sufficient to discriminate the
mechanistic expressions. Consequently, the selection of appropriate rate expression
would require both statistical and thermodynamic scrutinies.
B
2
242
Chapter 6: Reaction Studies
Table 6.3. Estimates of kinetic and reaction constants.
M T (K) k
rxn
( 10
8
) K
A
( 10
3
) K
B
( 10 ) B
3
R
2
773 5.47 198.27 192.16 0.999
823 6.64 0.83 2.16 0.957 1
873 11.25 0.42 0.18 0.958
773 2.39 174.91 0.15 0.972
823 8.22 72.32 1.48 0.943 2
873 17.03 12.90 6.33 0.939
773 10.60 0.16 1.16 0.951
823 19.70 0.55 1.44 0.912 3
873 40.50 1.50 2.66 0.909
773 0.18 46.50 0.916
823 0.65 24.76 0.925 4
873 1.61 7.15 0.959
A thermodynamical criterion was proposed by Boudart et al. [1967] based on
consistency of thermodynamic identities of adsorption coefficients evaluated from a
particular mechanistic model. The associated guideline is known as BMV criterion
which defines that
10 s -AS s 12.2 - 0.0012AH (6.3)
where AS = change in entropy (J mol
-1
K
-1
),
AH = change in enthalpy (J mol
-1
).
243
Chapter 6: Reaction Studies
The parameters AH and AS for each of propane and steam may be obtained from
g g
R
S
T R
H
K
A
+
A
=
1
ln
(6.4)
where K = adsorption constant,
R
g
= ideal gas constant,
T = reaction temperature (K).
The evaluated thermodynamic parameters, AH and AS, for propane (A) and steam (B)
are reported in Table 6.4. Evidently, Models 2 & 3 did not satisfy the thermodynamic
constraint, while Model 4 only barely survived it. In view of compliance to BMV
guideline, only Model 1 seems to provide the most meaningful description of the data.
Indeed, Table 6.5 reveals that Model 1 and its closest rival Model 4 yield activation
energy, E, estimates of 40.0 and 123.4 kJ mol
-1
respectively. In fact, the E-value of
Model 1 is closer to the range accounted by Praharso et al. [2004] for a nickel-based
catalyst (as illustrated in Fig. 6.2). Based on the foregoing discussion, it may therefore
be suggested that the steam reforming of propane involved adsorption of hydrocarbon
and steam on different sites. Nevertheless, lower E-estimates arising from Model 1 and
the previous power-law were attributed to the synergistic effect due to cobalt addition
into the nickel-based system, as well as the feature of enhanced mixing in the fluidised
bed reactor. Ultimately, this bimetallic synergism resulted in more stable activity along
with reduced propensity for deactivation [Hardiman et al., 2006a].
244
Chapter 6: Reaction Studies
Table 6.4. Thermodynamic coefficients and BMV assessment.
M
AH
A
(kJ mol
-1
)
AS
A
(J mol
-1
K
-1
)
AH
B
B
(kJ mol
-1
)
AS
B
B
(J mol
-1
K
-1
)
BMV
criterion
1 -243.23 -347.24 -394.03 -525.49 Yes
2 -145.15 -200.97 212.56 202.34 No
3 125.02 89.26 46.00 2.73 No
4 -104.21 -159.38 Borderline
A: propane; B: steam.
Table 6.5. Arrhenius parameters.
-ln k
rxn
M
T = 773 K T = 823 K T = 873 K
E (kJ mol
-1
)
1 16.7 16.5 16.0 40.0
2 17.6 16.3 15.6 110.7
3 16.1 15.4 14.7 75.0
4 20.1 18.8 17.9 123.4
Other researchers have also successfully employed the mechanistic approach for various
reaction systems, such as methanation during Fischer-Tropsch reaction [Adesina, 1988],
methane steam reforming [Xu and Froment, 1989; Opoku-Gyamfi, 1999] and iso-octane
steam reforming [Praharso et al., 2004]. Yet, transient response technique [Matsumoto
and Bennett, 1978; Kobayashi and Kobayashi, 1974; Chuang et al., 1997; Verykios,
2003] and surface science studies [Somorjai, 1981; Winslow and Bell, 1985; Chuang et
al., 1997; Verykios, 2003; Yang et al., 2005] were also proven valuable in providing
245
Chapter 6: Reaction Studies
insights during the search for a more sensible kinetic mechanism. Accordingly, Adesina
[1986 & 1988] recommended a combination of all these methods, as well as steady-
state analyses and thermodynamic scrutinies, to offer the most reliable means in
determining the final choice of a reaction mechanism.
6.3 Deactivation runs
As governed by thermodynamics, the event of coke-induced deactivation constantly
causes a drawback in the operation of catalytic processes. Hence, kinetic parameters
employed in rate expressions used for modelling, simulation and optimisation purposes
should be estimated under more realistic condition, in which the effect of time-
dependent activity decay are taken into consideration. In this section, it would be
demonstrated how reforming and deactivation rate constants were extracted
simultaneously using conversion-time data collected from deactivation-disguised
propane steam reforming. This was followed by post-mortem characterisation to
examine physical and chemical properties of the coked catalyst. These attributes may
then be correlated with the associated catalytic behaviour at a given operating condition.
6.3.1 Development of coupled reaction-deactivation model
Traditionally, reaction and deactivation kinetic parameters are obtained based on
separate runs performed at idealised conditions. Nevertheless, Krishnaswamy and
Kittrell [1979] emphasised that such an approach is inadequate, particularly in
hydrocarbon-processing operations, where both desired reaction and coke-induced
deactivation occur concurrently. In some cases, the existence of mass transport
limitation may also be significant, thus potentially distorting intrinsic kinetic during rate
246
Chapter 6: Reaction Studies
measurements [Satterfield and Sherwood, 1963; Fogler, 1999]. Following this line,
Levenspiel [1999] recently developed an integrated reaction-deactivation performance
model for a 1
st
order reaction undergoing exponential-type activity decay in a plug flow
reactor where the effect of pore diffusion on kinetic parameters may be incorporated in
term of Wagner modulus, |
W
. In view of extending Levenspiels approach for broader
applications, relevant expressions were derived for a general class of reactions with
different types of activity decay laws for both PFR and CSTR systems [Hardiman et al.,
2005a]. Subsequently, model discrimination may be carried out based on statistical and
reactor engineering considerations.
6.3.1.1 Generalisation of coupled reaction-deactivation model
The reaction rate of propane steam reforming, -r
SR
, may be expressed as,
( )
n m
SR W A
r k C C a t = (6.5)
where multiplication with the activity factor, a, admits reaction rate loss due to coke-
induced catalyst deactivation. Conventionally, the time-dependent activity, a, is defined
as the ratio of the reaction rate at a given time to the rate of the same reaction on a fresh
catalyst [Fogler, 1999]. The rate of activity decay can thus be written as:
p q
d W A
da
k C C a
dt
=
d
(6.6)
247
Chapter 6: Reaction Studies
where k
d
is an intrinsic deactivation coefficient and d is an empirical exponent. The
deactivation order, d, is a reflection on the general type of catalyst decay [Fogler, 1999],
namely linear (d = 0), exponential (d = 1) and hyperbolic (d = 2). In addition, activity
deterioration arising from loss of active sites by coke coverage may be correlated via a
reciprocal relationship with time [Shadman-Yazdi and Petersen, 1972; Corella et al.,
1981; Hardiman et al., 2003]. Thus, catalyst deactivation during hydrocarbon
processing may be described by one of the following decay laws:
t k
d
e a
'
= (exponential) (6.7)
t k
a
d
' +
=
1
1
(hyperbolic)
(6.8)
b
t a

=
0
| (reciprocal) (6.9)
For a plug flow reactor and continuous stirred tank reactor, the pertinent equations are:
0
A A
CSTR
SR
C C
'
r
t

=

(6.10)
A
PFR
SR
dC
'
r
t =

)
(6.11)
Substitution of Eqs. 6.5 and 6.6 into the reactor design equation of either PFR (Eq. 6.10)
or CSTR (Eq. 6.11) yields the expression for the concentration-time profile in terms of
kinetic and deactivation parameters [Levenspiel, 1999].
248
Chapter 6: Reaction Studies
In what follows, Hardiman et al. [2004] showed that, whilst C
W
(steam concentration)
and C
A
(propane concentration) are independent variables, Eq. 6.5 may however be re-
defined as,
a C k r
m
A SR

1
' = (6.12a)
where
n
A
W
C
C
k k
|
|
.
|

\
|
= ' and
1
m n m = + (6.12b)
Thus, k', is a pseudo-rate constant which has dependency on temperature and S:C ratio,
o
C
(i.e. k' = 3
-n
k o
C
n
). Under the same consideration, the transformation of Eq. 6.6
gives:
d
d
a k
dt
da
' = (6.13a)
where
1
3
q
A
p
C d
p
d
C k k o

= ' and
1
q p q = + (6.13b)
Upon integration,
249
Chapter 6: Reaction Studies
| | d
d
t k d a ' + = 1
1
) 1 ( 1 (for d = 1)
(6.14)
whilst,
) exp( t k a
d
' = (for d = 1)
(6.15)
For instance, if it is assumed that a laboratory scale fluidised reactor behaves as a well-
mixed system, Eq. 6.12 and either of Eqs. 6.14 or 6.15 can be introduced into Eq. 6.10
to yield,
| | d
d
m
A
A A
t k d C k
C C
' + '

= '
1
1
) 1 ( 1
1
0
t (for d = 1)
(6.16)
or,
) exp(
1
0
t k - C k
C C
d
m
A
A A
' '

= ' t (for d = 1)
(6.17)
Thus, a fit of the concentration history data to either Eqs. 6.16 or 6.17 would yield
parameter estimates for d, k', m, n, p, q and k
d
' via iterative nonlinear regression
analysis.
Nonlinear regression of these functions with POLYMATH 5.1, however, failed to give
a converged solution regardless of the initial guess for the parameter estimates. These
250
Chapter 6: Reaction Studies
highly-nonlinear systems suggest that the Jacobian matrix of the associated function
partial derivatives with respect to each of the parameters (which are themselves severely
nonlinear as may be seen from Eqs. 6.16 or 6.17) is prone to singularity making
nonlinear regression impracticable [Hardiman et al., 2004]. Consequently, it seems that
data treatment would be more suitably performed via linear regression analysis.
In a recent work, Monzn et al. [2003] proposed that the deactivation of a system with
activity decay to non-zero steady-state level may be given by:
d
r d
a a k
dt
da
) ( ' ' = (6.18a)
In particular, this equation may be employed as an alternative to Eq. 6.6 for the
mathematical process carried out between Eqs. 6.13 to 6.17, and so,
c o
A W d d
C C k k
r
= ' '
(6.18b)
whereupon, the steady-state residual activity, a
r
, is defined as,
|
A W r r
C C k a = (6.18c)
with and k
r
d
k
r
exhibit the usual Arrhenius dependency. Subsequently, this exercise
leads to:
251
Chapter 6: Reaction Studies
{ }
(

' ' + + '

= '

d
d
d
r r
m
A
A A
t k d a a C k
C C
1
1
1
) 1 ( ) 1 (
1
0
t (for d = 1)
(6.19)
where
1
3
c o o
o
A C d d
C k k
r

= ' ' and


1
c o c = +
(6.20a)
1
3
| |
o
A C r r
C k a

= and
1
| = + (6.20b)
or,
| | ) exp( ) - (1
1
0
t k a a C k
C C
d r r
m
A
A A
' ' + '

= ' t (d = 1)
(6.21)
Even so, there is evidently a greater number (9) of kinetic and deactivation parameters
obtained from Eqs. 6.19 and 6.21 compared to those in Eqs. 6.16 and 6.17. Obviously,
this suggests that nonlinear regression analysis using Eqs. 6.19 and 6.21 would require a
more rigorous computational effort attended by similar statistical constraints and a
larger number of data points than collected in this study [Hardiman et al., 2004].
As an example, let us consider a hyperbolic model (d = 2), thus from Eq. 6.16 it was
obtained that,
252
Chapter 6: Reaction Studies
| |
1
) 1
1
0

' + '

= '
t k C k
C C
d
m
A
A A
t
(6.22)
whereupon linearisation yields
t
k
k
k
C
C
d
A
A

1
1
1
0
|
.
|

\
|
' '
'
+
' '
=
|
|
.
|

\
|

t t
(6.23)
Specifically, a fit of the time-varying concentration data to Eq. 6.23 should provide the
estimates for reaction, k', and deactivation, k
d
', rate coefficients.
More step-by-step examples of model development are presented in Appendix F. A full
list of coupled reaction-deactivation performance equations for the various cases are
summarised in Table 6.6. Wagner modulus, |
W
, was incorporated to account for the
influence of pore resistance on the kinetics. Even though, in this study Weisz-Prater
analysis in Section 4.4.2 has previously revealed that pore diffusion resistance had no
significance effect on reaction rate measurements under the adopted reaction conditions.
Ultimately, this would permit the assumption that the effectiveness factor, q, is unity.
However, for cases in which true kinetics are distorted by substantial pore limitation,
the extent of the associated effect may be taken into account in term of Wagner
modulus, |
W
, which is defined as:
253
Chapter 6: Reaction Studies
254
where |
W
= Wagner modulus,

b
= density of fluidising gas (kg m
-3
),
d
p
= average particle diameter (m),
C
A
= measured propane concentration (mol L
-1
),
-r
SR
= measured rate of propane steam reforming (mol s
-1
g cat.
-1
),
D
eff
= effective diffusivity (m
2
s
-1
).
( )
eff
EXP A SR
p b W
D
C r
d
/
=
2
|

(6.24)
255
PFR CSTR
Activity, a,
decay law
n = 0 n = 1 n = 2 n = 0 n = 1 n = 2
Exponential
Table 6.6. Coupled reaction-deactivation models.
t k
d
e a
'
=
t k
k
C C
d A A
ln ) ln( '
|
|
|

| ' '
=
t
W
0
. \
|
t k
C
d
W A
ln ln ln
0
'
|
|
.

\
=
(
(

|
|
.

\
|
k
C
A
| | ' ' ( | | t
t k
C C
d
W A A
ln ln
0
'
|
.

\
=
|
.

|
k 1 1
|
|

| ' '
|
|

|
t
t k C C
d
W
A A
ln ) ln(
0
'
|
|
.

\
=
|
k | | ' 't
t k
k
C
A
ln 1 ln
0
' |
|

| ' '
= |
|

t
C
d
W A
|
.

\
|
.

\
|
t k
k
C C
d
W A A
A
ln ln
2
0
'
|
|
.
|

\
| ' '
=
|
|
.

|
t
C
1 | |
Hyperbolic
t k
a
d
' +
=
1
1 t
k
W d W
1 | | '
+ =
| |
k k C C
A A
0
|
.

\
' ' ' ' t t
t
k
k
k C
W d W
A
ln
1
0
|
.
|

\
|
' '
C
A
'
+
' '
=
|
|
|

| t
|
t
|
. \
t
k
k
k
W d W
1 1
1
|
.
|

\
|
' '
C C
A A 0
'
+
' '
=
|
|

t
|
t
|
|
.

\
t
k k C C
A A
0
|
.

\
' '
k
W d W
1 | | '
+
' '
=
t t
| | t
k
k
k C
W d W
A
1
1
0
|
.
|

\
|
' '
C
A
'
+
' '
=
|
|
|

t
|
t
|
. \
t
k
k
k
C C
W d W
A A
2
|
.
|

\
|
' '
'
+
' '
|
.

\
t
|
t
|
C
A 1
1
0
=
|
|

Reciprocal
b
t a

=
0
|
( ) t b
k
C C
0
|
|

| ' '
=
t |
W
A A
ln ln ln
0 |
.

\
|
t b
k C
A
ln ln ln ln
0 0
|
|

| ' '
= (
(

|
|

| t |
C
W A
|
.

\ (
|
.

\
| t b
k
C C
W A A
ln ln
1 1
ln
0
0

|
|
.
|

\
|
=
|
|
.

|
t |
( )
' ' | |
t b
k
C C
W
A A
ln ln ln
0
0

|
|
.

\
=
|
t | | | ' '
t b
C
W A
A
ln ln 1 ln
0 0

|
|
.

\
=
|
|
.

|
k C | | ' ' | | t |
t b
k
C C
W A A
A
ln ln ln
0
2
0

|
|
.
|

\
| ' '
=
|
|
.

|
t | C
1 | |
where:
0
0
A
A
F
W C
= ' t ;
W,n
1
q
|
= for strong pore diffusion; |
W,n
is Wagner modulus for n
th
order kinetics and
n
A SR
C a k r q ' =
Chapter 6: Reaction Studies
Chapter 6: Reaction Studies
6.3.1.2 Model discrimination
Selection of suitable model was based on temporal data of propane conversion over 10-
h reforming run. Low S:C feed ratio of 0.8 was employed to accelerate coking, thus
ensuring coke lay-down as the primary cause of deactivation over the experimental
duration. Further, since thermodynamic requires a temperature higher than 718 K, the
propane reforming runs were conducted at 773, 823 and 873 K. An upper limit of 923 K
is imposed to avoid the effect of sintering.
A typical propane conversion-time profile is presented in Fig. 6.5. These data were
subsequently fitted into the various coupled reaction-deactivation expressions listed in
Table 6.6 to secure reaction, k', and deactivation, k
d
', constants. The associated
statistical R
2
-values (reflecting the degree of data fitness into a particular model) were
obtained via regression analysis available in Microsoft Office Excel 2000 and are hosted
in Table 6.7. This analysis shows that all deactivation models derived based on
exponential law exhibited R
2
< 0.9, while hyperbolic and reciprocal activity decay
models provided better fits with R
2
= ~0.92 and ~0.97 respectively. This would seem to
suggest that the selection basis between these two decay mechanisms should not rely on
statistical scrutiny alone. Specifically, it lends credence for an additional criterion
involving reaction engineering compliance.
256
Chapter 6: Reaction Studies
0.0
0.2
0.4
0.6
0.8
1.0
0 1 2 3 4 5 6 7 8 9 1
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
0
Fig. 6.5. Typical propane conversion profile. Reaction conditions: temperature,
823 K; S:C feed ratio, 0.8:1.
Table 6.7. R
2
-values for the coupled reaction-deactivation models.
PFR CSTR Activity decay
law n = 0 n = 1 n = 2 n = 0 n = 1 n = 2
Exponential 0.8894 0.8669 0.8474 0.8894 0.8474 0.8317
Hyperbolic 0.9254 0.9220 0.9254 0.9254 0.9254 0.9429
Reciprocal 0.9769 0.9784 0.9777 0.9769 0.9777 0.9758
The deactivation and reaction rate constants for all three temperatures are summarised
in Tables 6.8 and 6.9 respectively. It is apparent that the deactivation coefficient, k
d
' (or
b), decreased with increased temperature irrespective of deactivating mechanism and
reactor type. As shown later in Table 6.10, this would imply that the deactivation
process is marked by a negative temperature coefficient. Indeed, the fact that
257
Chapter 6: Reaction Studies
deactivation rate is slower at higher temperatures is a direct implication of faster
removal of the culprit C
|
species (via gasification) as temperature rises within the
operating conditions used in this study [Bartholomew, 1982 & 2001; Hardiman et al.,
2004]. Meanwhile, reforming rate parameter, k', increased with temperature yielding a
positive coefficient (evinced later in Table 6.11).
258
Chapter 6: Reaction Studies
Table 6.8. Deactivation rate coefficients (k
d
' or b) for various coupled reaction-
deactivation models (s
-1
).
PFR CSTR Activity
decay law n = 0 n = 1 n = 2 n = 0 n = 1 n = 2
773 K 1.626 2.070 2.592 1.626 2.592 3.558
823 K 1.393 1.979 2.716 1.393 2.716 4.039
Exponential
(k
d
' 10
5
)
873K 0.691 1.119 1.697 0.691 1.697 2.703
773 K 2.258 3.167 4.488 2.258 4.488 8.137
823 K 1.782 2.821 4.492 1.782 4.492 9.299
Hyperbolic
(k
d
' 10
5
)
873K 0.797 1.417 2.471 0.797 2.471 5.308
773 K 2.280 2.936 3.713 2.280 3.713 5.147
823 K 2.005 2.887 4.007 2.005 4.007 6.009
Reciprocal
(b 10
1
)
873K 0.898 1.474 2.258 0.898 2.258 3.618
259
Chapter 6: Reaction Studies
Table 6.9. Reaction rate coefficients (k') for various coupled reaction-deactivation
models (units depend on reaction order).
PFR CSTR
Activity
decay law
n = 0
(k'10
6
)
n = 1
(k'10
3
)
n = 2
(k')
n = 0
(k'10
6
)
n = 1
(k'10
3
)
n = 2
(k')
773 K 5.925 3.251 1.843 5.925 4.538 3.476
823 K 7.316 4.470 2.910 7.316 7.170 7.027 Exponential
873K 8.610 5.532 3.915 8.610 10.024 11.669
773 K 6.118 3.425 2.005 6.118 4.938 4.186
823 K 7.432 4.608 3.086 7.432 7.603 8.253 Hyperbolic
873K 8.671 5.638 4.102 8.671 10.500 13.470
773 K 6.363 6.035 6.548 6.363 16.125 40.865
823 K 7.760 9.921 16.50 7.760 40.675 213.202 Reciprocal
873K 8.742 9.014 12.163 8.742 31.138 110.915
260
Chapter 6: Reaction Studies
In a bid to consider reaction engineering implications, the associated deactivation and
steam reforming kinetic constants, k
d
' (or b) and k', at the three temperatures employed
(cf. Tables 6.8 and 6.9) were further scrutinised to evaluate the Arrhenius parameters in
terms of activation energies and Arrhenius constants, as well as relevant correlation
coefficients, as provided in Tables 6.10 and 6.11.
Comparison of reforming activation energy estimates, E, in Table 6.11 showed that PFR
exhibited lower values compared to the CSTR. In fact, the PFR cases were characterised
by E-values less than 45 kJ mol
-1
suggesting the presence of pore diffusion resistance.
This is obviously inconsistent with previous Weisz-Prater analysis (cf. Section 4.4.2)
which confirmed that intraparticle mass transport was indeed negligible. Consequently,
it is apparent that the PFR-based models did not represent the true mechanism, hence
the fluidised bed reactor was more likely to possess a well-mixed behaviour as manifest
in a CSTR system.
Furthermore, these activation energy values were compared against literature values
which were reported in the range 55-70 kJ mol
-1
for Ni catalysts [Rostrup-Nielsen,
1984]. For similar nickel-based systems, the E-versus-carbon number graph shown in
Fig. 6.2 (obtained from Praharso et al. [2004]) gives an E-estimate for propane steam
reforming at approximately 65-70 kJ mol
-1
. Even so, the bimetallic Co-Ni employed in
this research may possibly result in lower activation energy than conventional nickel
catalysts because of synergism attributed by cobalt species [Hardiman et al., 2006a].
From the power-law analysis, the activation energy for the same Co-Ni catalyst
specimen was estimated as 63.2 kJ mol
-1
.
261
Chapter 6: Reaction Studies
Accordingly, these activation energy values are in a better agreement with the estimates
for 2
nd
order CSTR models evaluated as 65.7 and 58.6 kJ mol
-1
in light of hyperbolic
and reciprocal deactivation schemes respectively. Meanwhile, much lower E-values (<
45 kJ mol
-1
) were attained for n = 0 and n = 1. Subsequently, in view of the statistical
R
2
-coefficients between the hyperbolic and reciprocal models (cf. Tables 6.10 and 6.11),
it is manifest that deactivation during low S:C steam reforming may be best described
using hyperbolic law. Low R
2
-values observed in Table 6.10 are probably indicative of
non-linear behaviour of deactivation rate constant, k
d
', with respect to temperature in the
range 773-893 K [Lobo et al., 1973; Bartholomew, 1982 & 2001]. This is especially
evident in the system with low S:C ratio as the effect of catalyst decay is more
prominent.
Based on statistical regression analysis and reactor engineering considerations, temporal
data from propane steam reforming over a Co-Ni catalyst in a fluidised bed reactor are
therefore best represented by a coupled reaction-deactivation CSTR model with 2
nd
order kinetic involving hyperbolic deactivation behaviour. The overall 2
nd
order steam
reforming behaviour was attributed to the fact that propane was simultaneously being
consumed via decomposition (highly possible under low S:C conditions) in addition to
the actual steam reforming reaction. Whilst some literatures often reported 1
st
order
dependency with respect to the hydrocarbon substrate, it is important to note that this
was usually achieved based on data collected under conditions which minimised the
consequence of coking (S:C > 3) [Hardiman et al., 2005a].
262
Chapter 6: Reaction Studies
Table 6.10. Activation energy, Arrhenius coefficients and R
2
-values for
deactivation, k
d
'.
PFR CSTR Activity
decay law n = 0 n = 1 n = 2 n = 0 n = 1 n = 2
E
d
(kJ mol
-1
) -47.3 -33.9 -23.1 -47.3 -23.1 -14.8
ln A
Arrh
-18.3 -16.0 -14.1 -18.3 -14.1 -12.5 Exponential
R
2
0.857 0.776 0.635 0.857 0.635 0.413
E
d
(kJ mol
-1
) -57.7 -44.4 -32.8 -57.7 -32.8 -23.1
ln A
Arrh
-19.6 -17.1 -15.0 -19.6 -15.0 -12.9 Hyperbolic
R
2
0.889 0.830 0.718 0.889 0.718 0.497
E
d
(kJ mol
-1
) -51.4 -37.9 -27.1 -51.4 -27.1 -19.0
ln A
Arrh
-9.4 -7.0 -5.1 -9.4 -5.1 -3.5 Reciprocal
R
2
0.825 0.738 0.603 0.825 0.603 0.425
263
Chapter 6: Reaction Studies
Table 6.11. Activation energy, Arrhenius coefficients and R
2
-values for steam
reforming, k'.
PFR CSTR Activity
decay law n = 0 n = 1 n = 2 n = 0 n = 1 n = 2
E (kJ mol
-1
) 21.0 29.9 42.4 21.0 44.5 68.1
ln A
Arrh
-8.8 -1.1 7.2 -8.8 1.5 11.9 Exponential
R
2
0.999 0.994 0.992 0.999 0.997 0.997
E (kJ mol
-1
) 19.6 28.0 40.3 19.6 42.4 65.7
ln A
Arrh
-9.0 -1.3 7.0 -9.0 1.3 11.7 Hyperbolic
R
2
0.999 0.995 0.993 0.999 0.998 0.997
E (kJ mol
-1
) 17.9 23.2 36.1 17.9 38.2 58.6
ln A
Arrh
-9.2 -1.4 7.7 -9.2 2.0 13.1 Reciprocal
R
2
0.988 0.613 0.467 0.989 0.513 0.394
264
Chapter 6: Reaction Studies
6.3.2 Deactivation-influenced steam reforming analysis
The deactivation-disguised steam reforming experiments were carried out using three
S:C ratios between 0.8-1.6 and three reforming temperatures in the range of 773-873 K.
Fig. 6.6 displays the time-dependent propane conversion profiles for selected runs.
Clearly, steam reforming run conducted at higher S:C and temperature showed more
superior and stable conversion.
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
C
3
H
8

c
o
n
v
e
r
s
i
o
n

(
%
)

S:C=0.8 & T=773 K S:C=0.8 & T=873 K S:C=1.6 & T=773 K S:C=1.6 & T=873 K
Fig. 6.6. Typical propane conversion profiles. Reaction conditions: temperature,
773-873 K; S:C feed ratio, 0.8-1.6.
Figs. 6.7(a)-(d) present the associated product selectivity behaviours. In particular,
constant selectivity curves for H
2
and CO were observed with respect to time-on-stream
after about 1 h run time (cf. Figs. 6.7(a) and (b)). As shown in Fig. 6.7(c), the CO
reforming product was further converted to CO
2
via Boudouard (Eq. 2.7) and water-gas
shift (Eq. 2.8) pathways. In a similar way, from the plot in Fig. 6.7(d), the appearance of
265
Chapter 6: Reaction Studies
CH
4
(a by-product of dehydrogenation reaction, i.e. Eq. 2.6) provides an evidence that
carbon was being formed continually throughout the duration of steam reforming.
Fluctuations of data points for the run at S:C = 0.8 and 773 K were probably an
experimental artefact arising from flow maldistribution across the reactor tube as a
result of severe carbon deposition.
266
Chapter 6: Reaction Studies
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
H
2

s
e
l
e
c
t
i
v
i
t
y

S:C=0.8 & T=773 K
S:C=0.8 & T=873 K
S:C=1.6 & T=773 K
S:C=1.6 & T=873 K
(a)
0
0.1
0.2
0.3
0.4
0.5
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
C
O

s
e
l
e
c
t
i
v
i
t
y

S:C=0.8 & T=773 K
S:C=0.8 & T=873 K
S:C=1.6 & T=773 K
S:C=1.6 & T=873 K
(b)
Fig. 6.7(a)-(b). Transient selectivity profiles for component: (a) H
2
; and (b) CO.
267
Chapter 6: Reaction Studies
0
0.05
0.1
0.15
0.2
0.25
0.3
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
C
O
2

s
e
l
e
c
t
i
v
i
t
y

S:C=0.8 & T=773 K
S:C=0.8 & T=873 K
S:C=1.6 & T=773 K
S:C=1.6 & T=873 K
(c)
0
0.1
0.2
0.3
0.4
0.5
0.6
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
C
H
4

s
e
l
e
c
t
i
v
i
t
y

S:C=0.8 & T=773 K
S:C=0.8 & T=873 K
S:C=1.6 & T=773 K
S:C=1.6 & T=873 K
(d)
Fig. 6.7(c)-(d). Transient selectivity profiles for component: (c) CO
2
; and (d) CH
4
.
268
Chapter 6: Reaction Studies
The concentration-time data for all nine runs at varying S:C and reaction temperatures
(displayed in Appendix G) were fitted into the 2
nd
order CSTR model with hyperbolic
decay law (cf. Table 6.6) in order to obtain k' and k
d
' estimates. In general, the value of
k
d
' decreased with more steam feed and increased temperature (cf. Fig. 6.8).
Accordingly, under the operational conditions of this study, catalyst deactivation rate
was the highest at S:C = 0.8 and 773 K. As will be demonstrated in Section 6.3.3, the
trend shown in Fig. 6.8 is consistent with subsequent post-mortem analysis involving
carbon content measurement, thermogravimetric runs and XRD diffractograms.
0
5
10
15
20
25
0.6 0.8 1 1.2 1.4 1.6 1.8
S:C Ratio
k
d
'

x

1
0
5

(
s
-
1
)
773 K
823 K
873 K
Fig. 6.8. Profile of deactivation coefficient, k
d
'.
The decreasing k
d
' with respect to temperature implies a negative coefficient for the
coke-induced deactivation process. While carbon formation is favoured with increased
temperature, it seems that in the range 773-873 K examined, carbon removal rate by
gasification exceeded the carbon formation rate resulting in lower deactivation rate with
269
Chapter 6: Reaction Studies
increasing temperature since carbon deposition was the primary cause for loss in
catalyst activity. This is illustrated clearly by plots of carbon formation and
hydrogenation versus reciprocal temperature displayed in Fig. 6.9 [Bartholomew, 1982
& 2001]. This author also proposed that coverage of active sites was predominantly
attributed to build up of polymeric C
|
residue, which was formed via
dehydropolymerisation of surface C
o
species. From the corresponding graph, above 700
K (1/T < 1.43 10
-3
K
-1
) the gasification rate of C
|
via hydrogenation is higher than its
production rate (arising from conversion of C
o
), hence minimum deactivation takes
place. Conversely, between 600 and 700 K (1.43 10
-3
K
-1
< 1/T < 1.66 10
-3
K
-1
),
polymerisation rate of C
o
into C
|
exceeds the gasification rate of C
|
. In this temperature
regime, the accumulation of C
o
species also occurs due to slower hydrogenation. Even
so, the C
o
species can be gasified more rapidly below 600 K.
Fig. 6.9. Rates of formation (ln scale) and hydrogenation of C
o
and C
|
versus
reciprocal temperature [Bartholomew, 1982 & 2001].
270
Chapter 6: Reaction Studies
Interestingly, Lobo et al. [1972] have also reported negative coefficient over the
temperature range similar to that in this study, implicating lower carbon deposition rate
with rise in temperature. Their experimental observations are plotted in Figs. 6.10 and
6.11 which show coke deposition rate versus temperature for systems involving reaction
of hydrogen with 1-butene and propylene, respectively.
Fig. 6.10. Arrhenius plot for the rate of deposition on nickel; 1-butene, 100 torr;
hydrogen, 25 torr [Lobo et al., 1972].
271
Chapter 6: Reaction Studies
Fig. 6.11. Arrhenius plots for the rates of deposition from various propylene-
hydrogen mixtures on nickel [Lobo et al., 1972].
The trend of k' values in Fig. 6.12 shows that propane reforming increased with both
S:C ratio and temperature and thus, a positive activation energy was obtained.
272
Chapter 6: Reaction Studies
0
5
10
15
20
25
30
35
40
0.6 0.8 1 1.2 1.4 1.6 1.8
S:C Ratio
k
'

(
L
2

m
o
l
-
1

s
-
1

g

c
a
t
.
-
1
)
773 K
823 K
873 K
Fig. 6.12. Profile of reaction constant, k'.
Furthermore, the interaction between the two independent variables, temperature and
S:C ratio, on k
d
' and k' were investigated via two-way ANOVA statistical analysis. As
revealed in Table 6.12, the calculated F-values for temperature, S:C ratio and their
interaction, are greater than the tabulated values at 95% confidence level [Box et al.,
1978] signifying that deactivation and reaction kinetic parameters are dependent on
each of these factors.
273
Chapter 6: Reaction Studies
Table 6.12. Calculated F-values from two-way ANOVA with error variance as
denominator.
Calculated F-values
a
k
d
' k'
Temperature
19.10
(2,9)
27,677.79
(2,9)
S:C ratio
44.31
(2,9)
43,279.32
(2,9)
Temperature-S:C interaction
8.90
(4,9)
938.87
(4,9)
a
Calculated
ance error vari
iance factor var
= F .
The numbers in brackets are the degrees of freedom for numerator and denominator,
respectively.
F
2,9
= 4.26 at 95% confidence, F
4,9
= 3.63 at 95% confidence. Source: Box et al. [1978].
274
Chapter 6: Reaction Studies
Time-averaged product ratios may be examined in order to understand the nature of
complex behaviour during low S:C hydrocarbon reforming [Hardiman et al., 2004].
Figs. 6.13(a)-(d) illustrate the 3D graphs for various product ratios as a function of S:C
ratio and temperature. The H
2
:CO ratio in Fig. 6.13(a) reveals that steam reforming was
more competitive at high S:C at low end of temperature range used. It was also seen
that, under this operational regime, the H
2
:CO ratio exceeds the stoichiometric threshold
of 2.33 (as stipulated by Eq. 2.5). Accordingly, it points out that either H
2
was being
produced via another route or CO was being consumed by a companion reaction.
Indeed, Fig. 6.13(b) suggests the role of water-gas shift and Boudouard reactions in
transforming CO into CO
2
become increasingly dominant with decreased temperature
as indicated by low H
2
:CO
2
. Evidently, the ratio of CO over CO
2
would provide a more
direct sign on the competitiveness of the CO conversion into CO
2
. This was evidenced
in Fig. 6.13(c) which shows greater CO:CO
2
ratio at high temperature, thus consistent
with thermodynamics. In accordance to Le Chateliers principle, steam-rich reforming
would also shift water-gas shift reaction towards product, thereby reducing CO:CO
2
.
Furthermore, Fig. 6.13(d) would suggest that, at higher temperature, CH
4
was more
susceptible to further decomposition resulting in carbon and hydrogen. As expected, a
lean S:C ratio feed favoured propane dehydrogenation rather than reforming as
evidenced by low H
2
:CH
4
proportion.
275
Chapter 6: Reaction Studies
760
780
800
820
840
860
880
0.8
1
1.2
1.4
1.6
0
10
20
30
40
(a)
Temperature (K)
S:C ratio
Ratio of H
2
:CO
in product
750
800
850
900
0.8
1
1.2
1.4
1.6
0
5
10
15
(b)
Temperature (K)
S:C ratio
Ratio of H
2
:CO
2
in product
750
800
850
900
0.8
1
1.2
1.4
1.6
0
1
2
3
4
(c)
Temperature (K)
S:C ratio
Ratio of CO:CO
2
in product
750
800
850
900
0.8
1
1.2
1.4
1.6
1
2
3
4
5
6
(d)
Temperature (K) S:C ratio
Ratio of H
2
:CH
4
in product
Fig. 6.13. Surface plots of time-averaged product ratios: (a) H
2
:CO; (b) H
2
:CO
2
;
(c) CO:CO
2
; and (d) H
2
:CH
4
.
276
Chapter 6: Reaction Studies
Defining,
ratio C : S =
C
o
(6.25)
100
773
=
T
T
o (6.26)
The various product ratios, y
i/j
, in Figs. 6.13(a)-(d) may be correlated to the two
operating variables via bilinear equations:
C T T C CO H
y o o o o 22 . 12 03 . 0 18 . 15
/
2
+ =
(6.27)
C T T C CO H
y o o o o 29 . 16 34 . 26 19 . 3
2 2
/
+ =
(6.28)
C T T C CO CO
y o o o o 12 . 4 32 . 7 17 . 0
2
/
+ =
(6.29)
C T T C CH H
y o o o o 14 . 1 02 . 5 49 . 1
4 2
/
+ =
(6.30)
The 3D yield plots for each key product component are also presented in Figs. 6.14(a)-
(d). The surface plot of H
2
in Fig. 6.14(a) shows highest value at moderate S:C ratio (ca.
1.2) and mild reaction temperature (ca. 823 K) implying reincorporation of H
2
into
surface carbon (to yield CH
x
) under low S:C and temperature [Hardiman et al., 2004].
As a result, unsaturated CH
x
species may polymerise to form carbonaceous layers,
which gives rise to deactivation. The CO yield in Fig. 6.14(b) reaffirms that the CO
species (produced by main reforming reaction) was further reacted in water-gas shift (at
high S:C and low temperature) and Boudouard (at low temperature) reactions.
Interestingly, the CO
2
yield (cf. Fig. 6.14(c)) appears to display a mirror image of the
277
Chapter 6: Reaction Studies
pattern revealed by the CO yield (cf. Fig. 6.14(b)). This is probably a reflection on the
fact that composition balance between these two closely-associated species is strongly
dependent on reaction parameters. Fig. 6.14(d) indicates that, under steam-rich and
high-temperature environment, CH
4
may itself undergo steam reforming as signalled by
the low CH
4
yield in this region.
By the same token, the ratios of various products per mole C
3
H
8
reacted were correlated
with S:C and temperature as;
C T T C H C H
y o o o o 04 . 3 31 . 4 35 . 1
8 3 2
/
+ =
(6.31)
C T T C H C CO
y o o o o 72 . 0 23 . 1 09 . 0
8 3
/
+ =
(6.32)
C T T C H C CO
y o o o o 28 . 0 16 . 0 37 . 0
8 3 2
/
+ =
(6.33)
C T T C H C CH
y o o o o 17 . 1 04 . 1 66 . 0
8 3 4
/
+ =
(6.34)
278
Chapter 6: Reaction Studies
750
800
850
900
0.8
1
1.2
1.4
1.6
1
2
3
(a)
Temperature (K)
S:C ratio
Ratio of
H
2
:converted C
3
H
8
750
800
850
900
0.8
1
1.2
1.4
1.6
0
0.2
0.4
0.6
0.8
(b)
Temperature (K)
S:C ratio
Ratio of
CO:converted C
3
H
8
760
780
800
820
840
860
880
0.8
1
1.2
1.4
1.6
0
0.2
0.4
0.6
0.8
(c)
Temperature (K) S:C ratio
Ratio of
CO
2
:converted C
3
H
8
760
780
800
820
840
860
880
0.8
1
1.2
1.4
1.6
0
0.5
1
1.5
(d)
Temperature (K)
S:C ratio
Ratio of
CH
4
:converted C
3
H
8
Fig. 6.14. Surface plots of time-averaged ratios per mole C
3
H
8
consumed for
component: (a) H
2
; (b) CO; (c) CO
2
; and (d) CH
4
.
279
Chapter 6: Reaction Studies
6.3.3 Post-mortem catalyst examination
The physicochemical properties of carbon species formed during propane reforming
runs were probed using BET, H
2
chemisorption, total organic carbon (TOC) content
analysis, XRD, TEM, as well as carbon reactivity analyses via gravimetric temperature-
programmed (TPO-TPR and TPR-TPO-TPR) runs. The combination of these
techniques provides insights into the relationship between the nature and quantity of
carbonaceous residues and the respective reaction conditions.
6.3.3.1 General characterisation
Table 6.13 hosts the BET areas, H
2
chemisorptive parameters and TOC values, of
freshly-calcined, freshly-reduced and regenerated catalysts. During reduction of
calcined specimen, cobalt and nickel oxides, as well as corresponding aluminates, are
transformed into metallic Co, Ni, and alumina. As shown, it appears that the
disappearance of oxidic compounds gave rise to larger surface area in the reduced state.
In particular, the data in this table are consistent with deactivation coefficient, k
d
', in
Fig. 6.8 where TOC increased with decreasing temperature and S:C ratio. As illustrated
in Eq. 6.35, the oxidative coke burn-off process is highly exothermic and may indeed
lead to alumina support degradation, aggloromeration of active metal particles and re-
formation of aluminate compounds.
2 2
CO O C +
923
H A = -393.5 kJ mol
-1
(6.35)
280
Chapter 6: Reaction Studies
Accordingly, the amount of heat released is proportional to the initial quantity of carbon
deposit which in turn controls the extent of sintering. Not surprisingly, the loss in
surface area upon regeneration is greater for specimens with lower S:C ratio. This is
also in agreement with H
2
-chemisorption analysis which indicated smaller dispersion,
smaller surface area and larger particle size for regenerated sample with a history of
heavy coke deposition (cf. Table 6.13).
Table 6.13. BET surface areas, H
2
chemisorptive properties and total carbon
contents of freshly-calcined, freshly-reduced and regenerated spent catalysts.
S:C T (K)
S
BET
(m
2
g
-1
)
PD
(%)
S
m
(m
2
g
-1
)
APS
(nm)
Carbon content
(%)
0.8 773 132.39 1.47 1.97 68.60 60.83
0.8 873 123.43 1.69 2.27 59.49 55.15
1.6 773 141.58 3.68 4.92 27.43 15.09
1.6 873 140.54 8.04 10.76 12.54 0.00
Freshly-calcined 118.12 5.62 7.52 17.94 0.00
Freshly-reduced 147.45 - - - 0.00
XRD measurements of the freshly-calcined, freshly-reduced and coked catalysts are
presented in Fig. 6.15. Similarity in peak widths of Co and Ni phases (suggesting
homogenous crystallite size) between freshly-reduced and all four cokes samples prior
to regeneration confirmed that the sintering did not occur during steam reforming, hence
unlikely to be the major cause of deactivation. Significantly, the presence of filamentous
carbonaceous species was manifest as indicated by a carbon peak at 26. The intensity
281
Chapter 6: Reaction Studies
282
of these carbon peaks was directly proportional to TOC level. Furthermore, it can be
seen that full reduction of oxidic phases was not achieved in the fluidised bed reactor
during H
2
activation, nor steam reforming under adopted conditions, since NiAl
2
O
4
and
CoAl
2
O
4
phases remained to exist in all samples. Similarly, steam reforming condition
was also not capable to fully oxidise Ni and Co metallic species because of the presence
of H
2
and CO reducing species as reaction products. Consequently, it seems that steam
reforming environment is composed of a balance between oxidising and reducing
properties [Hardiman et al., 2005b].
The morphology of carbonaceous residues was examined via TEM analysis. For
comparison, the image of a fresh catalyst surface is shown in Fig. 6.16(a). Meanwhile,
Fig. 6.16(b) shows filamentous coke build up as a result of reactor operation at low S:C.
The mechanism of formation for such a filament was illustrated previously in Fig. 2.15.
As shown, the elongation of coke fibre was posited to take place with the metallic
particle as the head, in which the diameter of filament tubes is comparatively equal to
the size of crystallite at the tip (20-30 nm). Observations of carbon residues with similar
attributes have been reported in Kepinski et al. [2000], Moulijn et al. [2001] and Van
Doorn and Moulijn [1990] over nickel- and cobalt-based supported catalysts. Moreover,
the fact that some active metal particles were located on the surface of carbonaceous
layers lends credence to the non-zero level of propane conversion (as illustrated in Fig.
6.6) up to 10 h, even in systems with acute coke lay-down (ca. 55-61%). Nonetheless,
the sample utilised in reforming at S:C = 1.6, displayed in Fig. 6.16(c), generally
seemed to be free from significant coke presence and revealed a relatively similar
feature as the fresh specimen.
283
10 20 30 40 50 60
2-theta angle (deg)
I
n
t
e
n
s
i
t
y
Freshly-reduced
S:C =0.8, T =773 K
S:C =0.8, T =873 K
S:C =1.6, T =773 K
S:C =1.6, T =873 K
NiAl
2
O
4
NiAl
2
O
4
Carbon
Carbon
Carbon
Ni
Co
Ni
Ni
Ni
Co
Co
Co
Ni
Co
Co
Co
Co
Ni
Ni
Ni
NiAl
2
O
4
NiAl
2
O
4
NiAl
2
O
4
NiAl
2
O
4
Ni
Co
Co
Ni
NiAl
2
O
4
NiAl
2
O
4
NiAl
2
O
4
NiAl
2
O
4
CoAl
2
O
4
CoAl
2
O
4
CoAl
2
O
4
CoAl
2
O
4
CoAl
2
O
4
Freshly-calcined
NiAl
2
O
4
NiO
NiAl
2
O
4
CoAl
2
O
4 NiCo
2
O
4
Co
3
O
4
Fig. 6.15. X-ray diffractograms of freshly-calcined, freshly-reduced and coked catalysts.
Chapter 6: Reaction Studies
Chapter 6: Reaction Studies
Fig. 6.16(a)-(b). TEM of Co-Ni catalyst: (a) fresh state; and (b) after reforming at
S:C = 0.8.
284
Chapter 6: Reaction Studies
Fig. 6.16(c). TEM of Co-Ni catalyst: (c) after reforming at S:C = 1.6.
6.3.3.2 Thermogravimetric analysis
The reactivity of carbon deposits was investigated by thermogravimetric method using 2
schemes, namely oxidation (TPO) and reduction (TPR). The reaction between carbon
and oxygen during TPO is expressed as,
2 2
CO O C +
923
H A = -393.5 kJ mol
-1
(6.35)
whereas in TPR, hydrogen reactant is used,
4 2
CH 2H C +
923
H A = -74.5 kJ mol
-1
(6.36)
285
Chapter 6: Reaction Studies
The corresponding TPO and TPR profiles for the four coked catalysts studied are
presented in Figs. 6.17 and 6.18. Despite containing higher coke content, it appears that
catalysts collected from low S:C runs may be rejuvenated at a faster rate (as indicated
by steeper slopes). This behaviour was possibly due to the fact that the coke deposit
from reforming under a particular reaction condition exhibits a unique reactivity. As
pointed out by Kepinski et al. [2000], the rate of carbon removal can indeed, be related
via a simple function with the rate of deposition. Even so, the more lightly coked
specimens attained complete conversion within a shorter run time. Moreover, the
corresponding slopes (indicative of coke gasification rate) demonstrated that carbon was
more rapidly removed under TPO rather than TPR.
20
30
40
50
60
70
80
90
100
110
120
0 1 2 3 4 5 6
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
w
t

%
)
373
473
573
673
773
873
973
T
e
m
p
e
r
a
t
u
r
e

(
K
)
(a)
(c)
(d)
(b)
Fig. 6.17. TPO weight profiles of spent catalysts under reaction S:C and
temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and (d) 1.6
and 873 K.
286
Chapter 6: Reaction Studies
20
30
40
50
60
70
80
90
100
110
120
0 1 2 3 4 5 6
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
w
t

%
)
373
473
573
673
773
873
973
T
e
m
p
e
r
a
t
u
r
e

(
K
)
(a)
(d)
(c)
(b)
Fig. 6.18. TPR weight profiles of spent catalysts under reaction S:C and
temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and (d) 1.6
and 873 K.
The resulting weight changes during TPO and TPR are summarised in Table 6.14.
Evidently, the weight loss percents (ascribed to coke gasification) are consistent with
TOC and XRD observations. Further, the lower weight drops attained in TPR compared
to TPO would seem to suggest that some carbon species seemed to be more resistant to
H
2
than O
2
.
287
Chapter 6: Reaction Studies
Table 6.14. Weight change analysis of spent catalysts after different thermal
treatments.
Reforming conditions Weight loss (%)
Catalyst
S:C T (K) TPO TPR
(a) 0.8 773 60.95 59.65
(b) 0.8 873 57.32 55.15
(c) 1.6 773 14.91 19.45
(d) 1.6 873 2.50 4.10
Nevertheless, it is possible that the observed weight changes during TPO of coked
catalysts may be due to metal oxidation (resulting from the presence of O
2
) of the
reduced metallic phase(s). In order to assess the extent of metal oxidation, theoretical
weight drop percentages were evaluated for two extreme cases: (1) no metal oxidation
occurred, and (2) all metal phases were re-oxidised in conjunction with coke
gasification [Hardiman et al., 2006b]. In the same way, H
2
-TPR may also be attended
by the reduction of metallic oxide, thus contributing to the observed weight losses. For
this purpose, the two extreme cases considered were: (1) no reduction of metal oxide
occurred, and (2) all metal oxides were reduced during reductive carbon removal
[Hardiman et al., 2006b]. The weight change estimates calculated based on these
conjectures are hosted in Table 6.15.
The TPO experimental results, shown in Table 6.14, reveal that weight changes were
closer to the estimates proposed according to Case 1, hence re-oxidation of metallic
phase(s) appeared to yield no significant contribution to weight change during TPO
288
Chapter 6: Reaction Studies
regeneration run. Using the same reasoning, TPR weight losses in the heavily-coked
catalysts (S:C = 0.8) are also more comparable with Case 1, therefore weight behaviours
were not substantially affected by reduction of oxide phase(s). This was supported by
the proposition that a fraction of carbon residue remained on the surface being unable to
react with H
2
, thus providing a barrier which hindered reduction of oxide components.
In fact, for the catalysts with lighter carbon deposition (S:C = 1.6), weight loss estimates
of Case 2 are in a better agreement with the experimental data. This suggests that, in the
catalysts containing minimum coke deposit, the hydrogen species used as a coke
gasification agent would also play an important role in facilitating the reduction of
oxide species.
Table 6.15. Predicted weight changes of spent catalysts with and without effects
resulting from bulk phase oxidative/reductive reactions upon carbon gasification.
Weight loss (%)
TPO TPR Catalyst
Case 1 Case 2 Case 1 Case 2
(a) 60.83 58.64 60.83 63.01
(b) 55.15 52.65 55.15 57.65
(c) 15.09 10.36 15.09 19.82
(d) 0 5.57
a
0 5.57
a
Weight gain
The difference in coke reactivity towards air (TPO) and H
2
(TPR) would suggest the
existence of multiple carbon species possessing distinct physicochemical attributes.
289
Chapter 6: Reaction Studies
This was substantiated by TPO weight derivative plots in Fig. 6.19(a), which shows 2
peaks - a sharp one at lower temperature (650-700 K) and a broader, shoulder peak at
higher temperature (700-750 K), even though precise peak location is strongly
dependent on steam reforming history. According to Bartholomews nomenclature,
these peaks corresponded to atomic C

and polymeric C

carbon species respectively


[Bartholomew, 1982 & 2001].
In order to understand the nature of metallic phases following carbon deposition and
subsequent removal, a H
2
-TPR analysis was immediately performed at the conclusion
of TPO. The reduction derivative spectra for the four coked specimens under study, as
well as the fresh catalyst, are displayed in Fig. 6.19(b). The TPR curve of the fresh
system depicts two principal peaks which are ascribed to a solution of Ni and Co oxides
(560-730 K), and metal aluminate compounds (850-970 K), respectively. For the more
heavily-deposited catalysts, metallic oxide and aluminate species suffered severe
deterioration during the highly-exothermic carbon burn-off using oxygen in air. During
this process, an enormous energy was released in form of heat, which resulted in
sintering and hence, loss of metallic sites. In this regard, for the four coked catalysts
being investigated, the comparison of TPR peak intensities demonstrated consistency
with the previous H
2
-chemisorptive data (in Table 6.13). Furthermore, metallic phase
compositions in Fig. 6.19(b) may be used as the basis to predict the activity of
rejuvenated catalyst.
290
Chapter 6: Reaction Studies
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
423 523 623 723 823 923 1023
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t
%
/
m
i
n
923
(a)
(b)
(c)
(d)
Fig. 6.19(a). TPO weight derivative profiles of spent catalysts under reaction S:C
and temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and
(d) 1.6 and 873 K.
291
Chapter 6: Reaction Studies
423 523 623 723 823 923 1023
Temperature (K)
923
Fresh
(a)
(b)
(c)
(d)
Fig. 6.19(b). TPR (following TPO) weight derivative profiles of fresh and spent
catalysts under reaction S:C and temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873
K; (c) 1.6 and 773 K; and (d) 1.6 and 873 K.
292
Chapter 6: Reaction Studies
Following the same line, weight derivative plots were also derived for the TPR runs
(based on Fig. 6.18). As shown in Fig. 6.20(a), the corresponding TPR derivative
profiles of coked catalysts (a) and (b) were marked by a sharper peak at lower
temperature range (700-900 K) and a shoulder peak at higher range (> 900 K).
However, in the samples (c) and (d) with less coke deposits, the two peaks appeared to
merge, forming a single entity.
Subsequently, the disparity in weight drops between TPO and TPR runs of the catalysts
containing heavy carbon lay-down (refer to Table 6.14) demonstrates that carbonaceous
phases deposited under the various reforming conditions may exhibit unique chemical
attributes (C
1-z
H
z
ranging from pure carbon, z = 0, to simple hydrogen adatoms, z = 1)
providing an implication to different reactivity with respect to O
2
and H
2
. These are
further detailed as,
( ) O H
2
CO 1 O
4
2
H C
2 2 1
|
.
|

\
|
+
|
.
|

\
|
+

z
z
z
z z
(6.37a)
and
( )
4 2 1
CH 1 H
2
5 4
H C z
z
z z

|
.
|

\
|
+

(6.37b)
respectively [Hardiman et al., 2006b].
293
Chapter 6: Reaction Studies
As the previous regeneration scheme implemented a combined TPO-TPR technique, in
this case the H
2
-TPR run was followed by TPO (to gasify remaining carbon residue
resistant to hydrogen) and subsequently, a final TPR was conducted to examine the
resulting solid-state phase composition - a series of TPR-TPO-TPR. This technique was
performed particularly as an effort to explore an alternative regeneration route for
allowing complete carbon removal, while the associated risk of sintering is lower than
that of TPO-TPR. Figs. 6.20(b) and (c) display the TPO (following TPR) and TPR
(following TPO-TPR) weight derivative curves respectively.
Recalling the pathway of coke formation, the deactivation-causing C
|
layer was
originated from polymerisation of atomic C

pool. In Fig. 6.20(b), the oxidation of the


remnant carbonaceous deposit arising from the hydrogenation of C
|
pool (to C

) in the
first H
2
-TPR was signalled as a shoulder TPO peak at about 600-700 K, while the larger
peak at 700-850 K corresponds to the composite metal oxide phase [Hardiman et al.,
2006b]. Thus, it is not surprising that the first peak was not visible in samples with
lower coke concentration (curves (c) and (d)) as the C
|
carbon phase was practically
insignificant in these catalysts.
In comparison with the TPO-TPR set (cf. Fig. 6.19(b)), two primary peaks were also
observed at identical temperature ranges in the final H
2
-TPR of the TPR-TPO-TPR
series (presented in Fig. 6.20(c)). The first peak between 560-730 K (representing a
solution of Ni and Co oxides) varied in the symmetry and shape between catalysts as an
indication of phase composition difference. However, the peak areas (a measure of
concentration) seem to be relatively similar in all four catalysts, hence active metal
294
Chapter 6: Reaction Studies
dispersion was seemingly independent of thermal history. Significantly, it would seem
to suggest that the extent of sintering was not prominent during H
2
-reactivation
(compared to oxidation) due to much lesser heat release. From previous, the amounts of
energy produced for oxidative (Eq. 6.35) and reductive (Eq. 6.36) coke gasification are
given as 393.5 and 74.5 kJ mol
-1
, respectively. Accordingly, catalysis rejuvenation via
initial treatment with H
2
-TPR followed by TPO would be expected to yield a superior
activity recovery.
295
Chapter 6: Reaction Studies
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
423 523 623 723 823 923 1023 1123 1223 1323 1423 1523 1623 1723 1823 1923
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t
%
/
m
i
n
923
(c)
(d)
(b)
(a)
Fig. 6.20(a). TPR weight derivative profiles of spent catalysts under reaction S:C
and temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c) 1.6 and 773 K; and
(d) 1.6 and 873 K.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
423 523 623 723 823 923 1023
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t
%
/
m
i
n
923
(b)
(d)
(a)
(c)
Fig. 6.20(b). TPO (following TPR) weight derivative profiles of spent catalysts
under reaction S:C and temperature of: (a) 0.8 and 773 K; (b) 0.8 and 873 K; (c)
1.6 and 773 K; and (d) 1.6 and 873 K.
296
Chapter 6: Reaction Studies
423 523 623 723 823 923 1023
Temperature (K)
923
Fresh
(a)
(b)
(c)
(d)
Fig. 6.20(c). TPR (following TPR-TPO) weight derivative profiles of fresh and
spent catalysts under reaction S:C and temperature of: (a) 0.8 and 773 K; (b) 0.8
and 873 K; (c) 1.6 and 773 K; and (d) 1.6 and 873 K.
297
Chapter 6: Reaction Studies
6.3.3.3 Kinetic investigation of coke burn-off
Thermogravimetric coke burn-off data for the representative four samples were
quantitatively scrutinised in order to relate coke reactivity with reforming parameters.
This analysis would lead to the evaluation of pre-exponential factor, , activation
energy, , and kinetic coefficient, , for oxidative carbon removal (TPO). The
value of was computed according to the methodology of Liu et al. [2001 &
2002] which has been modified by Hardiman et al. [2006b] to allow the estimation of
- as outlined in Eqs. 5.18-5.21. Using the same approach as previous, the solid-
state kinetic parameter, , may also be calculated via Avrami-Erofeev (A) model
with n = 3 (cf. Eq. 5.8) [Brown, 2001]. The relevant values of , and
are summarised in Table 6.16.
2
,O reg
A
2
,O reg
E
2
,O reg
k
2
,O reg
E
2
,O reg
A
2
,O reg
k
2
,O reg
A
2
,O reg
E
2
,O reg
k
The -values reveal that the activation energy for oxidative regeneration over the
bimetallic Co-Ni catalyst fell in a narrow range between 54-58.5 kJ mol
2
,O reg
E
-1
. Much lower
estimate of the pre-exponential factor, , (a measure of the turn-over frequency,
TOF) in the catalyst used at 873 K lends credence to the fact that the main carbon
residue, C
2
,O reg
A

, is about 4-5 times less reactive than the C

pool formed at 773 K [Hardiman


et al., 2006b]. The gasification kinetic parameters of the spent sample obtained from
reaction run with S:C = 1.6 and 873 K (i.e. catalyst (d)) were not evaluated since the
corresponding solid contained negligible coke lay-down.
Liu et al. [2001 & 2002] investigated the activation energy of coke burn-off over
Mo/MCM-22 catalysts with varying metal loading. On the 2% Mo catalyst, the
298
Chapter 6: Reaction Studies
observed activation energy was found to be 127 kJ mol
-1
, however, a lower value of
35.4 kJ mol
-1
was obtained as Mo concentration increased to 10%. The high activation
energy estimate was comparable to the burning-off of an acid solid which possesses a
characteristic E-value of 114 kJ mol
-1
[Liu et al., 2002], thus suggesting that the site of
carbon lay-down in the low metal-loading catalyst was mainly centred on the acidic
support. On the other hand, carbonaceous residue was deposited predominantly on the
metallic sites or Mo
2
C species on the catalyst with high metal loading. Consequently,
the low apparent activation-energy (54-58.5 kJ mol
-1
) of the Co-Ni system (which in
fact contained high metal loading at 20% Co+Ni) supported on a basic Al
2
O
3
material
was a clear evidence that carbon was mostly yielded on the metallic Co/Ni sites.
Prior to the determination of oxidative coke burn-off kinetic coefficients, , it was
necessary to first identify the significance of any external mass transfer resistance. This
was quantitatively assessed using the Weisz-Prater criterion [Fogler, 1999], as given in
Eq. 4.9. Table 6.17 summarises the parameters used in the calculation. These
parameters were specifically chosen to simulate the worst possible cases during
oxidation and reduction. Subsequently, the LHS of Eq. 4.9 was evaluated as 1.752 10
2
,O reg
k
-
3
and 5.562 10
-5
for oxidative and reductive carbon gasification respectively,
signifying negligible external mass transfer resistance in both regeneration processes.
Moreover, this also eliminates the possibility of mass-transfer limitation as underlying
mechanism responsible for the low -estimate.
2
,O reg
E
From Table 6.16, the value of increased in the range 2.9-12 h
2
,O reg
k
-1
with decreasing
coke content. This dependency may be described by a power-law as given by,
299
Chapter 6: Reaction Studies
98 . 0
,
2

= C k k
carbon O reg
(6.38)
where k
carbon
= reactivity coefficient of carbonaceous deposit (h
-1
),
C = coke content (%).
which may then be re-written as:
C k k
O reg carbon
~
2
,
(6.39)
The expression given in Eq. 6.39 is in conformity with the finding of Kepinski et al.
[2000] which proposed that carbon removal rate is related to deposition rate by means
of a simple correlation. Furthermore, based on the data from the representative coked
catalysts, the associated value of k
carbon
was computed as 172.44 h
-1
.
Table 6.16. Pre-exponential factors, , activation energies, , and kinetic
constants, , of the coke burn-off process for catalysts coked under various
conditions.
2
,O reg
A
2
,O reg
E
2
,O reg
k
Pre-exponential factor,
(K
Activation energy,
(kJ mol
Kinetic constant,
(h
Catalyst
2 -1 -1 -1
2
,O reg
A
2
,O reg
E
2
,O reg
k min ) ) )
(a) 160.2 58.5 2.9
(b) 36.2 54.0 3.6
(c) 170.9 58.3 12.0
300
Chapter 6: Reaction Studies
Table 6.17. Parameters used in the Weisz-Prater analysis.
Gasifying agent
Parameter Remarks
O H
2 2
A
r'
(mol g solid
-1
s
-1
)
6.986 10
-5
1.861 10
-5
Observed gas consumption rate at
maximum carbon gasification
4 4 Upper limit
p
(g cm
-3
)
R
(cm)
4.705 10
-2
4.705 10
-2
Maximum radius of coked particles
assuming 70% coke content and
= 0.5
p,coked p, fresh
D
eff
(cm
2
s
-1
)
1.273 10
-1
4.488 10
-1
Estimated as 0.1 D
AB
where D
AB
is
the molecular diffusivity with pair
of gases of O -N and H -N
2 2 2 2
for
TPO and TPR respectively and was
calculated using the method of Bird
et al. (1960)
C
A
(mol L
-1
)
2.773 10
-3
6.602 10
-3
Calculated from ideal gas relation
at P = 1 atm and T = 923 K for
21% O and 50% H reactant gases
2 2
301
Chapter 6: Reaction Studies
6.4 Reactor optimisation
Traditionally, the determination of deactivation and reforming kinetic parameters
required for reactor optimisation was based on separate experimental runs [Grubecki
and Wjcik, 2001; Ogunye and Ray, 1971a & b]. However, during deactivation-
influenced steam reforming, main reforming reaction and coke-induced deactivation are
two inseparable processes in which one of the products or reactants is a precursor
species for catalyst decay [Hardiman et al., 2006c]. Nevertheless, as demonstrated
earlier in this chapter, the extraction of deactivation and reforming rate coefficients
from the same set of temporal conversion data can be achieved using a technique
suggested by Levenspiel [1999], which has been expanded for a more general case in
Hardiman et al. [2005a].
6.4.1 Theoretical considerations
The optimum operational strategy was developed on the basis of minimising
deterioration in conversion level due to simultaneous catalyst decay with time-on-
stream [Krishnaswamy and Kittrell, 1979; Szwast and Sieniutycz, 2001]. This was
performed by implementation of a pre-determined reaction temperature profile.
Mathematical framework used in the derivation of the optimal temperature-time relation
was provided in Hardiman et al. [2006c]. This is also detailed in the following.
Assuming there is no change in reaction mechanism over modest temperature range, the
activity, a, at any time, t, is given by;
) ( ) ( ) (
0
T k T k t a =
(6.40)
302
Chapter 6: Reaction Studies
where T = temperature at t = 0 (K).
0
Accordingly, this policy implies that the reaction temperature must be raised to T, in
order to ensure that the deactivated reaction rate is equal to the fresh reaction rate.
The kinetic constant, k, is dependent on temperature as defined in Arrhenius expression,
and so
(
(

|
|
.
|

\
|

=
T T R
E
a
g
1 1
exp
0
(6.41)
Accordingly, this can be re-arranged to yield
( )
0
1
ln
1
T
a
E
R
T
g
+ =
(6.42)
Recalling Eq. 6.13a,
d
d
a k
dt
da
' = (6.13a)
which upon integration, subsequent substitution into Eq. 6.42 gives:
1
0
1 ln
1 1

(

|
.
|

\
|
+ =
G
t
H T
T (6.43a)
303
Chapter 6: Reaction Studies
where H and G are defined as
g
d
R
E d E
H
) 1 (
=
(6.43b)
] ) 1 ( [
0
E d E k
E
G
d d

=
(6.43c)
respectively. In what follows, Eqs. 6.43 describe the time-dependent optimal
temperature policy which can be imposed to minimise conversion loss in a catalytic
reactor undergoing carbon-induced deactivation. In particular, parameters E, E
d
and
corresponding to uncontrolled runs may be obtained from multilinear regression of Eqs.
6.44 and 6.45 (which were originated from Eqs. 6.13 and 6.12 respectively), namely:
0
d
k
C
g
d
d d
p
T R
E
k k o ln
1
) ln( ln
1
+ = '
(6.44)
and,
C
g
n
T R
E
k k o ln
1
) ln( ln
0
+ ' = '
(6.45)
Note that is essentially the deactivation coefficient, k
0
d
k ', at temperature T = T .
d 0
304
Chapter 6: Reaction Studies
6.4.2 Model implementation
Consequently, substitution of deactivation and reforming rate parameters (cf. Figs. 6.8
and 6.12) into Eqs. 6.44 and 6.45 respectively gave various parameters in Table 6.18.
Deactivation process during reforming in the range 773-873 K was marked by a
negative coefficient (E
-1
= -83.6 kJ mol
d
) suggesting lower carbon accumulation with
increased temperature. On the other hand, steam reforming exhibits a positive
coefficient (E = 50.2 kJ mol
-1
). Negative order, p, of -1.494 with respect to steam-to-
carbon ratio signifies the fact that deactivation was effectively inhibited by steam
addition since it promoted carbon removal by gasification. Conversely, it is not
surprising that the reaction order associated with steam reforming, n, reveals a positive
dependency on steam-to-carbon ratio as reflected by an order of 1.070.
Table 6.18. Deactivation and reaction parameters of propane steam reforming.
Parameter Value
1
d
k (s
-1
)
-10
2.586 10
-1
E ) (kJ mol -83.6
d
p -1.494
Correlation coefficient 0.899
k
0
' (L
2 -1 -1
mol s g cat.
-1
) 1.487 10
4
-1
E (kJ mol ) 50.2
n 1.070
Correlation coefficient 0.950
305
Chapter 6: Reaction Studies
It is of strategic economic significance to be able to carry out steam reforming at the
lowest-possible S:C ratio without substantial carbon deposition. Such an operation,
however, still requires the reactor to run at a stoichiometrically-feasible condition.
Consequently, a S:C ratio of 1 was selected as the lower limit to test the optimum
dynamic temperature control policy. Meanwhile, the upper S:C was chosen at a slightly
higher value of 1.5. The temperatures employed (773 and 813 K) were the lower section
of the range used in the uncontrolled reforming runs (773-873 K). This was done since
it was expected that temperature would increase with time during optimum policy runs.
Following this line, the values of (i.e. k
0
d
k ' at temperature T = T
d 0
) for various sets of
initial temperature, T
0
, and S:C were computed by substitution of these variables, along
with other coefficients in Table 6.18, back to Eq. 6.44. The corresponding -estimates
for the four cases of dynamic-temperature experiments are given in Table 6.19.
0
d
k
Table 6.19. Estimate of for the four cases of optimal operational policy (s
-1
0
d
k ).
S:C Initial temperature, T
0
(K)
0
d
k 10
5
1.0 773 11.480
1.0 813 6.055
1.5 773 6.264
1.5 813 3.304
In Section 6.3.1.2, it has been suggested that carbon-induced catalyst deactivation
during propane steam reforming under low S:C ratio may be best described via a
306
Chapter 6: Reaction Studies
hyperbolic-type decay (i.e. second-order, d = 2). Taking this into account, as well as
activation energy estimates, E and E
d
, (from Table 6.18) and (from Table 6.19) into
Eqs. 6.43a-c leads to the temperature-time relation required to maintain propane
conversion level for a given initial temperature, T
0
d
k
0
, and S:C ratio. The proposed
temporal temperature profiles for the four cases at various T
0
and S:C are presented in
Fig. 6.21. In general, the temperature curves show a rapid increase in the first 10 h and
eventually a nearly-linear slope beyond 15 h reaction time. Additionally, it was noted
that lower S:C would require faster temperature ramping. This is consistent on the basis
that high temperature thermodynamically favours coke gasification process, thus
creating a more effective regeneration environment functional to remove heavy carbon
residues expected under the poor S:C condition. A practical upper limit of 1000 K was
implemented in order to avoid melt down of thermally-sensitive rig components (e.g.
reactor furnace).
773
793
813
833
853
873
893
913
933
953
973
0 5 10 15 20 25 30
Time-on-stream (h)
C
a
t
a
l
y
s
t

b
e
d

t
e
m
p
e
r
a
t
u
r
e

(
K
)
S:C=1.0 & To=773 K
S:C=1.0 & To=813 K
S:C=1.5 & To=773 K
S:C=1.5 & To=813 K
Fig. 6.21. New reactor operational policy for various S:C ratio (1.0 and 1.5) and
initial temperature cases (773 and 813 K).
307
Chapter 6: Reaction Studies
Figs. 6.22(a)-(d) display the conversion behaviour under the controlled furnace
temperature. From Figs. 6.22(a)-(b), dynamic runs performed at S:C = 1 showed a rapid
increase in conversion, which peaked at 5 h run time, followed by sharp decrease and
subsequent stabilisation after 10 h. The sharp gain in the first 5 h was a direct
implication of the high rate temperature rise as formulated in the temperature-time
policy. This was then counteracted by a drop to level similar to the commencement of
experiment. These conversion bumps were probably due to the fact that the parameters,
E, E
d
and , utilised in the derivation of optimal policy were determined from time-
averaged uncontrolled reaction data over 10 h, while in reality these parameters would
also change with time-on-stream. However, as reaction time exceeds 10 h, conversion
curves appeared to stabilise to non-zero level until the conclusion of experiment at 30 h.
0
d
k
As shown in Figs. 6.22(c)-(d), the optimal policies of the two experiments at S:C = 1.5
suggested a less rapid initial temperature rise. Near full conversion of propane was
achieved in the two cases after 5 h, which was maintained until about 20 h run time.
Subsequently, activity started to deteriorate slowly towards the end of run.
308
Chapter 6: Reaction Studies
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20 25 30
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
773
793
813
833
853
873
893
913
933
953
973
C
a
t
a
l
y
s
t

b
e
d

t
e
m
p
e
r
a
t
u
r
e

(
K
)
(a)
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20 25 30
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
773
793
813
833
853
873
893
913
933
953
973
C
a
t
a
l
y
s
t

b
e
d

t
e
m
p
e
r
a
t
u
r
e

(
K
)
(b)
Fig. 6.22(a)-(b). Transient propane conversion profiles under dynamic operational
policy with S:C ratio and initial temperature of: (a) 1.0 and 773 K; and (b) 1.0 and
813 K.
309
Chapter 6: Reaction Studies
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20 25 30
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
773
793
813
833
853
873
893
913
933
953
973
C
a
t
a
l
y
s
t

b
e
d

t
e
m
p
e
r
a
t
u
r
e

(
K
)
(c)
0.0
0.2
0.4
0.6
0.8
1.0
0 5 10 15 20 25 30
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
773
793
813
833
853
873
893
913
933
953
973
C
a
t
a
l
y
s
t

b
e
d

t
e
m
p
e
r
a
t
u
r
e

(
K
)
(d)
Fig. 6.22(c)-(d). Transient propane conversion profiles under dynamic operational
policy with S:C ratio and initial temperature of: (c) 1.5 and 773 K; and (d) 1.5 and
813 K.
310
Chapter 6: Reaction Studies
The coke contents for the optimum runs are presented in Table 6.20. This result reflects
the previous trend shown in the unforced reaction runs (cf. Table 6.13) implying that
less carbon was obtained with increased temperature and S:C ratio. Though, for S:C = 1,
the catalyst collected from run at T
0
= 813 K contained more coke yield (64.31%) than
the specimen at lower T = 773 K (62.47%).
0
It was pointed out in Bartholomew [1982 & 2001] that considerable temperature rise, as
well as the longer experimental duration, may have contributed to the conversion of C
|
into the less reactive C
C
, which has faster formation rate relative to removal rate with
increasing temperature.
c
C C C
(6.46)
Specifically, this author also highlighted that deactivation of steam reforming catalysts
due to encapsulation of active metal surfaces by C
C
films notably takes place at > 873
K. From Fig. 6.21, it is clear that all four temporal temperature policies required the
reactor to be heated above 873 K (with different crossing point from run to run).
Though, it would appear that C
C
may have been favoured specifically in systems with
S:C = 1, in which the reaction temperatures were eventually ramped to higher levels
than the reforming at S:C = 1.5. Furthermore, the magnitude of coke content among the
four specimens was shown to reveal an opposite sequence to the level of final
conversion. This inverse dependency may be correlated with a linear expression as
presented in Fig. 6.23.
311
Chapter 6: Reaction Studies
Table 6.20. Total carbon analysis and final conversion percents following 30 h run
time under dynamic temperature operation.
Initial temperature, T
S:C
0
(K)
Final propane
conversion (%)
Carbon content (%)
1.0 773 62.47 54
1.0 813 64.31 50
1.5 773 59.29 64
1.5 813 53.27 76
y =-2.3844x +2.0367
R
2
=0.9902
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
0.50 0.52 0.54 0.56 0.58 0.60 0.62 0.64 0.66
Carbon content
F
i
n
a
l

p
r
o
p
a
n
e

c
o
n
v
e
r
s
i
o
n
Fig. 6.23. Correlation between final propane conversion of dynamic policy runs
with carbon content.
In order to monitor the shifts of balance between various reactions during unsteady-state
runs, a range of time-varying product ratios in reactor exit stream are provided in Figs.
6.24(a)-(d). The corresponding time-averaged product ratios are given in Table 6.21. As
312
Chapter 6: Reaction Studies
seen in Fig. 6.24(a), initial H
2
:CO ratios significantly decreased with time. In particular,
reforming under low S:C and higher temperature gave a lower H
2
:CO ratio (i.e. reduced
propensity for propane steam reforming). However, it is noticed that in all four cases, a
H
2
:CO range between 3-6 was finally attained as time-on-stream reaches 20 h. Even so,
this still greatly exceeded the stoichiometric value (2.33).
It seemed that both water-gas shift and Boudouard reactions played an important role at
low temperature. As temperature was ramped up in accordance with the optimum
policy, the two reactions became thermodynamically less favourable, thus decelerating
the conversion of CO into CO . This is in agreement with the increasing H :CO
2 2 2
and
CO:CO
2
with respect to time, as depicted in Figs. 6.24(b) and (c), which implicated
slower generation rate of CO . Similarly, Fig. 6.24(a) reveals that H
2 2
:CO ratios
decreased with time due to a rise in CO concentration.
Fig. 6.24(d) shows that the H :CH
2 4
ratio rose with time-on-stream. Within the first 10 h,
the H
2
:CH
4
ratio increased rather gently with values maintained roughly between 1.8 to
2.5 suggesting that propane dehydrogenation was the more dominant reaction.
However, as temperature increased, the steam reforming reaction became more
significant and thus the H production rate from both C
2 3
H and intermediate CH
8 4
led to
rapid rise in H :CH ratio.
2 4
313
Chapter 6: Reaction Studies
0
5
10
15
20
25
30
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

H
2
:
C
O

i
n

p
r
o
d
u
c
t
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(a)
0
20
40
60
80
100
120
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

H
2
:
C
O
2

i
n

p
r
o
d
u
c
t
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(b)
Fig. 6.24(a)-(b). Transient profiles under dynamic operational policy for product
ratios: (a) H
2
:CO; and (b) H :CO
2 2
.
314
Chapter 6: Reaction Studies
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

C
O
:
C
O
2

i
n

p
r
o
d
u
c
t
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(c)
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

H
2
:
C
H
4

i
n

p
r
o
d
u
c
t
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(d)
Fig. 6.24(c)-(d). Transient profiles under dynamic operational policy for product
ratios: (c) CO:CO ; and (d) H :CH
2 2 4
.
315
Chapter 6: Reaction Studies
Table 6.21. Time-averaged concentration ratios of H :CO, H
2 2
:CO , CO:CO
2 2
and
H
2
:CH
4
.
T S:C (K) H :CO H :CO CO:CO H :CH
0 2 2 2 2 2 4
1.0 773 9.72 23.42 6.11 8.90
1.0 813 4.65 31.90 8.89 18.04
1.5 773 14.79 6.78 1.16 5.80
1.5 813 8.52 10.46 2.34 4.12
Time-dependent curves of various product yields are also presented in Figs. 6.25(a)-(d).
In addition, Table 6.22 summarises the time-averaged product yields for the four
operating conditions. The temporal H
2
yield, presented in Fig. 6.25(a), indeed exhibited
an increasing pattern in agreement with the stoichiometric dictation of the two most
important H
2
-producing reactions - propane steam reforming (Eq. 2.5) and
dehydrogenation (Eq. 2.6).
Figs. 6.25(b) and (c) confirm that the concentration balance between CO and CO
2
shifted during temperature-varying runs. The CO yield in the outlet gas increased while
CO decreased. Since CO
2 2
is formed from either water-gas shift or Boudouard reaction
(both of which are favoured at low temperature), it is logical that the increased
temperature with time will reduce CO production.
2
Although propane dehydrogenation is thermodynamically favoured at higher
temperature (thus producing more CH , i.e. Eq. 2.6), Fig. 6.25(d) further implicates that
4
316
Chapter 6: Reaction Studies
increased temperature at a greater extent would also facilitate further conversion of CH
4
via steam reforming.
317
Chapter 6: Reaction Studies
0
2
4
6
8
10
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

H
2
:
c
o
n
v
e
r
t
e
d

C
3
H
8
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(a)
0
0.4
0.8
1.2
1.6
2
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

C
O
:
c
o
n
v
e
r
t
e
d

C
3
H
8
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(b)
Fig. 6.25(a)-(b). Transient profiles under dynamic operational policy of component
ratios per mole C H
3 8
converted for species: (a) H ; and (b) CO.
2
318
Chapter 6: Reaction Studies
0
0.2
0.4
0.6
0.8
1
1.2
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

C
O
2
:
c
o
n
v
e
r
t
e
d

C
3
H
8
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(c)
0
0.4
0.8
1.2
1.6
2
0 5 10 15 20 25 30
Time-on-stream (h)
R
a
t
i
o

o
f

C
H
4
:
c
o
n
v
e
r
t
e
d

C
3
H
8
S:C=1.0 & T=773 K
S:C=1.0 & T=813 K
S:C=1.5 & T=773 K
S:C=1.5 & T=813 K
(d)
Fig. 6.25(c)-(d). Transient profiles under dynamic operational policy of component
ratios per mole C H
3 8
converted for species: (c) CO ; and (d) CH .
2 4
319
Chapter 6: Reaction Studies
Table 6.22. Time-averaged ratios of H
2
, CO, CO
2
and CH per mole C H
4 3 8
converted.
T S:C (K) H /C H conv. CO/C H conv. CO /C H conv. CH /C H conv.
0 2 3 8 3 8 2 3 8 4 3 8
1.0 773 2.88 0.60 0.29 0.61
1.0 813 2.47 0.63 0.15 0.56
1.5 773 2.74 0.44 0.43 0.77
1.5 813 2.87 0.51 0.41 0.81
6.4.3 Post-mortem characterisation
The weight percent profiles of oxidative coke burn-off using thermogravimetric unit are
displayed in Fig. 6.26. The magnitudes of weight losses were consistent with the TOC
values (listed in Table 6.20). Nevertheless, the slopes of the gasification curves
appeared to exhibit identical steepness indicating similar coke removal rate in all four
used catalysts.
As shown in Fig. 6.27, weight derivative curves of the coke gasification process evinced
identical shapes in which carbon peaks were seemingly to be dominated by naphthalenic
C pool with temperature peak at about 700-750 K. The C

species were also observed
as small humps at lower temperature between 650-700 K. As the reforming runs
conducted at S:C = 1 were ramped up to higher temperature during dynamic
experiments, their weight derivative profiles revealed a shoulder peak in the range 800-
850 K, which was ascribed to C
C
carbon [Bartholomew, 1982 & 2001]. The presence of
this carbonaceous species was manifest as a result of high-temperature steam reforming
in conjunction with longer reaction run time. Though it may intuitively seem that the C
C
320
Chapter 6: Reaction Studies
peak of curve (b) was somewhat shorter than curve (a), the catalyst treated at S:C = 1
and T
0
= 813 K suffered from higher degree of total coke deposition (cf. Table 6.20). In
addition, this particular run in fact yielded the greatest conversion loss upon 30 h
duration among all four samples.
The estimates of pre-exponential factor, , activation energy, , and kinetic
coefficient, , for the TPO profiles are provided in Table 6.23. The -values
between 56-61 kJ mol
2
,O reg
A
2
,O reg
E
2
,O reg
k
2
,O reg
E
-1
were comparable to the range reported previously for the same
Co-Ni catalyst coked during uncontrolled steam reforming (54-58.5 kJ mol
-1
).
Nevertheless, it would appear that the coke properties, and , have inverse
dependency on operating temperature, in which these two parameters gave lower values
with increased temperature. On the other hand, they have a positive dependency on S:C
ratio.
2
,O reg
E
2
,O reg
A
321
Chapter 6: Reaction Studies
0
20
40
60
80
100
120
0 1 2 3 4 5 6
Time (h)
W
e
i
g
h
t

p
e
r
c
e
n
t

(
w
t

%
)
373
473
573
673
773
873
973
1073
T
e
m
p
e
r
a
t
u
r
e

(
K
)
(b)
(c)
(a)
(d)
Fig. 6.26. TPO weight profiles of spent catalysts obtained from optimal policy
dynamic runs at S:C and T
0
of: (a) 1.0 and 773 K; (b) 1.0 and 813 K; (c) 1.5 and
773 K; and (d) 1.5 and 813 K.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
423 523 623 723 823 923 1023
Temperature (K)
d
(
w
e
i
g
h
t

p
e
r
c
e
n
t
)
/
d
(
t
)
,

w
t
%
/
m
i
n
923
(c)
(a)
(b)
(d)
Fig. 6.27. TPO weight derivative profiles of spent catalysts obtained from optimal
policy dynamic runs at S:C and T
0
of: (a) 1.0 and 773 K; (b) 1.0 and 813 K; (c) 1.5
and 773 K; and (d) 1.5 and 813 K.
322
Chapter 6: Reaction Studies
Table 6.23. Pre-exponential factors, , activation energies, , and kinetic
constants, , of the coke burn-off process for catalysts used in optimal policy
dynamic runs.
2
,O reg
A
2
,O reg
E
2
,O reg
k
Pre-exponential factor,
(K
S:C T
0
(K)
2
,O reg
2
,O reg
2
,O reg
Activation energy,
(kJ mol
Kinetic constant,
(h
2 -1 -1 -1
A E k min ) ) )
1.0 773 694.8 58.3 1.4
1.0 813 501.8 56.3 1.4
1.5 773 1268.7 61.0 1.5
1.5 813 975.7 59.7 1.5
Nomenclature
a activity
a' reaction order with respect to propane
a residual activity
r
2
,O reg
A pre-exponential factor (K
2 -1
min )
A Arrhenius constant
Arrh
APS active particle size (nm)
b deactivation exponent in the reciprocal decay law
b' reaction order with respect to steam
C coke content (%)
C concentration of reactant gas A (mol L
-1
)
A
323
Chapter 6: Reaction Studies
0
A
C initial concentration of reactant gas A (mol L
-1
)
C concentration of steam (mol L
-1
)
W
d deactivation order
d particle diameter (m)
p
D
AB
molecular diffusivity (cm
2 -1
s )
D
eff
effective diffusivity (m
2 -1
s or cm
2 -1
s )
-1
E activation energy for steam reforming (kJ mol )
-1
E activation energy for deactivation (kJ mol )
d
-1
2
,O reg
E activation energy for oxidative coke removal (kJ mol )
F statistical F-value
0
A
F initial flow rate of reactant gas A (mol s
-1
)
G as defined in Eq. 6.43c
H as defined in Eq. 6.43b
AH propane enthalpy change (J mol
-1
)
A
AH
B
steam enthalpy change (J mol )
-1
B
-1
AH amount of heat absorbed or released in a reaction at temperature T (kJ mol )
T
n+m
k intrinsic reaction rate constant (mol s
-1
g cat.
-1
(L mol
-1
) )
k' steam reforming kinetic constant ) cat. g s L mol (
1 1 1
1 1
m m
k
0
' as defined in Eq. 6.45 ) cat. g s L mol (
1 1 1
1 1
m m
frequency factor in power-law model (mol s g cat. kPa )
-(a'+b') -1 -1
1
0
k
k
carbon
reactivity coefficient of carbonaceous deposit (h
-1
)
k
d
intrinsic deactivation rate coefficient (s
-1
)
k deactivation rate coefficient in exponential and hyperbolic decay laws (s
-1
)
d
'
324
Chapter 6: Reaction Studies
k deactivation rate coefficient in Eq. 6.18b (s
-1
)
d
''
0
d
k deactivation rate coefficient at temperature T = T
0
(s
-1
)
1
d
k as defined in Eq. 6.44 (s
-1
)
r
d
k intrinsic deactivation constant ) s L mol (
1
1 1
c c
k intrinsic residual deactivation constant ) s L mol (
1
1 1

r
2
,O reg
k kinetic rate coefficient for oxidative carbon burn-off (h
-1
)
k
rxn
steam reforming kinetic constant in M1, M2, M3 and M4 as shown in Table 6.2
(unit depends on type of model)
k kinetic coefficient for processes in thermogravimetric analysis (h
-1
)
s
K propane adsorption constant
A
K
B
steam adsorption constant B
m, n, p, q reaction orders
m integer as defined in
2 1
H C
m m
1
m integer as defined in
2 1
H C
m m
2
p
A
partial pressure of propane (kPa)
p
B
partial pressure of steam (kPa) B
P pressure (atm)
PD percent metal dispersion
-r'
A
consumption rate of reactant gas A during coke gasification (mol g solid
-1
s
-1
)
-r
SR
rate of propane steam reforming (mol s
-1
g cat.
-1
)
R particle radius (cm)
R
2
statistical R-squared value
R ideal gas constant
g
325
Chapter 6: Reaction Studies
-1
AS
A
propane entropy change (J mol K
-1
)
AS
B
steam entropy change (J mol K )
-1 -1
B
S
BET
BET surface area (m
2
g
-1
)
S
m
metallic surface area (m
2
g
-1
metal)
t reaction run time (s)
T temperature (K)
T temperature at t = 0 (K)
0
W weight of catalyst (g cat.)
X propane conversion
y
i/j
product ratio of i to j
Greek letters
S:C ratio o
C
dimensionless temperature o
T
|
0
empirical parameter in the reciprocal decay law (s
b
)
o, c, |, reaction orders
q effectiveness factor
Wagner modulus for n
th
order kinetics |
W,n

b
density of catalyst bed (kg m
-3
)
particle density (g cm
-3
)
p
t' gas residence time (g cat. s L
-1
)
326
Chapter 6: Reaction Studies
References
Aasberg-Petersen, K., Bak Hansen, J.-H., Christensen, T.S., Dybkjaer, I., Seier
Christensen, P., Stub Nielsen, C., Winter Madsen, S.E.L. and Rostrup-Nielsen,
J.R. (2001). Technologies for large-scale gas conversion. Appl. Catal. A: Gen.,
221, 379-387.
Adesina, A.A. (1986). Application of periodic operation to the Fischer-Tropsch
synthesis of hydrocarbons. Ph.D. thesis, University of Waterloo, Ontario,
Canada.
Adesina, A.A. (1988). Tools for mechanistic investigation: application of the BMV-KD
criterion to the methanation reaction. J. Nig. Soc. Chem. Eng., 7, 258-263.
Bartholomew, C.H. (1982). Carbon deposition and steam reforming. Catal. Rev. Sci.
Eng., 24, 67-112.
Bartholomew, C.H. (2001). Mechanisms of catalyst deactivation. Appl. Catal. A: Gen.,
212, 17-60.
Bird, R.B., Stewart, W.E. and Lightfoot, E.N. (1960). Transport Phenomena. John
Wiley and Sons, Inc., New York.
Birtill, J.J. (2003). But will it last until the shutdown? Deciphering catalyst decay!
Catal. Today, 81, 531-545.
Boudart, M., Mears, D.E. and Vannice, M.A. (1967). Kinetics of heterogenous catalytic
reactions. Industrie Chimique Belge., 32, 281-284.
Box, G.E.P., Hunter, W.G. and Hunter, J.S. (1978). Statistics for Experimenters. Wiley,
New York.
Brown, M.E. (2001), Introduction to Thermal Analysis: Techniques and Applications.
Kluwer Academic Publishers, The Netherlands.
Chuang, S.S.C., Brundage, M.A. and Balakos, M.W. (1997). Mechanistic study in
catalysis using dynamic and isotopic transient infrared spectroscopy:
CO/H
2
/C
2
H
4
reaction on Mn-Rh/SiO
2
. Appl. Catal. A: Gen., 151, 333-354.
327
Chapter 6: Reaction Studies
Corella, J., Asua, J.M. and Bilbao, J. (1981). Kinetics of catalyst deactivation. Can. J.
Chem. Eng., 59, 647-648.
Crowe, C.M. (1976). Optimization of reactors with catalyst decay and the constant
conversion policy. Chem. Eng. Sci., 31, 959-962.
Farranto, R.J. and Bartholomew, C.H. (1997). Fundamentals of Industrial Catalytic
Processes. Blackie, London.
Fogler, H.S. (1999). Elements of Chemical Reaction Engineering. Prentice Hall
International, Inc., New Jersey.
Grubecki, I. and Wjcik, M. (2001). Analytical solution of the Euler-Lagrange equation
in an optimization problem for a batch reactor with deactivating catalyst. Chem.
Eng. Sci., 56, 6617-6621.
Hardiman, K.M., Mohammed, M.M. and Adesina, A.A. (2003). Deactivation kinetics of
cobalt-nickel catalysts in a fluidised bed reformer, in: Jackson, S.D., Hargreaves,
J.S.J. & Lennon, D. (Eds.), Catalysis in Application. The Royal Society of
Chemistry, Cambridge, UK, pp. 16-23.
Hardiman, K.M., Tan, T.Y., Adesina, A.A., Kennedy, E.M. and Dlugogorski, B.Z.
(2004). Performance of a Co-Ni catalyst for propane reforming under low steam-
to-carbon ratios. Chem. Eng. J., 102, 119-130.
Hardiman, K.M., Trujillo, F.J. and Adesina, A.A. (2005a). Deactivation-influenced
propane steam reforming: reactor analysis and parameter estimation. Chem. Eng.
Proc., 44, 987-992.
Hardiman, K.M., Hsu, C.-H., Ying, T.T. and Adesina, A.A. (2005b). The influence of
impregnating pH on the postnatal and steam reforming characteristics of a Co-
Ni/Al catalyst. J. Mol. Catal. A: Chem., 239, 41-48.
2
O
3
Hardiman, K.M., Hsu, C.-H. and Adesina, A.A. (2006a). A mechanistic model for
propane steam reforming on a bimetallic Co-Ni catalyst in fluidized bed reactor,
in: Rhee, H.-K., Nam, I.-S. & Park, J.M. (Eds.), Studies in Surface Science and
Catalysis. Elsevier B.V., Amsterdam.
328
Chapter 6: Reaction Studies
Hardiman, K.M., Cooper, C.G., Adesina, A.A. and Lange, R. (2006b). Post-mortem
characterization of coke-induced deactivated alumina-supported Co-Ni catalysts.
Chem. Eng. Sci., 61, 2565-2573.
Hardiman, K.M., Alenazey, F.S.M., Adesina, A.A. and Lange, R. (2006c). Optimal
operation of a fluidised bed steam reformer with decaying catalyst. In
preparation.
Kao, Y.K. and Bankoff, S.G. (1975). Optimal control of a stirred-tank reactor with
rapidly-decaying catalyst. Chem. Eng. Sci., 30, 1315-1324.
Kepinski, L., Stasinska, B. and Borowiecki, T. (2000). Carbon deposition on Ni/Al
2
O
3
catalysts doped with small amounts of molybdenum. Carbon, 38, 1845-1856.
Kobayashi, M. and Kobayashi, H. (1974). Transient response method in heterogenous
catalysis. Catal. Rev. Sci. Eng., 23, 1-15.
Krishnaswamy, S. and Kittrell, J.R. (1979). Analysis of time-dependent data for
deactivating catalysts. Ind. Eng. Chem. Proc. Des. Dev., 18, 399-410.
Levenspiel, O. (1999). Commentaries: Chemical reaction engineering. Ind. Eng. Chem.
Res., 38, 4140-4143.
Liu, H., Li, T., Tian, B. and Xu, Y. (2001). Study of the carbonaceous deposits formed
on a Mo/HZSM-5 catalyst in methane dehydro-aromatization by using TG and
temperature-programmed techniques. Appl. Catal. A: Gen., 213, 103-112.
Liu, H., Su, L., Wang, H., Shen, W., Bao, X. and Xu, Y. (2002). The chemical nature of
carbonaceous deposits and their roles in methane dehydro-aromatization on
Mo/MCM-22 catalysts. Appl. Catal. A: Gen., 236, 263-280.
Lobo, L.S., Trimm, D.L. and Figueiredo, J.L. (1973). Kinetics and mechanism of
carbon formation from hydrocarbon on metals, in: Hightower, J.W. (Ed.),
Proceedings of the Fifth International Congress on Catalysis, vol. 2. North-
Holland Publishing Company, Amsterdam, pp. 1125-1137.
Matsumoto, H. and Bennett, C.O. (1978). The transient method applied to the
methanation and Fischer-Tropsch reactions over a fused iron catalyst. J. Catal.,
53, 331-344.
329
Chapter 6: Reaction Studies
Montgomery, D.C. (1997). Design and Analysis of Experiments. John Wiley & Sons,
Inc., New York.
Monzn, A., Romeo, E. and Borgna, A. (2003). Relationship between the kinetic
parameters of different catalyst deactivation models. Chem. Eng. J., 94, 19-28.
Moulijn, J.A., van Diepen, A.E. and Kapteijn, F. (2001). Catalyst deactivation: is it
predictable? What to do? Appl. Catal. A: Gen., 212, 3-16.
Ogunye, A.F. and Ray, W.H. (1971). Optimization of cyclic tubular reactors with
catalyst decay. Ind. Eng. Chem. Proc. Des. Dev., 10, 410-416.
Ogunye, A.F. and Ray, W.H. (1971). Optimization of recycle reactors having catalyst
decay. Ind. Eng. Chem. Proc. Des. Dev., 10, 416-420.
Opoku-Gyamfi, K. (1999). A novel thermally self-sustaining fluidised bed reactor for
hydrogen production from methane. Ph.D. thesis, University of New South
Wales, Sydney, Australia.
Praharso, Adesina, A.A., Trimm, D.L. and Cant, N.W. (2004). Kinetic study of iso-
octane steam reforming over a nickel-based catalyst. Chem. Eng. J., 99, 131-
136.
Rostrup-Nielsen, J.R. (1984). Catalytic Steam Reforming. Springer-Verlag, Berlin.
Sapre, A.V. (1997). Catalyst deactivation kinetics from variable space-velocity
experiments. Chem. Eng. Sci., 52, 4615-4623.
Satterfield, C.N. and Sherwood, T.K. (1963). The role of diffusion in catalysis.
Addison-Wesley, Reading, Massachusetts.
Shadman-Yazdi, F. and Petersen, E.E. (1972). Changing catalyst performance by
varying the distribution of active catalyst within porous supports. Chem. Eng.
Sci., 27, 227-237.
Somorjai, G.A. (1981). The catalytic hydrogenation of carbon monoxide: The formation
of Cl hydrocarbons. Catal. Rev. -Sci. Eng., 23, 189-202.
Szwast, Z. and Sieniutycz, S. (2001). Optimal temperature profiles for parallel-
consecutive reactions with deactivating catalyst. Catal. Today, 66, 461-466.
330
Chapter 6: Reaction Studies
Van Doorn, J. and Moulijn, J.A. (1990). Carbon deposition on catalysts. Catal. Today,
7, 257-266.
Verykios, X.E. (2003). Mechanistic aspects of the reaction of CO
2
reforming of
methane over Rh/Al catalyst. Appl. Catal. A: Gen., 255, 101-111.
2
O
3
Winslow, P. and Bell, A.T. (1985). Studies of the surface coverage of unsupported
ruthenium by carbon- and hydrogen-containing adspecies during CO
hydrogenation. J. Catal., 91, 142-154.
Xu, J. and Froment, G.F. (1989). Methane steam reforming, methanation and water-gas
shift: I. Intrinsic kinetics. AIChE J., 35, 88-96.
Yang, R., Zhang, Y., Iwama, Y. and Tsubaki, N. (2005). Mechanistic study of a new
low-temperature methanol synthesis on Cu/MgO catalysts. Appl. Catal. A: Gen.,
288, 126-133.
331
Chapter7: Conclusions and Recommendations
C
C
h
h
a
a
p
p
t
t
e
e
r
r
7
7
C
C
O
O
N
N
C
C
L
L
U
U
S
S
I
I
O
O
N
N
S
S
A
A
N
N
D
D
R
R
E
E
C
C
O
O
M
M
M
M
E
E
N
N
D
D
A
A
T
T
I
I
O
O
N
N
S
S
7.1 Conclusions
Catalyst impregnated at low pH was found to yield better metal distribution, higher
metal surface area and increased reducibility. Investigation of pH effect showed that
acidic impregnation of the alumina-supported catalyst enhanced the mobility of
positively-charged Co and Ni ions farther into the support matrix to suitable adsorption
sites. Not surprisingly, such a uniform distribution of active sites is also advantageous in
providing a more stable steam reforming under a long period of time. However,
impregnation at pH 2 was detrimental to BET surface area due to alumina dissolution
caused by chemical attack. Further, total surface area is also dependent on calcination
temperature whereby higher temperature treatment would greatly reduce surface area.
However, H
2
-chemisorption experiments showed that more uniform dispersion profile
and finer crystallites could generally be achieved from acidic impregnation and
calcination with shorter duration, higher temperature and slower heating rate. The
nature of acid interface, as arising from metal-metal and/or metal-support interactions,
was explored in terms of strength and concentration. Lower-pH impregnation and
longer calcination time are the main determinants for stronger acid sites. Interestingly,
332
Chapter7: Conclusions and Recommendations
acid population is greater in the catalysts impregnated with basic solution (at pH 8).
More acid sites were also produced in specimens with decreased calcination
temperature. Significantly, NH
3
-TPD runs showed that the acid sites on both
monometallic Co and Ni catalysts, as well as the bimetallic Co-Ni system, were
predominantly medium Lewis type.
Structural analyses examined via XRD, TEM and SEM revealed the superior feature of
metallic dispersion on the surface of bimetallic Co-Ni system. XRD patterns confirmed
that the acidic specimen was more easily reducible than the basic catalyst although both
systems possess similar bulk components. The fact that low-pH impregnation leads to
better active metal distribution was further substantiated by TEM micrograph, whereas
in the high-pH catalyst metal deposition occurred predominantly on the particle exterior
due to rapid surface adsorption. Moreover, SEM images also evinced uniform
distribution of metallic Co and Ni particles in the bimetallic sample.
In agreement with H
2
chemisorption and TEM, calcination derivative profile (obtained
from thermogravimetric analysis) of the catalyst impregnated at pH 2 showed a short
and broad peak corresponding to slow and mild decomposition of nitrate precursors as
they are uniformly dispersed throughout the support matrix. The behaviour was
especially marked for specimens calcined at slower heating rate, which also leads to
better metal dispersion. These propositions are also in agreement with quantitative
evaluation of the calcination curves using 3
rd
-order Avrami-Erofeev sigmoid model.
From the two-cycle reduction-oxidation run, it would seem that the reaction stability in
acid catalyst was somewhat related to its ability in reproducing initial phases following
333
Chapter7: Conclusions and Recommendations
treatment under alternating oxidising and reducing conditions. Even so, none of the
calcination variables was deemed significant in influencing the reduction and oxidation
kinetic coefficients as determined from a 3
rd
-order Avrami-Erofeev model.
Steam reforming test under conditions favouring deactivation confirmed that the
homogenous acid catalyst exhibited superior stability. On the other hand, the basic
specimen recorded high initial activity, followed by swift deterioration due to rapid
hydrocarbon adsorption and eventual coke build-up in the region close to the support
periphery. In general, the extent of carbon resilience (defined as the ratio of reforming
rate constant over deactivation rate constant) was generally greater in catalysts calcined
at lower temperature and slower heating rate.
It was found that the physicochemical properties of the bimetallic catalyst are not a
simple linear combination of the individual metallic constituents suggesting the
involvement of chemical or textural synergism. While bimetallic catalyst appeared to
give lower reforming rate compared to the respective monometallic catalysts, it was
however compensated for by a larger drop in deactivation rate. Therefore, the dilution
effect of Co species was in fact effective in improving coking resilience of a
conventional Ni-based system.
Post-mortem characterisation of the bimetallic catalysts demonstrated that coke content
was indeed linearly proportional to deactivation rate parameter - an evidence of carbon
deposition as the principal cause of activity decay. Interestingly, XRD analyses of
coked specimens indicated that bulk metallic phases of Co-Ni, Co and Ni systems
334
Chapter7: Conclusions and Recommendations
remained essentially unchanged following deactivation-influenced steam reforming.
Furthermore, as shown in coke burn-off experiments performed in a TGA unit, the rate
of carbon removal from these three catalysts increased with higher coke content.
However, burn-off activation energy and density of reacting site depended largely on
type and composition of metallic species.
Steady-state kinetics of propane steam reforming in fluidised bed reactor over the Co-Ni
catalyst under conditions of both low (0.8-1.6) and high (2.4-4.8) S:C ratios were
analysed in terms of power-law and mechanistic models. Based on the power-law
expression, the presence of propane and steam significantly enhanced the propensity for
reforming as evidenced by positive reaction orders of 1.20 and 0.82 for propane and
steam respectively. The corresponding activation energy for this scheme was given as
63.2 kJ mol
-1
. Moreover, mechanistic approach revealed that the steam reforming of
propane may be adequately described by Langmuir-Hinshelwood mechanism involving
different adsorption sites for hydrocarbon and steam. The activation energy based on
this kinetic mechanism was estimated as 40 kJ mol
-1
. The lower E-values posited in
these two schemes compared to the range 65-70 kJ mol
-1
suggested for a Ni-based
reforming catalyst highlights possible synergism in the bimetallic catalyst and the
enhanced reactor mixing associated with a fluidised bed system.
In view of the generalised Levenspiel integrated reaction-deactivation model, it was
demonstrated that steam reforming in a well-mixed fluidised bed reformer undergoing
significant carbon deposition (carried out at S:C = 0.8) may be described by a CSTR
model involving second-order propane steam reforming with hyperbolic activity decay.
335
Chapter7: Conclusions and Recommendations
Subsequently, deactivation and reaction kinetic coefficients estimated using this model
under reaction conditions of 773-873 K and S:C = 0.8-1.6 signalled a strong
dependency on both S:C and temperature as also substantiated by statistical ANOVA
treatment. The deactivation was characterised by a negative activation-energy, which
lends credence to the fact that the removal of culprit coke species C

occurred at a faster
rate than its formation (from dehydropolymerisation of C

) within the temperature range


adopted in this study. This finding was supported by investigations on the nature and
quantity of carbonaceous deposits performed via total carbon content measurement,
XRD, TEM, as well as temperature-programmed runs. TEM images showed the
presence of C

as a polymeric filament with metal crystallite at its tip. Significantly,


thermogravimetric analysis suggested that minimum sintering and substantial recovery
of initial oxide phases following oxidative regeneration can be achieved with the
implementation of a reduction step prior to oxidation. Further, the oxidative coke burn-
off was shown to exhibit an activation energy in the narrow range 54-58.5 kJ mol
-1
regardless amount of carbon deposition implying the active Co/Ni crystallites as the
predominant sites of coke lay-down. Additionally, based on rate coefficients, the kinetic
analysis presented a simple relationship between coke deposition and removal
processes. The kinetic rate coefficients of the burn-off runs increased from 2.9 to 12 h
-1
with decreasing coke content.
The optimum operational policy via pre-determined temperature control was
demonstrated to be an effective strategy for minimising conversion loss despite the
severe coke lay-down as seen in the carbon content measurements. The high-
temperature environment and the longer experimental duration during transient steam
336
Chapter7: Conclusions and Recommendations
reforming runs also appeared to bring about the formation of C
C
species from the aging
of C

. In principle, transient monitoring of various product ratios and yields in reactor


exit stream revealed that propensity for steam reforming decreased with time, while
water-gas shift and Boudouard reactions became less favourable with increasing
temperature.
7.2 Recommendations
Based on the findings and observations of this research project, the following
suggestions are proposed for future studies:
- Whilst cobalt addition to an alumina-supported nickel reforming catalyst
prepared by acidic impregnation has improved reaction stability, further study
may be required to find the optimum proportion between cobalt and nickel
metal loadings;
- Cobalt and nickel constituents are available from various precursors (e.g.
acetate, hydroxide, oxalate, sulfate, carbonate, etc). Different precursor
potentially introduces unique electronic properties to the resulting catalyst
system. Consequently, it may be of a special interest to investigate the effect
of precursor, especially in a bimetallic system with the involvement of metal-
metal interaction;
- In some cases with prolonged time-on-stream and higher temperature run, it
may be necessary to isolate the sintering effect from the overall deactivation
rate parameter. A possible technique is to load catalyst sample in fluidised
bed reactor under flowing inert at hydrodynamic and reaction conditions, but
337
Chapter7: Conclusions and Recommendations
without flowing reactant gases. Subsequently, pulses of reactant gases are
introduced at a certain interval while the conversion is being monitored. The
conversion profile may be essentially regarded as the time-dependent activity
behaviour under the exclusive effect of sintering;
- As has been revealed in the thermal analysis of spent catalysts, a two-step
reduction-oxidation regeneration technique resulted in minimum loss of
metallic sites. Therefore, it will be interesting to re-use these regenerated
catalysts for steam reforming to observe whether they exhibit similar catalytic
behaviours as the fresh system;
- Since deactivation, k
d
', and reaction, k', kinetic coefficients vary with time,
their definitions in the optimal reactor policy may be more appropriately
treated as time-dependent functions rather than time-averaged rate constants;
- The estimates of k
d
' and k' were evaluated from steady-state reforming runs in
the temperature range of 773-873 K. Nonetheless, in the policy verification
runs it was required that temperature was subsequently increased beyond 873
K. As the nature of coke formed depends strongly on temperature, it is
expected that different type of coke residues may be formed during the
transient runs. Accordingly, the suggested temperature-time policy should be
constructed based on steady-state data collected up to the highest temperature
reached in the transient runs. This was however not carried out in this project
due to the risk of sintering. The use of alumina support pre-treated at higher
temperature (e.g. o-Al
2
O
3
) may possess greater resistance against sintering,
however, there needs to be a more careful judgment in compromising the
lower surface area due to high pre-treatment temperature;
338
Chapter7: Conclusions and Recommendations
- Optimal dynamic temperature policy may be developed based on maintaining
stable H
2
or CO concentration in product stream instead of propane reactant
conversion. In addition, S:C ratio may also be used as the variable to be
controlled for keeping constant conversion;
- Carbon-induced deactivation occurs in many important petroleum and
petrochemical operations. It will be important for economic considerations
that the strategies adopted in this project can be extended to other catalytic
processes with significant coking problems, such as dry reforming, partial
oxidation, catalytic cracking, dehydrogenation, hydrotreating, MTG,
dearomatisation, etc.
339
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
A
A
S
S
A
A
M
M
P
P
L
L
E
E
C
C
A
A
L
L
C
C
U
U
L
L
A
A
T
T
I
I
O
O
N
N
F
F
O
O
R
R
P
P
R
R
E
E
P
P
A
A
R
R
A
A
T
T
I
I
O
O
N
N
O
O
F
F
S
S
T
T
O
O
C
C
K
K
S
S
O
O
L
L
U
U
T
T
I
I
O
O
N
N
D
D
U
U
R
R
I
I
N
N
G
G
C
C
A
A
T
T
A
A
L
L
Y
Y
S
S
T
T
P
P
R
R
E
E
P
P
A
A
R
R
A
A
T
T
I
I
O
O
N
N
Catalyst type = 5Co-15Ni/80Al
2
O
3
Metal salt used = Co(NO
3
)
2
6H
2
O (MW = 291.03), and;
Ni(NO
3
)
2
6H
2
O (MW = 290.81)
Pore volume of Al
2
O
3
= 0.8 ml g
-1
Basis = 100 g Al
2
O
3
Thus,
Weight of Co(NO
3
)
2
6H
2
O salt required =
1
1
mol g 58.93
mol g 291.03
80
5
g 100

= 31.11 g
Weight of Ni(NO
3
)
2
6H
2
O salt required =
1
1
mol g 58.71
mol g 290.81
80
15
g 100

= 92.87 g
Volume of water added = 2 0.8 ml g
-1
100 g = 160 ml
340
Appendix
The calculated amounts of salts added in the preparation of catalysts employed are
summarised in Table A1. Further, Table A2 shows metallic concentrations for cobalt
and nickel in impregnation solutions.
Table A1. Amount of salts added in mono- and bi-metallic catalysts.
Amount of Salt Added (g)
Catalyst
Co(NO
3
)
2
6H
2
O Ni(NO
3
)
2
6H
2
O
20Co/80Al
2
O
3
124.44 -
20Ni/80Al
2
O
3
- 123.83
5Co-15Ni/80Al
2
O
3
31.11 92.87
Table A2. Metallic concentration in impregnation solution (based on 100 g Al
2
O
3
).
Metal Co Ni
Salt formula Co(NO
3
)
2
6H
2
O Ni(NO
3
)
2
6H
2
O
Molecular weight (g mol
-1
) 291.03 290.81
Water added in a single
impregnation (ml)
160 160
Concentration in impregnation
solution (g metal ml
-1
)
0.15626 & 0.03906 for
20Co and 5Co-15Ni
catalysts respectively
0.15625 & 0.11718 for
20Ni and 5Co-15Ni
catalysts respectively
341
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
B
B
M
M
E
E
T
T
H
H
O
O
D
D
O
O
L
L
O
O
D
D
Y
Y
F
F
O
O
R
R
S
S
O
O
L
L
V
V
I
I
N
N
G
G
C
C
O
O
M
M
P
P
L
L
E
E
X
X
C
C
H
H
E
E
M
M
I
I
C
C
A
A
L
L
-
-
R
R
E
E
A
A
C
C
T
T
I
I
O
O
N
N
E
E
Q
Q
U
U
I
I
L
L
I
I
B
B
R
R
I
I
A
A
As multiple reactions are involved, the equilibrium composition was more conveniently
predicted via minimisation of the total Gibbs energy. An example of calculation is
presented in the following based on the methodology given in Perry and Green [1997].
Assuming a system with ideal gases, for chemical species i it is required that:
0 ln
0
0
= +
|
|
.
|

\
|
E
+
A
_
k
ik
g
k
i i
i
g
f
a
T R P
P
n
n
T R
G
i

(i = 1, 2, 3, , N) (B.1)
where
0
i
f
G A = standard Gibbs energy change of formation for species i (J mol
-1
),
R
g
= gas constant,
T = temperature (K),
P = pressure (kPa),
PP
0
= standard-state pressure of 100 kPa,
n
i
= number of molecules of species i,

k
= Lagrange multiplier of the k
th
element,
a
ik
= number of atoms of the k
th
element present in each molecule of species i.
342
Appendix
Additionally, material balance for element k may be described as:
i ik k
i
n a A =
_
(k = 1, 2, 3, , w)
(B.2)
where A
k
= total number of atomic masses of the k
th
element in the feed.
There are N equilibrium equations and there are w material-balance equations giving a
total of N + w equations. The unknowns in these equations are the n
i
, of which there are
N, and the
k
, of which there are w, yielding a total of N + w equations. Therefore, there
are equal number of equations and unknowns permitting for a simultaneous technique in
solving these equations.
As an example, the equilibrium composition for steam reforming of propane at 800 K
and 1 bar would be calculated. The initial feed is assumed to contain 1 mol of C
3
H
8
and
3 moles of H
2
O (S:C feed ratio = 1). From Sandler [1999], the values of
0
i
f
G A are
estimated as:
3 8
0 1
82,128 J mol
C H
f
G

A =
4
0 1
22, 569 J mol
CH
f
G

A =
2
0 1
216, 410 J mol
H O
f
G

A =
0 1
190, 700 J mol
CO
f
G

A =
2
0 1
408, 010 J mol
CO
f
G

A =
343
Appendix
2
0 1
8, 339 J mol
H
f
G

A =
The required values of A
k
and a
ik
are summarised in Table B.1.
Table B.1. Values of A
k
and a
ik
.
Element k
Carbon Oxygen Hydrogen
A
k
= number of atomic masses of k in the system
A
C
= 3 A
O
= 3 A
H
= 14
Species i a
ik
= number of atoms of k per molecule of i
C
3
H
8
3
,
8 3
=
C H C
a 0
,
8 3
=
O H C
a 8
,
8 3
=
H H C
a
CH
4
1
,
4
=
C CH
a 0
,
4
=
O CH
a 4
,
4
=
H CH
a
H
2
O
0
,
2
=
C O H
a 1
,
2
=
O O H
a 2
,
2
=
H O H
a
CO
1
,
=
C CO
a 1
,
=
O CO
a 0
,
=
H CO
a
CO
2
1
,
2
=
C CO
a 2
,
2
=
O CO
a 0
,
2
=
H CO
a
H
2
0
,
2
=
C H
a 0
,
2
=
O H
a 2
,
2
=
H H
a
Since P = 1 bar, the six equations for the six species (Eq. B.1) may be written:
C
3
H
8
:
3 8
3 8 82,128
ln 0
C H
C H
i i
n
RT n RT RT

+ + +
E
=
CH
4
:
4
4 22, 569
ln 0
CH
C H
i i
n
RT n RT RT

+ + +
E
=
344
Appendix
H
2
O:
2
2 216, 410
ln 0
H O
O H
i i
n
RT n RT RT

+ + +
E
=
CO:
190, 700
ln 0
CO C O
i i
n
RT n RT RT

+ + +
E
=
CO
2
:
2
2 408, 010
ln 0
CO
C O
i i
n
RT n RT RT

+ + +
E
=
H
2
:
2
2 8, 339
ln 0
H
H
i i
n
RT n RT

+ +
E
=
Further, the material-balance equations for the three elements (Eq. B.2) are given by:
C: 3 3
2 4 8 3
= + + +
CO CO CH H C
n n n n
O: 3 2
2 2
= + +
CO CO O H
n n n
H: 14 2 2 4 8
2 2 4 8 3
= + + +
H O H CH H C
n n n n
These equations can then be solved simultaneously via computer simulation.
Subsequently, by taking into account
2 2 2 4 8 3
H CO CO O H CH H C
i
i
n n n n n n n + + + + + =
_
(B.3)
and
i i
i
i
n
n
y
E
=
(B.4)
345
Appendix
the following results were obtained:
8 3
H C
y = -0.038 (0.000)
RT
C

= -4.043
4
CH
y = 0.200
O H
y
2
= 0.036
RT
O

= 34.079
CO
y = 0.216
2
CO
y = 0.049
RT
H

= 0.824
2
H
y = 0.537
_
i
i
y = 1.000
The negative sign of propane composition, , was a computational artefact of
numerical solution which also indicates complete consumption of propane reactant and
consequently, it was rounded off to zero. Moreover, the values of
8 3
H C
y
k
/RT have no
significance, however they are included to show the full results.
346
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
C
C
T
T
R
R
A
A
N
N
S
S
P
P
O
O
R
R
T
T
R
R
E
E
S
S
I
I
S
S
T
T
A
A
N
N
C
C
E
E
C
C
A
A
L
L
C
C
U
U
L
L
A
A
T
T
I
I
O
O
N
N
S
S
C.1 Determination of mass transfer coefficient (k )
c
The superficial gas velocity was determined from;
1 2
2
1
s m 10 66 4
K 298 s 60 cm 14 3
K 873 min 1 min ml 300

=


= = .
. A
V
U
Therefore, Reynolds number, Re, may be evaluated by,
0611 0
s Pa 10 12 7
m 10 31 2 s m 10 66 4 m kg 10 04 4
5
4 2 2 3 1
.
.
. . .

Ud
Re
g
p g
=


= =


For fluidised beds with Re = 0.01-15,000, the following correlation [Dwivedi and
Upadhyay, 1977] may be used:
( )
29 . 17
) 0611 . 0 (
365 . 0
0611 . 0
765 . 0
5 . 0
1 365 . 0 765 . 0 1
386 . 0 82 . 0 386 . 0 82 . 0
=
(

+ =
(

+ =
Re Re
J
D
c
347
Appendix
Adopting the parameters listed in Table 4.6,
752 . 6
s m 10 61 . 2 m kg 10 04 . 4
s Pa 10 12 . 7
1 2 5 3 1
5
=


= =

AB g
D
Sc

Subsequently, the mass transfer coefficient, k


c
, may be calculated from;
1
1 2
s m 226 . 0
) 752 . 6 (
s m 10 66 . 4 29 . 17
) (
3
2
3
2


=

= =
Sc
U J
k
D
c
C.2 Determination of heat transfer coefficient (h
max
)
Using the coefficients in Table 4.6, Fr, Ar and Pr coefficients may be evaluated as;
240 . 0
s m 8 . 9 m 10 31 . 2
s m ) 10 332 . 2 (
) (
2 4
2 2 2 2
2
=

= =


g d
U
Fr
p
mf
5 . 38
s m kg ) 10 12 . 7 (
s m 8 . 9 m kg ) 404 . 0 4000 ( m kg 10 04 . 4 m ) 10 31 . 2 (
) ( ) (
2 2 2 2 5
2 3 3 1 3 3 4
2
3
=

=


g
g s g p
g d
Ar


972 . 0
K m W 10 97 . 3
s m kg 10 12 . 7 K kg J 10 42 . 5
Pr
1 1 2
1 1 5 1 1 2
=


= =


g
g pg
C

348
Appendix
Thus, Nu
max
was obtained from the correlation [Sathiyamorthy and Raja Rao, 1992] as
follows:
79 . 9 ) 972 . 0 (
10 19 . 2
10 06 . 3
) 5 . 38 ( ) 240 . 0 ( 0046 . 0 Pr 0046 . 0
3 . 0
5 . 0
2
6
63 . 0 42 . 0 3 . 0
5 . 0
63 . 0 42 . 0
max
=
|
|
.
|

\
|

=
|
|
.
|

\
|
=

vg
vs
C
C
Ar Fr Nu
The heat transfer coefficient, h
max
, may then be computed from:
1 2
4
1 1 2
max
max
K m W 7 . 682 , 1
m 10 31 . 2
K m W 10 97 . 3 79 . 9


= =
p
g
d
Nu
h
349
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
D
D
C
C
A
A
L
L
C
C
U
U
L
L
A
A
T
T
I
I
O
O
N
N
S
S
O
O
F
F
T
T
H
H
E
E
O
O
R
R
E
E
T
T
I
I
C
C
A
A
L
L
C
C
A
A
T
T
A
A
L
L
Y
Y
S
S
T
T
W
W
E
E
I
I
G
G
H
H
T
T
C
C
H
H
A
A
N
N
G
G
E
E
D
D
U
U
R
R
I
I
N
N
G
G
T
T
H
H
E
E
R
R
M
M
O
O
G
G
R
R
A
A
V
V
I
I
M
M
E
E
T
T
I
I
C
C
R
R
U
U
N
N
D.1 Calcination
Catalyst type = 5Co-15Ni/80Al O
2 3
O and NiO Major oxide phases = Co
3 4
Basis = 100 g Al
2
O
3
In the dried impregnated system,
Amount of Co(NO
3
) =
1
1
mol g 58.93
mol g 182.94
80
5
g 100

= 19.40 g
2
Amount of Ni(NO
3
)
2
=
1
1
mol g 58.71
mol g 182.72
80
15
g 100

= 58.35 g
Total weight of moisture-free impregnated solid = 100 + 19.40 + 58.35
= 177.75 g
In the freshly-oxidised system,
350
Appendix
Amount of Co
3
O =
1
1
mol g 176.79
mol g 240.79
80
5
g 100

= 8.51 g
4
Amount of NiO =
1
1
mol g 58.71
mol g 74.71
80
15
g 100

= 23.86 g
Total weight of freshly-calcined system = 100 + 8.51 + 23.86 = 132.37 g
Therefore,
%
.
. .
100
75 177
37 132 75 177
|
.
|

\
|
= 25.53% Weight change for complete calcination =
D.2 Reduction
Catalyst type = 5Co-15Ni/80Al O
2 3
O and NiO Major oxide phases = Co
3 4
Basis = 100 g Al
2
O
3
In the freshly-calcined system,
Amount of Co
3
O =
1
1
mol g 176.79
mol g 240.79
80
5
g 100

= 8.51 g
4
Amount of NiO =
1
1
mol g 58.71
mol g 74.71
80
15
g 100

= 23.86 g
Total weight of freshly-calcined system = 100 + 8.51 + 23.86 = 132.37 g
In the freshly-reduced system,
Amount of Co =
80
5
g 100 = 6.25 g
351
Appendix
Amount of Ni =
80
15
g 100 = 18.75 g
Total weight of alumina + reduced metals = 100 + 6.25 + 18.75 = 125 g
Therefore,
% 100
37 . 132
125 37 . 132
|
.
|

\
|
Weight change for complete reduction = = 5.57%
352
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
E
E
D
D
E
E
V
V
E
E
L
L
O
O
P
P
M
M
E
E
N
N
T
T
O
O
F
F
L
L
A
A
N
N
G
G
M
M
U
U
I
I
R
R
-
-
H
H
I
I
N
N
S
S
H
H
E
E
L
L
W
W
O
O
O
O
D
D
A
A
N
N
D
D
E
E
L
L
E
E
Y
Y
-
-
R
R
I
I
D
D
E
E
A
A
L
L
K
K
I
I
N
N
E
E
T
T
I
I
C
C
M
M
O
O
D
D
E
E
L
L
S
S
The mechanism of hydrocarbon steam reforming was suggested by Hardiman et al.
[2006a] to take place as follows:
1 1 1
S H C S H C 2S H C
2 1 2 1
+ +
m m m y m x y x
(E.1)
2 2 2 2
S H S OH 2S O H + + (E.2)
2 1 2 1
S H S O H C S OH S H C
2 1 2 1
+ +
m m m m
(E.3)
1 1 1 1 1 1
S H C S CHO S S O H C
2 1 2 1
+ +
m m m m
(E.4)
2 1 2 1
S H S CO S S CHO + + (E.5)
(in excess H O)
2 1 2 1
S H S COO S OH S CO + + (E.6) 2
1 1
S CO S CO + | (E.7)
1 2 1
S CO S COO + | (E.8)
2 2 2 2
2S H S H S H + | + (E.9)
353
Appendix
where m , m , x and y are integers and S & S
1 2 1 2
are 2 types of sites.
E.1 Dual-site adsorption of propane and steam (Model 1)
For propane substrate (x = 3 and y = 8), Eqs. E.1-E.3 may be re-written as:
H + 2S C H S + CH S (E.10) C
3 8 1 2 5 1 3 1
2 2 2 2
S H S OH 2S O H + + (E.11)
S + OH S CH O S + H S (E.12) CH
3 1 2 3 1 2
For site S
1
,
2
H C H C CH H C
8 3 8 3 3 5 2
p K =
(E.13)
3 5 2
CH H C
= As propane was being consumed, it can be assumed that , and so,
2
H C H C
2
CH
8 3 8 3 3
p K =
(E.14)
8 3 8 3 3
H C H C CH
p K u =
(E.15)
Site balance around S
1
suggests
_
=
+ + =
3
3
CH
CH
1
i
i
u u
(E.16)
354
Appendix
is MARI (the most abundant reactive species), Eq. E.16 becomes, Accordingly, if CH
3
0
3
CH
~
_
= i
i
u
3
CH
1 + = u , since
(E.17)
Substitution of Eq. E.15 into E.17 yields,
8 3 8 3
H C H C
1
1
p K +
= u
(E.18)
For site S
2
,
2
O H O H H OH
2 2
p K ' = ' '
(E.19)
H OH
' = ' As steam was being consumed, it can be assumed that , and so,
2
O H O H
2
H
2 2
p K ' = '
(E.20)
O H O H H
2 2
p K u u ' = '
(E.21)
Site balance around S
2
suggests
_
=
' + ' + ' =
H
H
1
i
i
u u
(E.22)
Accordingly, if H is MARI (the most abundant reactive species), Eq. E.22 becomes,
355
Appendix
0
H
~ '
_
= i
i
u
H
1 ' + ' =u , since
(E.23)
Substitution of Eq. E.21 into E.23 yields,
O H O H
2 2
1
1
p K +
= ' u
(E.24)
Assuming that Eq. E.12 is the rate determining step,
H CH
3
u u ' = k r
(E.25)
Finally, by substituting Eqs. E.15, E.18, E.21 and E.24 into E.25, the kinetic model
ascribed to dual-site adsorption of propane and steam was obtained as follows:
( )( )
O H O H H C H C
O H H C
2 2 8 3 8 3
2 8 3
1 1 p K p K
p p k
r
rxn
+ +
=
O H H C rxn
K K k k
2 8 3
= , where
(E.26)
E.2 Dual-site adsorption with non-dissociative water adsorption
(Model 2)
For non-dissociative water adsorption,
p K ' = '
O H O H O H
2 2 2
(E.27)
356
Appendix
Site balance around S
2
suggests
O H
2
1 ' + ' = u
(E.28)
Substitution of Eq. E.27 into E.28 yields,
O H O H
2 2
1
1
p K +
= ' u
(E.29)
Similarly, the kinetic model of the dual site with non-dissociative water adsorption was
derived as:
( ) ( )
O H O H H C H C
O H H C
2 2 8 3 8 3
2 8 3
1 1 p K p K
p p k
r
rxn
+ + +
=
O H H C rxn
K K k k
2 8 3
= , where
(E.30)
E.3 Adsorption of both propane and steam on the same site (Model 3)
If propane and water are adsorbed on the same site, site balance suggests
H CH
3
1 u u ' + + =
(E.31)
Subsequently, substitution of Eqs. E.15 and E. 21 into E.31 yields
357
Appendix
O H O H H C H C
2 2 8 3 8 3
1
1
p K p K + +
= u
(E.32)
Thus, the adsorption of both propane and steam on the same site may be described by:
( )
2
O H O H H C H C
O H H C
2 2 8 3 8 3
2 8 3
1 p K p K
p p k
r
rxn
+ +
=
O H H C rxn
K K k k
2 8 3
= , where
(E.33)
E.4 Eley-Rideal model (Model 4)
This specifies that water is unadsorbed, whereupon
2
O H O H
2
O H
2 2 2
p K = '
(E.34)
O H O H O H
2 2 2
p K = ' u
(E.35)
Substituting of Eq. E.35 into the generalised form of E.25 gives the Eley-Rideal kinetic
model:
( )
8 3 8 3
2 8 3
H C H C
O H H C
1 p K
p p k
r
rxn
+
=
O H H C rxn
K K k k
2 8 3
= , where
(E.36)
358
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
F
F
E
E
X
X
A
A
M
M
P
P
L
L
E
E
S
S
O
O
F
F
D
D
E
E
R
R
I
I
V
V
A
A
T
T
I
I
O
O
N
N
F
F
O
O
R
R
L
L
E
E
V
V
E
E
N
N
S
S
P
P
I
I
E
E
L
L
C
C
O
O
U
U
P
P
L
L
E
E
D
D
R
R
E
E
A
A
C
C
T
T
I
I
O
O
N
N
-
-
D
D
E
E
A
A
C
C
T
T
I
I
V
V
A
A
T
T
I
I
O
O
N
N
M
M
O
O
D
D
E
E
L
L
F.1 Case 1: 1
st
order reforming with hyperbolic decay law in a PFR
For a 1
st
order steam reforming kinetics (n = 1) with hyperbolic activity decay law (d =
2) in a PFR, we have;
2
a k
dt
da
d
' = (F.1)
which integrates to
t k
a
d
1
1
' +
=
(F.2)
When reaction is 1
st
order,
A SR
C a k r q ' =
(F.3)
359
Appendix
) 1 (
0
X C a k r
A SR
' = q
(F.4)
The PFR design equation is
) )
'
=

=
) 1 (
0
0
X C a k
dX
r
dX
F
W
A SR A
q
(F.5)
0
0
X A
0
A
k' a C W
1
ln( 1 X ) ln
F 1 X
q
| |
= =
|

\ .
where
0
0
A
A
F
W C
= ' t
(F.6)
|
.
|

\
|

= ' '
X
a k
1
1
ln t q
(F.7)
|
|
.
|

\
|
=
' +
' '
A
A
d
C
C
t k
k
0
ln
1
t q
(F.8)
Noting that
W
|
q
1
= , Eq. (F.8) may be re-arranged to yield:
t
k
k
k
C
C
W d W
A
A
ln
1
0
|
.
|

\
|
' '
'
+
' '
=
|
|
.
|

\
| t
|
t
|
(F.9)
( ) | |
A A
C C
0
ln 1 from which a plot of versus t may be used to obtain relevant parameter
estimates.
360
Appendix
F.2 Case 2: 2
nd
order reforming with reciprocal decay law in a CSTR
For a 2
nd
order steam reforming kinetics (n = 2) with reciprocal activity decay law in a
CSTR, we have;
b
t a

=
0
| (F.10)
and,
2

A SR
C a k r q ' = (F.11)
2 2
) 1 (
0
X C a k r
A SR
' = q
(F.12)
The CSTR design equation is
2 2
) 1 (
0 0
X C a k
X
r
X
F
W
A SR A
'
=

=
q
(F.13)
whereupon,
A
A
A
A
A
A
A
A
A
A
C
C
C
C
C
C
C
C
X
X
F
W C a k
0 0
0
0
0
0
2
2 2
2
1
) 1 (

|
|
.
|

\
|
=
|
|
.
|

\
|

=
' q
(F.14)
thus,
361
Appendix
A A
A
C C
C
a k
1

2
0
= ' ' t q
(F.15)
and,
A A
A
-b
C C
C
t k
1

2
0
0
= ' ' t q |
(F.16)
Consequently, linearisation provides;
t b
k
C C
C
W A A
A
ln - ln
1
ln
0
2
0
|
|
.
|

\
| ' '
=
|
|
.
|

\
|

|
t |
(F.17)
( ) ( ) | |
A A A
C C C 1 ln
2
0
from which a plot of versus ln t may be used to obtain relevant
parameter estimates.
362
Appendix
A
A
p
p
p
p
e
e
n
n
d
d
i
i
x
x
G
G
C
C
O
O
N
N
C
C
E
E
N
N
T
T
R
R
A
A
T
T
I
I
O
O
N
N
-
-
T
T
I
I
M
M
E
E
D
D
A
A
T
T
A
A
F
F
O
O
R
R
V
V
A
A
R
R
I
I
O
O
U
U
S
S
S
S
:
:
C
C
R
R
A
A
T
T
I
I
O
O
A
A
N
N
D
D
T
T
E
E
M
M
P
P
E
E
R
R
A
A
T
T
U
U
R
R
E
E
G.1 Propane conversion-time data for T = 773 K and S:C = 0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
363
Appendix
G.2 Propane conversion-time data for T = 823 K and S:C = 0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
G.3 Propane conversion-time data for T = 873 K and S:C = 0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
364
Appendix
G.4 Propane conversion-time data for T = 773 K and S:C = 1.2
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
G.5 Propane conversion-time data for T = 823 K and S:C = 1.2
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
365
Appendix
G.6 Propane conversion-time data for T = 873 K and S:C = 1.2
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
G.7 Propane conversion-time data for T = 773 K and S:C = 1.6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
366
Appendix
G.8 Propane conversion-time data for T = 823 K and S:C = 1.6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
G.9 Propane conversion-time data for T = 873 K and S:C = 1.6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 1 2 3 4 5 6 7 8 9 10
Time-on-stream (h)
1

-

(
C
A
/
C
A
0
)
367

You might also like