Irina Makhkamova -Complete PhD Thesis-final-For GS Submission-30 May 2011
PhD Thesis Complete
-
Upload
florian-huber -
Category
Documents
-
view
347 -
download
0
Transcript of PhD Thesis Complete
Florian Huber
Nanocrystalline copper-based mixed oxide catalysts for water-gas shift Doctoral thesis for the degree of philosophiae doctor Trondheim, August 2006 Norwegian University of Science and Technology Faculty of Natural Sciences and Technology Department of Chemical Engineering
The force of habit always tries to close our eyes.
Amazement and questioning is the beginning of
every philosophy, the encounter with the mystery the
greatest experience.
Jean Henri Fabre
Acknowledgements
i
Acknowledgements
First of all, I would like to thank my supervisors Professor Anders Holmen and
Associate Professor Hilde. J. Venvik for their support, encouragement and guidance
throughout my work with this thesis. You have given me a great opportunity to develop
and learn new things. The last four years have been very instructive, from a professional
as well as a personal point of view.
Adjunct Professor John Walmsley, Sintef Materials and Chemistry, is gratefully
acknowledged for the TEM work as well as for interesting and helpful discussions. I
also would like to thank Associate Professor Magnus Rønning for giving me the
opportunity to use X-ray absorption spectroscopy and for interesting and helpful
discussions. Professor De Chen is also gratefully acknowledged for helpful discussions.
Egil Haanæs, Department of Chemical Engineering, and Elin Nilsen, Department of
Material Technology, are gratefully acknowledged for their assistance with the TGA
and the XRD devices, respectively. Zuhair Kamil Sallom, Sintef Materials and
Chemistry, is gratefully acknowledged for his assistance with microwave spray
drying/pyrolysis. The project team at the Swiss-Norwegian Beam Lines at the European
Synchrotron Radiation Facility in Grenoble, in particular Wouter van Beek, are
gratefully acknowledged for their assistance. The Ugelstad laboratory and in particular
Torbjørn Vrålstad are gratefully acknowledged for assistance with the BET device.
Hilde Meland and Cathrine Bræin Nilsen are gratefully acknowledged for preparing
some of the Cu-Zn-based catalysts. Magnus Thomassen, Department of Material
Technology, is gratefully acknowledged for assistance with preliminary cyclic
voltammetry experiments. Edvard Bergene, Asbjørn Lindvåg, Rune Myrstad, Edd A.
Blekkan, Ingrid Aartun, Bozena Silberova and Rune Lødeng are gratefully
acknowledged for interesting discussions and/or technical assistance.
The financial support from the Research Council of Norway, the Department of
Chemical Engineering, and Statoil ASA through the Gas Technology Center NTNU-
SINTEF is gratefully acknowledged.
Acknowledgements
ii
I want to thank my colleagues at NTNU and SINTEF for their support and the
interesting discussions we have had during the last four years.
Finally, I very much want to thank my family and my friends for their patience, support
and encouragement. Thank you for being there for me – laughing with me in good times
and supporting me during hard times.
Every path is only a path, and it is not an infringement
against oneself or others to abandon it if your heart
tells you so.
Carlos Castaneda: The Teachings of Don Juan
Abstract
iii
Abstract
The present work consists of six individual chapters dealing with preparation and
characterization of copper-based mixed oxide catalysts, and their catalytic performance
under water-gas shift conditions.
Paper I investigates the effect of cerium oxide on Cu-Zn-based mixed-oxide catalysts.
The Cu reducibility as well as the TOF of the catalysts, based on N2O chemisorption,
was found not to be significantly improved by the addition of CeO2. Ceria impregnated
on the hydrotalcite-type precipitate prior to calcination was found to improve copper
dispersion and stability of the mixed oxide catalyst.
Paper II comments on passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts. These
catalysts are often subjected to passivation procedures prior to characterization to retain
the reduced state of the active metal. Passivation with N2O or O2 to create a protective
oxide layer results in a certain degree of sub-surface oxidation. The extent of bulk
oxidation depends on the type of oxidant as well as the size of the metal particles, as
shown for copper catalysts. Passivation strategies that involve CO and/or CO2 without
the formation of significant amounts of coke or wax around the metal particles may
comprise both formation of carbonaceous species and an oxide layer. Encapsulation of
reduced metal particles by a protective layer of carbon is found to efficiently preserve
the metallic state, as demonstrated for metallic nickel and iron with carbon nanofibers.
Passivation and encapsulation of reduced catalysts are not ideal strategies for
characterization of reduced systems, but require evaluation by a parallel in situ approach.
Paper III deals with preparation and characterization of Cu-Ce-Zr mixed oxide catalysts.
Cu0.23Ce0.54Zr0.23-mixed oxides were prepared by homogeneous co-precipitation with
urea. The resulting material exhibits high surface area and small nanocrystalline primary
particles with fluorite-type structure. STEM-EDS analysis shows that Cu and Zr are
inhomogeneously distributed throughout the ceria matrix. The EXAFS analysis
indicates the existence of CuO-type clusters inside the ceria-zirconia matrix. This type
of mixed oxide materials should therefore be described as heterogeneous single-phase
Abstract
iv
materials rather than homogeneous solid solutions. The pore structure and surface area
of the mixed oxides are affected by preparation parameters during both precipitation
(stirring) and the following heat treatment (drying and calcination). TPR measurements
show that most of the copper is reducible and not inaccessibly incorporated into the bulk
structure. Reduction-oxidation cycling shows that the reducibility improves from the
first to the second reduction step, probably due to a local phase segregation in the
metastable mixed oxide with gradual copper enrichment at the surface of the Ce-Zr
particles during heat treatment.
Paper IV compares Cu-Zn-Al and Cu-Ce-Zr mixed oxide catalysts under water-gas shift
reaction conditions. The catalysts were prepared by two different methods, co-
precipitation and flame spray pyrolysis. Cu-Ce-Zr catalysts are found not to be
generally superior to Cu-Zn-Al catalysts in terms of activity or short-term stability.
Instead, the difference in activity seems to be related to structural characteristics of the
catalysts as well as the reaction conditions. The apparent activation energy of the Cu-
Ce-Zr MMO catalysts appears to be less affected by increased concentrations of CO2
and H2 than the Cu-Zn-Al MMO catalysts.
Paper V investigates the effect of carbon nanofibers as catalyst dispersant. Carbon
nanofiber (CNF) and Cu-Ce-Zr mixed metal oxide containing nanocomposite catalysts
have been prepared by homogeneous co-precipitation with urea. The CNF-containing
nanocomposite catalysts exhibit similar overall catalytic activity and stability as the
corresponding CNF-free catalyst. 13 wt% of the MMO could be replaced by CNF
without decreasing the overall activity and stability of the catalyst. The specific activity
of the nanocomposites based on the total metal oxide content is similar or higher than
the activity of the CNF-free material, depending on the CNF content. Similar activation
energies are obtained for the CNF-free and CNF-containing materials. CO
chemisorption studies showed no significant CO chemisorption on CNF. CNF may be
regarded as inert dispersant material improving the dispersion of the mixed oxide
particles.
Abstract
v
Paper VI deals with the reducibility of copper in Cu-Ce-Zr and Cu-Zn-Al mixed oxide
catalysts prepared by homogeneous co-precipitation with urea, and the impact of Pt
impregnated on these mixed oxide catalysts. The WGS activity of Cu-Ce-Zr and Cu-Zn-
Al mixed oxide catalysts can be related to the Cu reducibility in these catalysts. Low-
temperature reducibility correlates with low-temperature activity. Pre-reduction is not
absolutely necessary when performing the WGS reaction above the Cu reduction
temperature of the catalyst. An adequate reduction procedure may, however, be applied
in order to optimize the CO conversion. Pt had no significant effect on the Cu-Ce-Zr
mixed oxide catalyst, but altered the properties of the Cu-Zn-Al mixed oxide catalyst. Pt
shifted the Cu reduction to lower temperatures, indicating the existence of an interaction
between Pt and Cu in the bimetallic catalyst. The effect of Pt on the WGS activity and
stability was dependent on the pre-treatment procedure as well as the reaction
conditions.
Table of contents
vi
Table of contents
Acknowledgement i
Abstract iii
Table of contents vi
List of publications and presentations viii
List of symbols and abbreviations xi
1. Introduction 1 1.1. Hydrogen society 1
1.2. Fuel processing for fuel cell applications 1
1.3. Scope of the present work 3
2. Theory and literature 5 2.1. Water-gas shift reaction (WGS) 5
2.2. WGS catalysts for fuel processing 8
2.2.1. Catalyst requirements 8
2.2.2. New catalyst materials 9
2.3. Mixed metal oxide catalysts 11
3. Experimental 13 3.1. Catalyst preparation 13
3.2. Characterization of catalyst materials 13
3.2.1. Inductively coupled plasma atomic emission spectroscopy (ICP-AES) 13
3.2.2. X-ray diffraction (XRD) 14
3.2.3. Nitrogen adsorption-desorption isotherms 14
3.2.4. Thermogravimetric analysis (TGA) 15
3.2.5. Volumetric TPR measurements 15
3.2.6. Transmission electron microscopy (TEM) 15
3.2.7. Transmission X-ray absorption spectroscopy (XAS) 16
3.3. Setup for water gas shift testing 18
Table of contents
vii
4. Results and discussion 23 4.1. Pre-studies 23
4.1.1. Total pore volume and pore size distribution 23
4.1.2. XRD particle size estimates with Win-crysize 25
4.1.3. Reproducibility of the WGS testing – By-pass effects 26
4.2. Summary of results and discussion 30
4.2.1. Relating catalyst structure and composition to the water-gas shift
activity of Cu-Zn-based mixed-oxide catalysts (Paper I) 30
4.2.2. Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based
catalysts (Paper II) 34
4.2.3. Preparation and characterization of nanocrystalline, high-surface
area Cu-Ce-Zr mixed oxide catalysts from homogeneous
co-precipitation (Paper III) 38
4.2.4. Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for
water-gas shift (Paper IV) 42
4.2.5. Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for water-gas shift:
Carbon nanofibers as dispersing agent for the mixed oxide
particles (Paper V) 43
4.2.6. The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide
catalysts for water-gas shift (Paper VI) 44
5. Concluding remarks 47
6. Suggestions for further work 50
References 51
List of appendices 62
List of publications and presentations
viii
List of publications and presentations
Publications
This thesis is based on the following articles, referred to in the text by their roman
numerals. Reprints of the published articles or manuscripts are given as Appendices I –
VI.
I M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Relating
catalyst structure and composition to the water-gas shift activity of Cu-Zn-
based mixed-oxide catalysts, Catalysis Today 100 (2005), 249-254.
II F. Huber, Z. Yu, S. Lögdberg, M. Rønning, D. Chen, H. Venvik, A. Holmen,
Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts,
Catalysis Letters, in press.
III F. Huber, H. Venvik, M. Rønning, J. Walmsley, A. Holmen, Preparation and
characterization of nanocrystalline, high-surface area Cu-Ce-Zr mixed oxide
catalysts from homogeneous co-precipitation, Manuscript in preparation.
IV F. Huber, H. Meland, M. Rønning, H. Venvik, A. Holmen, Comparison of Cu-
Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift, submitted.
V F. Huber, Z. Yu, J. Walmsley, D. Chen, H. Venvik, A. Holmen,
Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for water-gas shift: Carbon
nanofibers as dispersing agent for the mixed oxide particles, Applied Catalysis
B: Environmental, accepted.
VI F. Huber, J. Walmsley, H. Venvik, A. Holmen, The effect of platinum in Cu-
Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift, Manuscript in
preparation.
List of publications and presentations
ix
Presentations
The following presentations were given at national and international meetings based on
results obtained during this work:
H. Venvik, T. Gjervan, I. Aartun, K. A. Johnsen, F. Huber, E. Tangerås, H. Dyrbeck, M.
Fathi, A. Holmen: Propane to hydrogen via stepwise catalytic reactions. Oral
presentation, NKS Catalysis Symposium, Hafjell, November 28 – 29, 2002.
F. Huber, D. Chen, M. Rønning, H. Meland, H. Venvik, A. Holmen: Activity of different
hydrotalcite-based copper catalysts in the water-gas shift reaction. Poster presentation,
International Summer School: Towards a Hydrogen-based Society, Humlebæk,
Denmark, August 9 -15, 2003 and EuropaCat VI, Innsbruck, Austria, August 31 -
September 04, 2003.
M. Rønning, H. Meland, F. Huber, D. Chen, H. Venvik, A. Holmen: Effect of catalyst
composition on the structure and activity of Cu-based water-gas shift catalysts. Poster
presentation, EuropaCat VI, Innsbruck, Austria, August 31 - September 04, 2003.
F. Huber, M. Rønning, H. Meland, D. Chen, H. Venvik, A. Holmen: Structure-activity
relations in water-gas shift catalysts from hydrotalcite precursors. Oral presentation,
NKS Catalysis Symposium, Bergen, November 20 – 21 2003.
M. Rønning, F. Huber, H. Meland, D. Chen, H. Venvik, A. Holmen: Relating catalyst
structure and composition to the water-gas shift activity of Cu-Zn-based mixed-oxide
catalysts. Oral presentation, 11th Nordic Symposium on Catalysis, Oulu, Finland, Mai
23 - 25, 2004.
F. Huber, M. Rønning, H. Meland, D. Chen, H. Venvik, A. Holmen: Thermogravimetric
analysis of copper-based catalysts. Oral presentation. Thermal Analysis Seminar,
Trondheim, June 24, 2004 and Norwegian Hydrogen Seminar, Kvitfjell, November 15
– 16, 2004.
List of publications and presentations
x
F. Huber, H. Venvik, A. Holmen: Cu-based mixed metal oxide catalysts for clean fuel
applications – On preparation, characterization and water-gas shift testing. Oral
presentation, Haldor Topsøe A/S, Lyngby, Denmark, Mai 10, 2006.
F. Huber, H. Meland, Z. Yu, M. Rønning, J. Walmsley, D. Chen, H. Venvik, A.
Holmen: Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for clean fuel applications.
Oral presentation, 12th Nordic Symposium on Catalysis, Trondheim, Norway, Mai 28 -
30, 2006.
F. Huber, Z. Yu, J. Walmsley, M. Rønning, H. Venvik, D. Chen, A. Holmen:
Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for clean fuel applications: Carbon
nanofibers as dispersant for the mixed oxide particles. Oral presentation, ISNEPP 2006,
Hong Kong, China, June 18 - 21, 2006.
List of symbols and abbreviations
xi
List of symbols and abbreviations
Latin symbols
B Line width at half maximum (FWHM) [° or radian]
Be Instrumental line broadening [° or radian]
BET BET surface area [m2/g]
dC Size of the core of passivated metal particles [nm]
dP Particle size estimate [nm]
E Energy [eV]
Ea Activation energy [kJ/mole]
E0 Energy shift to correct for deviation from theoretical edge
value [eV]
I0 Incident X-ray intensity [-]
It Transmitted X-ray intensity [-]
k Wave number of the photoelectron [Å-1]
K Scherrer constant [-]
Keq Equilibrium constant [-]
N Coordination number [-]
R Distance of the central atom to the neighbouring atom (shell) [Å]
stv Surface-to-volume ratio [m2/m3]
T Temperature [K or °C]
xp Fraction of passivated metal atoms [-]
Greek symbols
λ Wavelength [Å]
∆σ2 Debye-Waller factor (disorder in the neighbour distance) [Å2]
Θ Bragg angle [° or radian]
Abbreviations
AFAC Amplitude reduction factor
BET Brunauer Emmet Teller
BJH Barrett Joyner Halenda
List of symbols and abbreviations
xii
CCD Charge-coupled device
CNF Carbon nanofibers
DFT Density functional theory
EDS Energy dispersive X-ray spectroscopy
EG Ethylene glycol
EPR Electron paramagnetic resonance spectroscopy
ESRF European synchrotron radiation facility
EXAFS Extended X-ray absorption fine structure
FT Fourier transform
FWHM Full width half maximum
HCP Homogeneous co-precipitation
HTS High temperature water-gas shift
ICP-AES Inductively coupled plasma atomic emission spectroscopy
LFC Liquid flow controller
LTS Low temperature water-gas shift
MCFC Molten carbonate fuel cell
MFC Mass flow controller
MMO Mixed metal oxides
NLDFT Non-local density functional theory
NPD Nitrate precursor decomposition
OSC Oxygen storage capacity
PAFC Phosphoric acid fuel cell
PCA Principal component analysis
PEMFC Proton exchange membrane fuel cell
or polymer electrolyte membrane fuel cell
PSD Pore size distribution
RMS Root mean square
SOFC Solid oxide fuel cell
SNBL Swiss-Norwegian beam line
SSITKA Steady-state isotopic transient kinetic analysis
STEM Scanning transmission electron microscopy
TCD Thermal conductivity detector
List of symbols and abbreviations
xiii
TEM Transmission electron microscopy
TGA Thermogravimetric analysis
TOF Turnover frequency
TOS Time on stream
TPR Temperature programmed reduction
TPO Temperature programmed oxidation
US Ultrasonic field
VPI Effective core-hole lifetime
WGS Water-gas shift
XANES X-ray absorption near edge structure
XAFS X-ray absorption fine structure
XAS X-ray absorption spectroscopy
XLBA X-ray line broadening analysis
XPS X-ray photoelectron spectroscopy
XRD X-ray diffraction
Introduction
1
1. Introduction 1.1. Hydrogen society
The predicted depletion of fossil fuels and the increase in energy demand require the
development of alternative low-emission and low-carbon energy systems [1,2].
Hydrogen is expected to play a major role as a future energy carrier, in a society based
on the use of renewable energy to split water into oxygen and hydrogen, and storage and
use of hydrogen as a fuel. Fuel cells are expected to provide electricity and power
vehicles. The only emission from the fuel cell will be pure water [3,4]. It will take more
than a decade before renewable energy sources can compete effectively with fossil fuels
[1], and a transitional approach is therefore necessary to produce hydrogen. Conversion
of natural gas is currently the most economic way to produce hydrogen [4-7].
1.2. Fuel processing for fuel cell applications
Figure 1 illustrates the general concepts of processing gaseous, liquid, and solid carbon-
containing fuels for fuel cell applications. Pure hydrogen, reformate (hydrogen-rich gas
from fuel reforming), and methanol are the fuels available for current fuel cells [8].
Sulphur compounds in carbon-containing fuels poison the catalysts in the fuel
processing and in the fuel cells and must be removed. Synthesis gas, a mixture of CO
and H2, can be generated from reacting the fuel with steam, oxygen or a mixture of both
[9-11]. The product, often referred to as reformate, can be used directly for high-
temperature fuel cells such as SOFC and MCFC. Hydrogen is the preferred fuel for low-
temperature fuel cells such as PEMFC and PAFC, which can be obtained by fuel
reformulation on-site for stationary applications or on-board for automotive applications
[6,12-14].
Using fuel cells to power vehicles, given the cost and size requirements, will require
major breakthroughs in new technologies. The earliest market for fuel cells will be for
stationary power generation. Fuel cells must first be successful for this before they can
be applied to transportation applications. Figure 2 shows that stationary application is
on the critical path for transportation. Also shown is the parallel path to
commercialization for portable power applications, where methanol is likely the fuel of
choice. Molten carbonate and solid oxide fuel cells are also being developed for high-
Introduction
2
power applications (> 250 kW). These require less extensive fuel reformers because this
function occurs partly within the anode compartment [5].
Figure 1. The concepts and steps for fuel processing of gaseous, liquid and solid fuels
for high-temperature and low-temperature fuel cell applications [6].
Figure 2. The road to commercialization for fuel cells [5].
Introduction
3
Distributed power applications can use the current natural gas and commercial propane
infrastructure to generate H2 in homes and at industrial sites. Once this technology is
implemented in the mass market, transportation applications will be more feasible, but
breakthroughs in cost, size, and infrastructure are essential. It is expected that
automotive applications will require on-board storage of H2 distributed from a local
service station. The H2 will be generated from natural gas or from a liquid fuel designed
specifically for fuel cells (low sulphur, high paraffin content, without concern of
optimizing octane or cetane numbers) [9-13]. On-board reformers, using liquid fuels,
are viewed as a transitional solution until new H2 storage technologies are developed [5].
Concerning the fuel processor technology, microreactor, membrane reactor and
monolith reactor concept represent promising approaches for stationary and mobile
applications. For more information on this matter, it is referred to relevant literature
[5,15-27].
A technical challenge for fuel processing catalysts is the development of active, poison-
resistant materials that will result in small catalyst volumes, reduced start-up times,
durability under steady-state and transient conditions at the required temperatures, cost
reductions, and versatility to variations in fuel/feed composition [5,28].
1.3. Scope of the present work
This work has been a part of the Strategic Institute Program “Advanced catalyst/reactor
systems for conversion of hydrocarbons to hydrogen for fuel cells”, supported by the
Research Council of Norway through Grant No. 140022/V30 (RENERGI). The program
was carried out in a joint collaboration between SINTEF Materials and Chemistry and
the Department of Chemical Engineering at the Norwegian University of Science and
Technology (NTNU). The program has focused on synthesis gas production by partial
oxidation or oxidative steam reforming in microstructured and monolithic reactors
[16,17,29,30], and the water-gas shift reaction using conventional and new WGS
catalyst formulations. The WGS reaction has also been studied in a membrane reactor
with a self-supported, hydrogen permeable Pd-Ag membrane [31].
Introduction
4
The aim of the present work was to gain insight into certain aspects related to copper-
based catalysts used for clean fuel applications and small-scale hydrogen production
systems, including preparation, characterization and catalytic performance. The water-
gas shift reaction was used as test reaction. The results of the present study are however
not only relevant to research on the water-gas shift reaction, but in fact also relevant to
other heterogeneous catalytic reactions, since Cu-based catalysts are applied in several
areas of heterogeneous catalysis. Moreover, the results may also be of importance to
catalytic systems where another active metal such as Ni or Co is combined with the
support materials chosen here.
Theory and literature
5
2. Theory and literature 2.1. Water-gas shift reaction (WGS)
The water-gas shift is the reaction of steam with carbon monoxide to produce carbon
dioxide and hydrogen gas. It is a reversible and exothermic chemical reaction:
CO + H2O ↔ CO2 + H2 ∆H0298 = - 41.2 kJ/mol (1)
The WGS is one of the oldest catalytic processes employed in the chemical industry
[32]. The first report on the water-gas shift reaction was published in a British gas
patent from 1888 [33]. Already then it was described how the passage of carbon
monoxide and steam over red-hot refractory material was able to produce carbon
dioxide and hydrogen. Since the early 1940s the WGS reaction has represented an
important step in the industrial production of hydrogen or synthesis gas. The essential
role of the industrial WGS reaction is to enhance the production of hydrogen and
remove CO for refinery hydroprocesses, ammonia synthesis and bulk storage and
redistribution of hydrogen, as well as adjustment of H2/CO ratios in the production of
methanol and synthetic, liquid hydrocarbons through the Fischer-Tropsch synthesis.
Although the WGS reaction is not the primary reaction, it also plays a role in methanol
synthesis, reforming of hydrocarbons, Fischer-Tropsch synthesis, automotive exhaust
catalysis, and selective CO oxidation for removal of trace CO [34].
The WGS is equilibrium controlled, and since the reaction is equimolar, the effect of
pressure on the thermodynamics is minimal considering the range of pressure used for
fuel processing [34]. However, pressure may affect adsorption/desorption of the gas
molecules involved [35] and under certain conditions enhance the rate of reaction [36-
38]. The equilibrium constant Keq as a function of temperature, as estimated from
thermodynamic data [39], is shown in Figure 3. Keq decreases with increasing
temperature due to the exothermic nature of the reaction. Thus, the forward WGS
reaction is thermodynamically favoured by low temperature. Figure 4 shows the
equilibrium CO conversion for two different reactant compositions, a feed containing 5
mole-% CO and 25 mole-% H2O, and a simulated reformate feed containing 5 mole-%
CO, 25 mole-% H2O, 12 mole-% CO2 and 35 mole-% H2. This corresponds to feed
Theory and literature
6
compositions used for some of the activity tests performed in this thesis. The
equilibrium conversion in Figure 4 decreases with temperature, and the addition of CO2
and H2 to the reactants shifts the equilibrium towards the reactants.
1
10
100
1000
10000
100 150 200 250 300 350 400 450 500
temperature [ C ]
equi
libriu
m c
onst
ant [
- ]
Figure 3. Equilibrium constant Keq as function of temperature.
From a thermodynamic perspective, the extent of the forward WGS reaction is
maximized at low temperature, high H2O and low H2 concentrations. However, most
catalysts available today are kinetically limited at the low temperatures needed for
nearly complete CO conversion, thus, requiring large reactor volumes. Two or more
consecutive stages of WGS are often used to take advantage of both kinetics and
thermodynamics. By operating an HTS catalyst at higher temperatures the favourable
kinetics can be exploited and the volume of the catalyst can be minimized. By cooling
the syngas between the HTS and LTS stages an active catalyst can take advantage of the
thermodynamic equilibrium at low temperature [34].
The method of producing the syngas affects the WGS equilibrium. Autothermal
reforming produces a syngas with lower H2 concentration due to the dilution of nitrogen
as compared to steam reforming. The lower H2 concentration improves the equilibrium
Theory and literature
7
CO conversion whereas the high H2 concentration expected from conventional steam
reforming lowers the WGS equilibrium conversion [34].
0,00
0,20
0,40
0,60
0,80
1,00
100 150 200 250 300 350 400 450 500
temperature [ C ]
CO
con
vers
ion
[ - ]
CO-H2O
CO-H2O-CO2-H2
Figure 4. Equilibrium CO conversion for two different feed conditions: (1) 5 mole-%
CO and 25 mole-% H2O; (2) 5 mole-% CO, 25 mole-% H2O, 12 mole-% CO2 and 35
mole-% H2 (simulated reformate feed), corresponding to feed compositions used for the
activity tests reported in papers IV, V and VI.
Because of its industrial significance, the kinetics and mechanisms of the WGS reaction
have been extensively studied in the past [34,40-47]. Two types of mechanisms have
been proposed for the main catalyst systems, the associative and the regenerative (redox)
[40]. The associative mechanism involves reaction through an intermediate surface
species such as formate, carbonate, or bicarbonate. For instance, water may adsorb and
dissociate to form OH that reacts with CO to form HCOO, which eventually
decomposes to H2 and CO2. In the redox mechanism, successive oxidation and
reduction of the surface occurs. Adsorbed water dissociates into oxygen and hydrogen,
and the oxygen is then titrated by CO [38]. Experimental evidence exists supporting
both reaction pathways. It is possible that either reaction mechanism could proceed and
that the relative rates of the two pathways could depend on the catalyst and the reaction
Theory and literature
8
conditions [40]. From a microkinetic analysis on a Cu-based catalyst, it was concluded
that the associative mechanism may be dominant at lower temperatures while the redox
mechanism becomes important at higher temperatures [38,48]. Ovesen et al. [49]
investigated the microkinetics of the WGS over Cu at industrial conditions based upon
the redox mechanism. At low pressure, formate and carbonate type species could be
excluded from the model while at high pressure the synthesis and hydrogenation of
formate type surface species had to be included.
2.2. WGS catalysts for fuel processing
2.2.1. Catalyst requirements
Traditional industrial catalysts used for WGS are iron-based for HTS and copper-based
for LTS. Typical compositions are Fe3O4/Cr2O3 and Cu/ZnO/(Al2O3 or Cr2O3),
respectively [34,40,50-52]. Cu/ZnO catalysts suffer from the drawback of being
pyrophoric (i.e. oxidize spontaneously in air with extensive heat release) and susceptible
to poisons. HTS Fe-Cr catalysts exhibit low activity and suffer from reduced
equilibrium CO conversion at higher temperatures [28].
The requirements for fuel processing are relatively different from the ones for the
industrial use of WGS catalysts [5,19,28,34]. Industrially, the Fe/Cr and Cu/Zn/Al
catalysts are slowly activated by a controlled reduction process. The catalysts must be
purged with inert gas to prevent condensation and reoxidation upon shutdown. If any of
these conditions are not met, the performance of the catalyst is significantly limited.
Simple operation is essential for WGS catalysts for fuel processing applications. Table 1
gives a list over desired requirements of WGS catalysts used for production of hydrogen
for fuel cells. The requirements for mobile applications must compete with the
standards set by internal combustion engines, and the operability, size, weight and cost
targets are therefore rigorous. For stationary applications the catalyst attributes are less
constrained and larger volumes and higher costs are acceptable to compete with the cost
of electricity from the grid [34].
Theory and literature
9
Table 1. WGS catalyst requirements for mobile and stationary applications [34].
WGS catalyst attribute Mobile application Stationary application
Volume reduction Critical <0.1 l kW-1 Not as constrained
Weight reduction Critical <0.1 kg kW-1 Not as constrained
Cost Critical <$1 kW-1 Not as critical
Rapid response Critical <15 s Load following
Selectivity Critical Important
No reduction required Critical Important
Oxidation tolerant Critical Important
Condensation tolerant Important Important
Nonpyrophoric Important Eliminate purging
Attrition resistance Critical No constraint
Pressure drop Important Important
Poison tolerant Desired Desired
Conventional Cu/Zn/Al and Fe/Cr WGS catalysts may find a place in fuel processing
due to their proven success with long-term performance. Fe/Cr catalysts are known to
be poison tolerant, selective and stable at high temperature. The low temperature
activity of commercial Cu/Zn/Al catalysts represents a benchmark for new
developments [5,19,34,35,53-55]. Even though conventional Cu-based catalysts must be
treated with special care, engineered solutions are possible that take advantage of the
low temperature activity, low cost and proven performance [13,26,34,56-58].
2.2.2. New catalyst materials
Two trends can be distinguished in the development of new, non-pyrophoric WGS
catalysts for fuel processing [34]: low cost base metal formulations and precious metal
formulations. Alternative base metal formulations include transition metal carbides,
cobalt-vanadium oxides, Co on perovskite, Cu and Ni on ceria, Cu-Al spinel, and other
proprietary base metal formulations. Precious metal formulations are based on platinum
Theory and literature
10
group metals or gold supported on (mixed) metal oxides. For further reading on new
developments it is referred to [5,19,28,34,51,59].
The most prominent catalyst formulations are combinations of copper or precious
metals and metal oxides. Promising alternative oxide materials to ZnO (or Al2O3) are
CeO2 [51,52,59], Fe2O3 [60,61], TiO2 [59-62], ZrO2 [24,63,64], La2O3 [65] and V2O3
[64].
A possible strategy for discovering improved WGS catalyst materials based on
metal/metal oxide combinations is outlined in Figure 5.
Figure 5. Possible strategy for improving WGS catalyst materials based on metal/metal
oxide formulations.
A proper combination of metal and metal oxide may be the first task, assuming that the
supporting oxide material affects activity and/or stability of the catalyst [66]. The
combination of Cu and ZnO (or Al2O3) is a classical example. A large number of
alternative formulations have been proposed over the last years, including Cu and CeO2
or ZrO2 [51,52,63], Ru and ZrO2 [24,64] or Fe2O3 [67], Au and CeO2 [68,69], TiO2
[70,71] or Fe2O3 [72], Pt and CeO2 [53,59,73] or TiO2 [59,74], Pd and CeO2 [75], and
more. The performance of the metal/metal oxide formulation can be improved through
optimization of metal loading and metal dispersion on the oxide surface, and further
enhanced by addition of other elements. It has been reported that the catalyst
Improvement of WGS catalyst materials
Metal / Metal Oxide
Alloying Doping Bimetallic systems Mixed oxides Alkali promoter
Theory and literature
11
performance can be improved by using mixed metal oxides instead of single metal
oxides [52,56,76-84], as also applied for the classical CuO/ZnO/Al2O3 (or SiO2)
formulation [56,84]. The addition of a second metal resulting in alloy formation or
bimetallic systems can also be applied to improve the performance [35,85-89]. Finally,
the deposition of alkali metals on the metal surface or oxide support is known to
improve the catalyst performance [90-96] and has been successfully employed
commercially.
2.3. Mixed metal oxide catalysts
Mixed metal oxides (MMO) containing copper, cerium and zirconium are applied in
several areas of heterogeneous catalysis. Ce-Zr mixed oxides are extensively used in
three-way catalysts [79,97]. Cu-Ce-Zr-based mixed oxides of various composition are
applied within the field of hydrogen production: Water-gas shift [51,52,85,98], steam
reforming of methanol [99-102] and selective oxidation of CO [98,103-106]. In addition,
they are used as NO reduction catalyst [107], for oxidation of methane [108], wet
oxidation of phenol [109] and acetic acid [110], methanol synthesis [111,112], direct
oxidation of hydrocarbons in solid-oxide fuel cells (SOFC) [113] and storage of reactive
hydrogen for alkadiene hydrogenation [114].
Copper - in its reduced state - is typically regarded as the active catalyst component in
MMO materials, except in SOFCs, where copper is considered to function as electronic
conductor [113]. It is considered the most active metal for WGS on inert support
materials [35,115]. Ceria acts as a reducible oxide support, enhancing the catalytic
activity via catalytic metal-support interaction and/or better dispersion of the active
metal component [100,116]. The main advantage of ceria is its oxygen storage capacity
(OSC). CeO2-containing oxides are able to adsorb and release oxygen under oxidizing
and reducing conditions, respectively, according to the reaction [117]
22 2
2 2
. ( / ). ( / ) x
red H COCeO CeO
ox H O CO − (2)
Theory and literature
12
Ceria is also found to stabilize the catalyst against deactivation [100,116] due to a
higher thermal stability of the material and/or better dispersion of the active metal. ZrO2
is also known to improve catalytic activity and stability of MMO supported catalysts
[77,101]. Zirconium added to ceria to form Ce-Zr mixed oxides inhibits the thermal
sintering of CeO2 [79,117-119]. Introduction of Zr into the ceria lattice furthermore
enhances the reducibility of ceria [120-122], resulting in improved catalytic activity of
MMO supports compared to single oxide supports [81,123]. The amount of Zr also
affects surface area and crystallite size of the MMO [123,124] together with the
calcination temperature [125]. Cu inclusion in the fluorite lattice of ceria has been
reported to improve the reducibility of ceria [126,127] and affect particle size and
surface area [112,128], in addition to its function as an active catalyst material.
The catalytic performance of MMO thus depends on the interaction of the single
components. A pre-requisite for an efficient interaction is a homogeneous distribution of
the single components throughout the material without extensive segregation. The
catalytic activity often scales proportionally with the surface area of the active
components [85,129,130] since the reaction takes place at the surface of the catalyst
while the bulk is inaccessible to reactants. Obtaining a homogeneous distribution of the
components in conjunction with a high surface area of the material are therefore
important targets during catalyst preparation.
Co-precipitation of metal (hydrous) oxides in aqueous solution at high pH has been
applied successfully for many different metal oxide catalyst formulations. It has been
stated that co-precipitation results in more active and/or stable catalysts than
impregnation methods [131-135], because of a more homogeneous distribution. For
impregnation-deposition methods the extent of interaction between the different
components depends on the surface area of the support material and the amount of
material impregnated or deposited. The formation of separate phases at higher loadings
is likely - especially when the surface of the support is saturated [85]. At low metal
loadings, however, impregnation-deposition methods may be superior to co-
precipitation, since burying of active components in the bulk is avoided. Thus, the
optimum metal loading may be lower [133], presuming a good metal dispersion.
Experimental
13
3. Experimental 3.1. Catalyst preparation
Different methods have been used for the preparation of the copper-based mixed oxide
catalysts. Most of the preparation methods are described in detail in the corresponding
papers. This chapter contains therefore only additional details on preparations that are
not sufficiently described in the papers.
For the precipitation method used in Paper I, the metal nitrate solution contained 0.1035
mole Cu(NO3)2·3H2O, 0.1382 mole Zn(NO3)2·6H2O and 0.0750 mole Al(NO3)3·9H2O in
approx. 300 ml deionized water. The alkaline solution contained 0.075 mole Na2CO3
and 0.600 mole NaOH in approx. 400 ml deionized water. The nitrate solution was
added to the alkaline solution over a period of 60 – 90 min. Then the pH was adjusted
between 8 and 9. For incipient wetness impregnation with Ce-nitrate, 0.200 g
Ce(NO3)3·6H2O in 0.6 ml water and 0.274 g Ce(NO3)3·6H2O in 1.3 ml water were used
for impregnation before and after calcination, respectively. Additional information is
given in [136].
Catalyst CuZn-CP, used in Paper IV and mentioned in Paper II (estimated Cu particle
size 14 nm), was prepared by co-precipitation [56,137]. The metal nitrate solution
contained 40.3 g Cu(NO3)2·3H2O, 59.8 g Zn(NO3)2·6H2O and 62.5 g Al(NO3)3·9H2O in
1 l of distilled water. The alkaline solution contained 96 g (NH4)CO3 in 1l of distilled
water. The two solutions were mixed under stirring in a volumetric ratio of 1:1. The
light blue precipitate was filtered, washed with distilled water and dried over night at 90
°C. The dry matter was crushed and calcined at 350 °C for 1 h (heating rate: 3 °C/min).
3.2. Characterization of catalyst materials
3.2.1. Inductively coupled plasma atomic emission spectroscopy (ICP-AES)
ICP-AES was used to determine the actual catalyst composition. The analyses were
performed by Molab as. The samples were dissolved in hydrochloric acid before
analysis without any visible residues.
Experimental
14
3.2.2. X-ray diffraction (XRD)
XRD spectra for crystallite size estimation and phase identification were recorded on
Siemens diffractometers D-5000 (monochromatic CuKα-radiation) and D-5005
(dichromatic CuKα+β-radiation). Particle size estimates were calculated by X-ray line
broadening analysis (XLBA, line width at half maximum, B) using the Scherrer
equation [138]. In addition, software-based XLBA was used to determine average
crystal size estimates and crystal size distributions. The analysis was performed in two
steps. Selected experimental XRD peaks were simulated by means of the software
package Profile [139] using the Pearson VII model function. The contribution of CuKβ
to the peak intensities is removed in this step. The program Win-crysize [140], utilizing
the Warren-Averbach method [141], was then used to estimate the crystallite size taking
into account contributions from microstrain (scaled as mean square root of the average
squared relative strain). Contributions from instrumental line broadening were removed
using LaB6 as reference, since this material does not exhibit any broadening due to
crystallite size or strain. The width of the peaks is hence due to instrumental broadening.
3.2.3. Nitrogen adsorption-desorption isotherms
N2 adsorption-desorption isotherms were measured using a Micrometrics TriStar 3000
instrument. The data were collected at -196 °C. The BET surface area was calculated by
the BET equation in the relative pressure interval ranging from 0.01 to 0.30. The pore
volume was estimated by the Barrett-Joyner-Halenda (BJH) method [142] as the
adsorption cumulative volume of pores between 1.7 nm and 300.0 nm width. This
method is based on the assumption of cylindrical pores, and the capillary condensation
in the pores is taken into account by the classical Kelvin equation. The pore size
distributions were calculated by non-local density functional theory (NLDFT, original
DFT model with N2 [143], DFT Plus software package [144]) assuming a slit-like pore
geometry. For comparison, pore size distributions were also calculated from the
Harkins-Jura model (cylindrical geometry), a classical method based on the Kelvin
equation [145].
Experimental
15
3.2.4. Thermogravimetric analysis (TGA)
Thermogravimetric analysis was performed with a Thermogravimetric Analyser TGA
Q500 (TA Instruments) and a Thermogravimetric Analyser TGA 7 (Perkin Elmer).
Figure 6 outlines the basic components and the measurement principle of the TGA
devices.
Figure 6. Basic components and measurement principle of the TGA devices.
TGA was used for temperature-programmed oxidation (TPO) and reduction (TPR)
studies, as well as determination of the degree of reoxidation of copper upon passivation
and Cu dispersion measurements by means of selective oxidation via N2O surface
titration [146-148] taking into account the bulk oxidation of copper [149-151].
3.2.5. Volumetric TPR measurements
In paper VI, TPR experiments were carried out with the volumetric device CHEMBET
3000 (Quantachrome Instruments), based on the analysis of the gas stream by means of
a thermal conductivity detector (TCD).
3.2.6. Transmission electron microscopy (TEM)
TEM data were recorded on a JEOL 2010F transmission electron microscope. Small
amounts of catalyst sample were placed in sealed glass containers containing ethanol
Purge gas Reactant gas
Weight chamber
Furnace
Outlet
Thermocouple
Experimental
16
and immersed in an ultrasonic bath to disperse the individual particles. The resulting
suspension was dropped onto a holey carbon film, supported on a titanium mesh grid,
and dried. Conventional TEM images were recorded onto a CCD camera. Scanning
transmission electron microscope (STEM) images were acquired with a nominal probe
size of approx. 0.7nm. Bright field and dark field STEM images were acquired. Energy
dispersive x-ray spectroscopy (EDS) analysis and mapping were performed using an
Oxford Instruments INCA system. Drift compensation was employed to correct for
movement of the sample during the time taken for the acquisition of maps.
3.2.7. Transmission X-ray absorption spectroscopy (XAS)
XAS data were collected at the Swiss-Norwegian Beam Lines (SNBL) at the European
Synchrotron Radiation Facility (ESRF), France. Spectra were obtained at the Cu K-edge
(8.979 keV) using a channel-cut Si(111) monochromator. Higher order harmonics were
rejected by means of a chromium-coated mirror aligned with respect to the beam to give
a cut-off energy of approximately 15 keV. The maximum resolution (∆E/E) of the
Si(111) band pass is 1.4 x 10-4 using a beam of size 0.6 x 7.2 mm. Ion chamber
detectors with their gases at ambient temperature and pressure were used for measuring
the intensities of the incident (I0) and transmitted (It) X-rays.
The amounts of material in the samples were calculated to give an absorber optical
thickness close to 2 absorption lengths. The samples were ground and mixed with boron
nitride to achieve the desired absorber thickness. For X-ray absorption near edge
structure (XANES) measurements, the samples were placed in an in situ reactor-cell
(Figure 7)[152] and reduced in a mixture of 5 vol-% H2 or CO in He (purity: 99.995%;
30 ml/min flow rate at ambient temperature) heating from room temperature to 350 °C
by 6 °C/min. XANES profiles were collected during heating of the samples to follow
the reduction of CuO in H2 or CO.
Experimental
17
Figure 7. Lytle type in situ XAS cell for transmission and fluorescence detection [152].
The recorded XAS spectra were energy-calibrated, pre-edge background subtracted
(linear fit) and normalized using the WinXAS software package [153]. The software
package was also used for XANES analysis. Principal component analysis (PCA) and
linear combination XANES fits were applied for identification of the number and type
of phases in the recorded XANES spectra. The reference spectra of these phases were
used in a least-square fitting procedure to determine the fraction of each phase present
[154]. For Extended X-ray Absorption Fine Structure (EXAFS) analysis, the data were
converted to k-space using WinXAS, and the least-square curve fitting was performed
with the EXCURVE 98 program [155] based on small atom approximation [156] and ab
initio phase shifts calculated by the program. Details on the comparison of the small
atom approximation, a simplified approach resulting in a reduced computation time, and
Water cooling
Gas in
Gas out ~A– +
Liquid
Sample cell
Water cooling
Soller slits
Fluorescent X-rays
Thermocouple
Kapton window
X-rays
Experimental
18
the angle-averaged curved-wave theory have been discussed by [156] and [157], and
references therein.
For more details on X-ray absorption spectroscopy, theory and basic principles,
applications, techniques and data analysis, it is referred to relevant literature [157-168].
3.3. Setup for water-gas shift testing
The setup used for WGS reaction experiments of the catalyst materials is shown in
Figure 8 and Figure 9.
The setup consisted of a dosing system, based on mass flow controller (MFC) and liquid
flow controller (LFC), a water evaporator, two electrical furnaces for the externally
heated tubular fixed-bed reactors, a microchannel heat exchanger and an on-line GC-
analysis system. GC, dosing system and pressure regulation (by a back pressure
controller) were all computer-controlled. A carbonyl trap was connected to the outlet of
the CO gas cylinder to remove possible iron carbonyl formed. Water was dosed by the
LFC from a cylinder pressurized with He at approx. 3 bar overpressure. Helium exhibits
a low solubility in water, lower than Ar and N2. As a consequence, problems with gas
bubbles disturbing the operation of the LFC were less pronounced than experienced
with for example Ar. The water was injected into the vaporizing unit, and the water
vapour was then mixed into the gas stream. The microchannel heat exchanger was
operated with air and used as water condenser after the reactor unit. The dry exit gas
was analyzed with an Agilent G2891A Micro GC equipped with thermal conductivity
detectors. The Micro GC is equipped with four columns; A: Molecular Sieve 5Å (carrier
gas: argon), B: PoraPLOT U (helium), C: Alumina (helium) and D: OV-1 (helium). For
the WGS experiments, column A was used to detect CO, H2, N2 and CH4, column B to
detect CO2. The gas composition in the dry gas was quantified by using a range of
certified calibration gases as external standards.
Experimental
19
Figu
re 8
. Sch
emat
ic d
raw
ing
of th
e ex
perim
enta
l set
up u
sed
for t
estin
g th
e W
GS
perf
orm
ance
of t
he c
atal
yst m
ater
ials
.
N2
Vent
GC
Vent
CH
PC 2
69
SV
LFC
Rea
ctor
II<5
00 o C
Rea
ctor
I <1
000
o C
PC
MFC
O2
MFC
MFC
H2
MFC
HC
Vent
Vent
Vent
CH
PC 1
50
Cal
ib
Vent
MFC
CO
2
CO
Vent
He
Mic
ro h
eat
exch
ange
r
Wat
er ta
nk
Vapo
rizer
Wat
er ta
nk
By-
pass
Air
Car
bony
l tra
p
N2
N2
Vent
GC
GC
Vent
CH
PC 2
69
SV
LFC
SV
LFC
Rea
ctor
II<5
00 o C
Rea
ctor
I <1
000
o C
PC
MFC
O2
MFC
O2
O2
O2
MFC
MFC
MFC
H2
MFC
H2H2
MFC
HC
HC
Vent
Vent
Vent
CH
PC 1
50C
HPC
150
Cal
ibC
alib
Vent
MFC
CO
2C
O2
CO
CO
Vent
He
He
He
Mic
ro h
eat
exch
ange
r
Wat
er ta
nk
Vapo
rizer
Wat
er ta
nk
By-
pass
Air
Car
bony
l tra
p
Experimental
20
Figu
re 9
. Pic
ture
s of t
he e
xper
imen
tal s
etup
use
d fo
r tes
ting
the
WG
S pe
rfor
man
ce o
f the
cat
alys
t mat
eria
ls.
Experimental
21
The CO conversion as well as the carbon balance was calculated from the CO and CO2
concentrations in the dry exit gas. For the reactant mixtures containing CO, H2O and
balance N2 only, the CO conversion was obtained as follows:
2
2
100%( )CO
COXCO CO
= ⋅+
(3)
For the simulated reformate reactant mixture, containing CO/H2O/CO2/H2/N2, the
calculated CO conversion has to take into account the initial composition of the feed gas
(measured via by-pass; subscript ‘0’), as follows:
0
2 2,0
( ) 100%( )CO
CO COXCO CO CO
−= ⋅+ −
(4)
The selectivity was 100 % in all activity measurements. Traces of CH4 were scarcely
detected during the temperature scans as well as the following recording of the short-
term deactivation behaviour. The approach to equilibrium is calculated with the exit gas
concentrations, corresponding to a certain CO conversion, and the equilibrium constant
(Keq):
2 2
2
100%eq
CO HEquilibrium approachK CO H O
⋅= ⋅⋅ ⋅
(5)
Apparent activation energies (Ea) were determined by the integral method (irreversible
reaction, first order in CO and zero order in H2O, plug-flow) or the differential approach
with an Arrhenius-type plot. According to Keiski et al. [169], the water-gas shift
reaction is not a simple first-order reaction in CO, but a first-order rate equation
describes the phenomenon quite well.
Two different reactors were used for the WGS experiments. A tubular stainless steel
reactor with an inner diameter of 9 mm (Figure 10A, Reactor II in Figure 8) was used in
paper I. In order to minimize temperature gradients the catalyst samples (0.15 – 0.25 g)
Experimental
22
were diluted with inert SiC (300 – 600 µm, 4 g). The dilution should not be too high,
otherwise by-pass effects may significantly decrease the reaction extent, especially at
high CO conversion levels [170]. The catalyst-SiC mixture (approx. 2 ml) was placed
on quartz wool and covered with a layer of SiC particles (average diameter 1250 µm,
approx. 8 ml). Quartz-type reactors (Figure 10B, Reactor I in Figure 8), placed in a gold
insulated furnace (Thermcraft Trans Temp), were applied in the papers IV, V and VI.
Two quartz tubes with different inner diameters, 6 and 10 mm, were used as specified in
the papers. The catalyst powder was placed on quartz wool held in place by a quartz
sinter underneath. The free volume of the reactor was reduced by inserting an additional
quartz tube. Heat and transport limitations were considered by applying well-known
empirical evaluation criteria [171].
A)
T
SiC
Catalyst+ SiC
Quartz wool
Innersteal tubewith meshon top
B)
T
CatalystQuartz woolQuartz sinter
Figure 10. Schematic drawing of the tubular fixed bed reactor configurations: A)
Stainless steel reactor, B) quartz-type reactor.
Results and discussion
23
4. Results and discussion 4.1. Pre-studies
4.1.1. Total pore volume and pore size distribution
Table 2 shows the total pore volume for five Cu-Ce-Zr MMO catalysts prepared by
homogeneous co-precipitation (HCP). The pore volume was determined by three
different methods, namely the Harkin-Jura model, the original DFT model (with N2) and
the BJH method, in order to evaluate the impact of the calculation method. The DFT
model and the BJH method give different absolute values, but show similar trends when
arranging the five catalysts according to the size of their pore volume. The results
obtained with the Harkin-Jura model show certain deviations from the other two
methods in terms of the absolute value as well as the size trend.
Table 2. Total pore volume of five Cu-Ce-Zr MMO catalysts determined by three
different methods, namely the Harkin-Jura model, the original DFT model (with N2) and
the BJH method.
Total pore volume (cm3/g)
HCP-1a HCP-1b HCP-1c HCP-2a HCP-2b
Harkin-Jura 0.0598 0.0958 0.0554 0.2233 0.2126
Original DFT 0.0347 0.1710 0.0837 0.2247 0.1630
BJH method 0.1288 0.2678 0.1893 0.3345 0.2201
The pore size distributions (PSD) for the five MMO catalysts were calculated in two
different ways. Figure 11 shows the PSD calculated by original DFT model (with N2),
assuming a slit-like pore geometry. Figure 12 shows the PSD determined with the
Harkins-Jura model, assuming a cylindrical geometry. Although the Harkin-Jura model
and the DFT model gave different absolute values for the total pore volume (Table 2),
the main trends in the pore size distribution are, however, essentially the same for the
samples investigated.
Results and discussion
24
0
0,003
0,006
0,009
0,012
1 10 100 1000
pore width [ nm ]
pore
vol
ume
[ cm
³/g S
TP ]
HCP-1aHCP-1bHCP-1cHCP-2aHCP-2b
Figure 11. PSD for five Cu-Ce-Zr MMO catalysts calculated by non-local density
functional theory (NLDFT, Original DFT model with N2 [143], DFT Plus software
package [144]) assuming a slit-like pore geometry.
0
0,002
0,004
0,006
0,008
1 10 100 1000pore width [ nm ]
pore
vol
ume
[ cm
³/g S
TP ]
HCP-1a
HCP-1b
HCP-1c
HCP-2a
HCP-2b
Figure 12. PSD for five Cu-Ce-Zr MMO catalysts determined with the Harkins-Jura
model (cylindrical geometry), a classical method based on the Kelvin equation [145].
Results and discussion
25
For further details on the comparison of DFT and classical models based on the Kelvin
equation, it is referred to Rouquerol et al. [172] and Chytil et al. [173], and references
therein.
4.1.2. XRD particle size estimates with Win-crysize
In order to evaluate the particle size estimates obtained from the Win-crysize analysis,
two sensitivity studies were conducted. For a Cu0.23Ce0.54Zr0.23 mixed oxide prepared by
homogeneous co-precipitation (HCP), the Win-crysize procedure was varied to assess
the effect on the particle size estimate. For a Cu0.23Ce0.54Zr0.23 mixed oxide prepared by
nitrate precursor decomposition (NPD), the crysize estimate was compared with particle
size estimates obtained with the Scherrer equation.
Applying the Pearson VII and the Pseudo-Voigt 2 model function, fitting between three
and five peaks ((111), (200), (220), (311), (222)) with Profile and including between
three and four peaks ((111), (200), (220), (311)) in the Win-crysize analysis for the HCP
sample, the average particle size estimate varied between 2.9 and 3.4 nm. The XRD
spectrum of the NPD sample was recorded with both D-5000 and D-5005. Two peaks
were fitted with Profile (Pearson VII) and analyzed with Win-crysize ((111), (220)).
The particle size estimates were 3.7 nm and 3.5 nm for D-5000 and D-5005,
respectively. Analyzing three peaks ((111), (200) and (220)) resulted in 3.8 nm and 3.6
nm, respectively. Estimating the particle size by the Scherrer equation gave 5.0 nm and
4.9 nm for D-5000 and D-5005, respectively. These are average values obtained from
two peaks ((111), (220)), based on the FWHM values, the wavelength for Cu Kα λ =
1.5418 Å and the Scherrer constant K = 0.89 [138]. For the D-5005 spectrum, the
contribution of CuK to the peak intensities was removed prior to determining FWHM.
No background correction was implemented.
TEM and STEM-EDS data obtained for Cu-Ce-Zr MMO catalysts, which were prepared
by co-precipitation (Paper III), suggest that XLBA may somewhat underestimate the
primary crystal size of this type of mixed oxide, possibly with about 2 nm.
Results and discussion
26
4.1.3. Reproducibility of the WGS testing – By-pass effects
Initial studies on the WGS reaction over Cu-Ce-Zr MMO catalysts, prepared by the
homogeneous co-precipitation method (Paper III), in the quartz reactor (Reactor I, inner
diameter 10 mm) showed a poor reproducibility of the activity data, indicating that
some important issues were not taken into account under the initial experimental
conditions. A summary of the pre-study on reproducibility as well as some conclusions
are given in this chapter. The WGS reactant mixture contained 5 % CO and 25 % H2O
in nitrogen (see Paper V and VI). Under these conditions, 70 % CO conversion at 300
°C corresponds to an equilibrium approach of about 1 %.
In the initial experiments, the catalyst samples (< 200 µm, 0,15 g) were mixed with SiC
(50 – 150 µm, average diameter 88 µm) in a mass ratio of 1:1 and then poured into the
quartz reactor onto quartz wool held in place by a quartz sinter. The bed height was
approx. 2 mm. The corresponding reproducibility tests in the temperature range 200 –
300 °C showed strong variations (Figure 13), with an estimated standard deviation of 16
– 18 % at all temperatures, based on five runs. The apparent activation energies
determined by the integral approach (irreversible reaction, first order in CO and zero
order in H2O, plug-flow) varied between 25 and 31 kJ/mole.
A certain degree of segregation of the catalyst powder in the catalyst-SiC bed was
observed upon pouring the mechanical catalyst-SiC mixture into the quartz pipe. Some
of the catalyst powder was for example concentrated along the reactor walls or on top of
the bed. Berger et al. [174,175] investigated the influence of inert-diluted catalyst beds
on the conversion in a gas-solid laboratory micro-reactor for an irreversible reaction, i.e.
N2O decomposition over two different catalysts. Vertically and horizontally segregated
beds as well as mixed beds with different degrees of dilution were considered. Their
results showed that catalyst dilution should be applied with caution, since it may
significantly influence the conversion and lead to an erroneous interpretation. If the
catalyst and the diluting particles are not well-mixed, the conversion may be
significantly reduced due to by-pass and axial dispersion. Apparent activation energies
may also be reduced. The effects are stronger at high conversion levels.
Results and discussion
27
0
20
40
60
80
100
190 210 230 250 270 290 310
temperature [ C ]
CO
con
vers
ion
[ %
]
rep 1rep 2rep 3rep 4rep 5
Figure 13. Reproducibility of WGS measurements on Cu-Ce-Zr MMO catalysts (< 200
µm) diluted with SiC (50 – 150 µm, average diameter 88 µm) in a quartz reactor (inner
diameter 10 mm).
It is thus assumed that a major source for the poor reproducibility of the activity
measurements shown in Figure 13 can be attributed to by-pass effects in the diluted
catalyst bed. Because of a certain random segregation of the catalyst powder upon
pouring, catalyst powder located for example at the reactor walls may be by-passed,
thus resulting in a reduced conversion level. A significant part of the catalyst powder
exhibited a particle size smaller than 50 µm, while the SiC used exhibited particles
mainly in the range 50 – 150 µm. This fraction of small catalyst particles (< 50 µm is
probably particularly prone to segregation upon pouring. Consequently, when diluting
catalyst samples with inert material, the particle size of both materials should match.
A second study was carried out, where Cu-Ce-Zr catalyst samples were applied without
dilution with inert material. The catalyst powder (0.05 g, particle size < 50 µm, bed
height 0.5 – 1 mm) was placed between two layers of quartz wool. A standard deviation
of 6 – 9 % and apparent activation energies of 28 – 29 kJ/mole (integral method) was
obtained for three repeated experiments. Two of the runs produced similar CO
Results and discussion
28
conversion (13.3 % and 12.8 % at 220 °C, and 30.9 % and 30.1 % at 300 °C), while the
third run shows higher conversions (15.0 % at 220 °C and 34.8 at 300 °C).
Approximately 0.02 bar pressure drop was measured over the reactor for the two runs
with similar conversion, and approx. 0.07 bar for the third run with the higher
conversion. This may be an indication of a larger by-pass effect in the two similar
measurements compared to the third with higher conversion.
Three reproductions were also performed on the Cu-Ce-Zr MMO catalysts with
particles in the range 50 – 200 µm. The catalyst powder was placed on quartz wool
(without a covering layer of quartz wool). The amount of catalyst used was 0.10 g (bed
height approx. 1 mm), and no dilution was applied. The apparent activation energies
varied between 32 and 33 kJ/mole in the temperature range 200 – 300 °C. The estimated
standard deviation for these experiments was in the range 2 – 4 %. These values were
later reproduced for Cu-Ce-Zr MMO-CNF nanocomposite catalysts in Paper V and the
MMO catalysts in Paper VI, which were measured under similar reaction conditions.
The average conversions of the experiments with 0.05 g of the (< 50 µm)-fraction show
13 – 19 % lower values than the (50 – 200 µm)-fraction, when scaled up to 0.10 g with
the integral model. An experiment with 0.10 g of the (< 50 µm)-fraction also resulted in
a lower CO conversion than for the (50 – 200 µm)-fraction. This suggests that the (< 50
µm)-fraction suffers more from by-pass effects and should not be used for activity
measurements. The activation energy is not considerably affected by a small degree of
by-passing, especially at low conversion levels [174,175], but the apparent activation
energy will decrease as the extent of by-pass increases [174,175]. In a regime of mass
transport limitations, the apparent activation energy should decrease to about 5 kJ/mole
[176].
The use of Cu-Ce-Zr MMO catalysts with a particle size in the range 50 – 200 µm
without dilution gave the best reproducibility for the WGS reaction experiments in the
quartz reactor with 10 mm inner diameter. To avoid possible limitations from pore
diffusion at high conversion levels under the reaction conditions applied [171], the
upper boundary of the particle range should be decreased. A sieve with mesh size 125
µm was therefore introduced and used in subsequent experiments.
Results and discussion
29
Based on the experience gained during this pre-study, I recommend using the following
configurations for catalytic experiments. Even though most of the experiments
described in this study are carried out without catalyst dilution, I recommend applying
catalyst dilution with inert material in order to minimize local temperature gradients
from the heat produced or consumed by the reaction and to assure a certain bed height
thus avoiding an inhomogeneous bed height distribution over the reactor diameter. The
dilution should, however, be not too high in order to minimize random by-pass effects
[170]. Possible axial temperature gradients arising from a larger bed height, depending
on the heating device, have then also to be taken into account. The size of the catalyst
particles should be matched with the size of the inert material in order to minimize
demixing effects. Finally, the catalytic measurements should be carried out at low
conversion levels, smaller than 30 %, since a perfect mixing is not possible and by-pass
effects are more pronounced at higher conversions [174,175]. In this way, the
experiments are also carried out away from the chemical equilibrium, and transport
limitations are avoided.
Results and discussion
30
4.2. Summary of results and discussion
This section contains a summary of the results obtained in the Papers I – VI and some
comments on general trends and results in the literature. In some cases, additional
results are presented that are not given in the corresponding paper, because of shortage
of space or because these results were obtained after publication.
4.2.1. Relating catalyst structure and composition to the water-gas shift activity of
Cu-Zn-based mixed-oxide catalysts (Paper I)
Summary:
Four catalyst samples were characterized by means of XRD, in situ XANES and
thermogravimetric analysis in order to investigate the effect of cerium oxide on Cu-Zn-
based mixed-oxide catalysts. The activity of the catalyst samples was tested for the
forward water-gas shift reaction. The TOF of the catalysts, based on N2O chemisorption,
for the water-gas shift reaction was found not to be significantly improved by the
addition of CeO2. Ceria impregnated on the hydrotalcite-type precipitate prior to
calcination was found to improve copper dispersion and stability of the MMO catalyst.
The stabilization of Cu-based catalysts by adding further oxide compounds appears to
be a generally applicable approach. Kristiansen [56] reports that the impregnation of
Cu-Al spinel-based catalysts with ZnO can improve the stability of these catalysts
against sulphur and chlorine poisoning. Jung and Joo [177] report an increase in
stability of Cu-Zn-based catalysts for methanol dehydrogenation upon addition of Al or
Cr. The catalysts were prepared by co-precipitation. Saito et al. [77] report that the
addition of Zr to Cu-Zn-Al MMO catalysts does not significantly improve the activity
of the catalyst prepared by co-precipitation, but make them less affected by pre-
treatments such as calcination and pre-reduction. Cu-Zn-Zr-Al MMO catalysts are thus
considered to be more suitable for practical use. Zhang et al. report that the stability of
Cu-Al MMO catalysts for steam reforming of methanol could be improved by adding
Ce [100] or Zr [101]. The catalysts were prepared by co-precipitation. Wu et al. [178]
report that the activity of Cu-Zn-Al MMO catalysts improved after a certain induction
period by impregnation with an aqueous solution of B2O3. The stability of both, B2O3-
containing and B2O3-free Cu-Zn-Al MMO catalysts could be improved by co-
Results and discussion
31
precipitation of Cu-, Zn- and Al-salts with colloidal silica [178,179]. Colloidal silica
was also used for stabilization of a Cu-Zn-Zr-Al MMO catalyst for methanol synthesis
[180]. Chen et al. [181] report that the catalytic activity and stability of Cu/SiO2
catalysts for the high temperature reverse water-gas shift reaction could be improved by
adding iron. Fe was also claimed to prevent the oxidation of Cu via a spillover process
of oxygen. The catalysts were prepared by subsequent impregnation of SiO2 with Cu
and Fe. Iron was also claimed to have similar effects on Cu/Al2O3 and commercial Cu-
Zn-Al MMO catalysts.
Additional aspects and comments:
In the paper, we have reported that the extent of reduction of Cu-aC-350 determined by
TGA-TPR exceeded the maximum copper equivalent determined by ICP-AES
indicating a reduction of components other than copper oxide. A subsequent TGA
analysis (details below) suggested that part of the weight change can be assigned to
decomposition of nitrate precursor remaining after calcination. This illustrates that
reducing atmospheres are more efficient in decomposing nitrate precursors than
oxidizing atmospheres, in line with the decomposition chemistry of nitrates.
The TPR profile of Cu-aC-350 after oxidation treatment in air at 400 °C is shown in
Figure 14 together with the TPR profile of Cu-350. In contrast to the profile shown in
the paper, Cu-aC-350 now exhibits a similar reduction behaviour as Cu-350, since the
heat treatment at 400 °C removed most of the precursor remains, as shown below. The
coverage effect mentioned in the paper, which resulted in a delay of the Cu reduction
during TPR, can thus be related to the decomposition of the precursor remains during
this reduction step.
Table 3 shows the comparison between certain TGA results obtained without and with
the treatment in air at 400 °C after calcination. Comparing the weight loss during
oxidation treatment (subtracted with the weight loss during Ar-drying at 260 °C
described in the paper) with the weight loss corresponding to the difference of the Cu
reduction extent without and with oxidation treatment indicates that most of the
precursor remains removed during the oxidation treatment is also removed during the
Results and discussion
32
TPR step reported in the paper. The Cu dispersion has to be corrected by the amount of
precursor remains interpreted as Cu reduction extent in the paper. Thus, the Cu
dispersion is estimated to 8.9 or 7.9 % based on the Cu content determined by TGA-
TPR or ICP-AES, respectively. Consequently, the TOF of Cu-aC-350 has to be
multiplied by a corresponding correction factor resulting in a TOF somewhat below the
TOF values of the other three catalysts. The effect of the precursor remains on the N2O
titration following the TPR can be neglected. For both measurements, before and after
the oxidation treatment, the weight increase assigned to oxygen uptake was estimated to
0.26 · 10-2 mg-O/mg-sample. The Cu particle size estimated by XLBA has to be
corrected for the real thickness of the passivation layer.
-0,10
-0,08
-0,06
-0,04
-0,02
0,00
0 25 50 75 100 125 150
Time [ min ]
Nor
m. d
eriv
. wei
ght
[mg/
(mgC
u*m
in) ]
50
100
150
200
250
300
Tem
pera
ture
[ C
]
Cu-aC-350-ox-red
Cu-350
Temperature
Figure 14. TPR of Cu-aC-350 (7% H2 in Ar, heating rate: 2 °C/min, 260 °C for 2 h,
flow rate: 80 ml/min at ambient temp.) after oxidation treatment in air at approx. 400 °C
(15 min, heating rate: 10 °C/min, flow rate 80 ml/min at ambient temp.). The TPR of
Cu-350 is given for comparison.
Results and discussion
33
Table 3. Comparison of TGA results before and after oxidation treatment in air at 400
°C (15 min, heating rate: 10 °C/min, flow rate 80 ml/min at ambient temp.).
Weight loss upon oxidation treatment
(excluding Ar-drying at 260 °C) 6.74 mg/mg-sample
Cu mass frac. [-]
ICP-AES 0.26
TGA-TPR
- without ox. treat.
0.44
Precursor weight loss
(equivalent to Cu mass
frac. of 0.21)
- with ox. treat. 0.23 4.99 mg/mg-sample
Cu dispersion [%] (± 0.4 – 0.8 %)
TGA-N2O
- without ox. treat.
4.8
- with ox. treat. 8.9 based on TGA-TPR Cu content
- with ox. treat. 7.9 based on ICP-AES Cu content
Effect of nitrate precursor on N2O
titration
Without ox. treat. 0.26 · 10-2 mg-O/mg-sample
With ox. treat. 0.26 · 10-2 mg-O/mg-sample
Correction of Cu particle size estimated by Scherrer
equation
Passivation
degree [%] Particle size [nm]
Cu dispersion
from XRD [%]
Without ox. treat. 32 16 5.6
With ox. treat. 59 19 4.6
Results and discussion
34
4.2.2. Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts
(Paper II)
Summary:
Catalysts containing metals such as Cu, Ni, Fe, Co are often subjected to passivation
procedures prior to characterization to retain the reduced state of the active metal.
Passivation with N2O or O2 to create a protective oxide layer also results in a certain
degree of sub-surface oxidation. The extent of bulk oxidation depends on the type of
oxidant as well as the size of the metal particles, as shown for copper catalysts. The
final (oxidic) passivation layer requires a certain thickness in order to be stable and
prevent further bulk oxidation in ambient air atmosphere. The studies performed on
reduced metal catalysts based on Cu, Ni, Fe and Co suggest that the result of passivation
procedures should be monitored, in order to quantify the extent of bulk oxidation. Heat
released during exothermic oxidation reactions appear to be a critical parameter, since
the local temperature and hence bulk diffusion and the final extent of bulk oxidation
increases.
Passivation of reduced catalysts is not an ideal strategy for characterization of reduced
systems that are unstable in air. In situ measurements are preferred, but passivated
samples may be used to some extent, e.g. for estimating particle size with XRD, keeping
in mind the limited relevance and accuracy of data derived from this approach. The use
of passivated samples for a detailed X-ray line broadening analysis in correlation with
catalytic activity, distinguishing between particle size and strain effects, is questionable
and requires evaluation by a parallel in situ approach. Strain may also be introduced into
the crystal lattice by the oxidation of the outer metal layers during passivation. Possible
morphological changes depending on the reduction/oxidation potential of the
surrounding atmosphere may also complicate the interpretation of the characterization
results. Finally, the discussion of active reaction sites based on the characterization of
passivated samples is disputable.
Passivation strategies that involve CO and/or CO2 without the formation of significant
amounts of coke or wax around the metal particles may comprise both formation of
Results and discussion
35
carbonaceous species and an oxide layer (at the latest when exposed to air), depending
on the conditions used and the metal to be passivated.
Encapsulation of reduced metal particles by a protective layer of carbon is found to
efficiently preserve the metallic state, as demonstrated for metallic nickel and iron with
carbon nanofibers. For certain catalysts, CNF encapsulation could be preferred instead
of passivation by a protective oxide layer. Provided that the carbon layer is impermeable
to oxygen when exposed to air, the reduced metal particles may be preserved in their
metallic, hence active, state. Possible morphological changes as a result of a change in
the reduction/oxidation potential of the surrounding atmosphere should be of less
concern than for passivation with oxygen. However, the particle morphology might be
affected by the CNF growth process.
Additional aspects and comments:
The Ni-Fe catalyst used in this study was investigated by XAS. The catalyst contains Ni
and Fe in a ratio of Ni:Fe = 8:2 and Al at a molar fraction of 0.25. A least-squares fitting
of the XANES profile at the Ni K-edge with NiO and Ni foil as reference materials gave
a composition of 93 % NiO and 7 % Ni in the calcined sample (Table 4). In the paper it
was claimed that it is unlikely to find 7 % Ni in the calcined sample, and the amount of
Ni found was attributed to uncertainties in the fitting procedures. Figure 15 shows the
XANES profiles of the calcined Ni-Fe catalyst and the reference materials, NiO and Ni
foil. The similarity between the calcined catalyst and NiO is obvious. The deviation in
the high-energy region could be related to data treatment, including energy calibration,
background subtraction and normalization. A small contribution from the Ni profile
function is optimum for the linear combination in order to fit the catalyst profile with
NiO and Ni foil. This example indicates the limits of a simple least-squares fitting
approach with linear combinations of reference spectra. In the end, the reasonability of
the result has to be evaluated by the scientist, and certain error intervals have to be
accepted. We suggest interpreting the data with an error interval of ± 5 % or more.
Results and discussion
36
0,00
0,50
1,00
1,50
2,00
8,31 8,33 8,35 8,37 8,39
photon energy [ keV ]
norm
. abs
orpt
ion
[ a.u
. ]Ni-Fe cat
NiO
Ni foil
Figure 15. XAS spectra of calcined Ni-Fe catalyst, NiO and Ni foil recorded at the Ni
K-edge.
The linear combination was also performed with 3 reference materials, Ni foil, NiO and
NiAl2O4 spinel [182]. In the paper, it was argued against the existence of a Ni-Al spinel
phase in the Ni-Fe catalyst. Inclusion of Ni-Al spinel in least-squares fitting implies that
the spinel is partly reducible under the reduction conditions applied and leads to an
unrealistically low passivation degree. The results of both approaches are compared in
Table 4. The average particle size (dP), including core and passivation layer, was
estimated using a simple cubic model:
(6)
Determination of the reduction degree of Ni, Fe or Co by means of magnetic
measurements, as reported by Olafsen et al. for a Ni-Mg-Al mixed oxide [183], is an
interesting alternative method for evaluating the degree of passivation. In a sample with
a total metal loading of approx. 2 wt%, about 11 % of the Ni remained in the reduced
31
(1 )P Cp
d dx
= ⋅−
Results and discussion
37
state after oxidation in 2 % O2 in N2 for 1 h at ambient temperature. The non-reducible,
non-ferromagnetic fraction of Ni was estimated to about 13 %. The reduction degree of
the reducible Ni (87 % of Ni in the sample) after passivation can therefore be estimated
to about 13 %. The Ni particle size was estimated to 9 ± 2 nm.
Table 4. Comparison of two or three reference compounds (Ni foil, NiO, NiAl2O4) for
linear combination to reproduce the XANES profile of the calcined and reduced-
passivated Ni-Fe catalyst samples at the Ni K-edge.
Ni metal core [nm] 4.5 (with XRD, after passivation, [182])
Ni metal (%) NiO (%) NiAl2O4 (%)
2 reference compounds
Calcined 7 93
Reduced-passivated 70 30
Passivation degree [%] 30
Particle size [nm] 5.4
3 reference compounds
Calcined 0 62 38
Reduced-passivated 74 6 20
Passivation degree [%] 8
Particle size [nm] 4.7
Results and discussion
38
4.2.3. Preparation and characterization of nanocrystalline, high-surface area Cu-
Ce-Zr mixed oxide catalysts from homogeneous co-precipitation (Paper III)
Summary:
Cu0.23Ce0.54Zr0.23-mixed oxides were prepared by homogeneous co-precipitation with
urea. The resulting material exhibits high surface area and small nanocrystalline primary
particles. The material consists of a single fluorite-type phase according to XRD, in the
research literature often denoted as solid solution. As a result of the heterogeneous
nature of the co-precipitation process, however, STEM-EDS analysis shows that Cu and
Zr are inhomogeneously distributed throughout the ceria matrix. The EXAFS analysis
indicates the existence of CuO-type clusters inside the ceria-zirconia matrix. This type
of mixed oxide materials should therefore rather be referred to as heterogeneous single-
phase materials. The pore structure and surface area of the mixed oxides are affected by
preparation parameters during both precipitation (stirring) and the following heat
treatment (drying and calcination). TPR measurements show that most of the copper is
reducible and not inaccessibly incorporated into the bulk structure. Reduction-oxidation
cycling shows that the reducibility improves from the first to the second reduction step,
probably due to a local phase segregation in the metastable mixed oxide with gradual
copper enrichment at the surface of the Ce-Zr particles during heat treatment.
It would be interesting to study the dynamic structural change in situ during reduction
and reoxidation, as well as the structure of the mixed oxide in the reduced state (see for
example [184] for in situ XAFS study on dynamic structural changes in Ce-Zr mixed
oxides). Tanaka et al. [185] reported about a so-called ‘intelligent’ Pd-perovskite
catalyst, where Pd shifts between metallic Pd clusters and Pd oxide incorporated in
perovskite depending on the reduction/oxidation potential of the surrounding gas
atmosphere.
The results and discussion given in this paper are not necessarily limited to Cu-Ce-Zr
mixed oxides, but can to some degree apply to other catalyst formulations prepared by
co-precipitation, e.g. other single-phase materials such as Co-Ce-Zr, or multi-phase
materials such as Cu-, Ni- or Fe-based mixed oxides.
Results and discussion
39
Additional aspects and comments:
Figure 16 shows the effect of the metal composition on the surface area of Cu-Ce mixed
oxides, for the surface-to-volume ratio (stv), the mole-based BET (m-BET) and the
specific BET surface area. The BET data are taken from Shen et al. [112] and Lamonier
et al. [128].
0
0,4
0,8
1,2
1,6
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8Cu mole fraction x [-]
surf
ace
area
ratio
A/A
(x=0
or 0
.12)
[-]
BET [Lam]stv [Lam]m-BET [Lam]BET [Shen]stv [Shen]m-BET [Shen]
Figure 16. Effect of material density and composition on the surface area of Cu-Ce
mixed oxides. The BET data are taken from literature [112,128]. The surface area data
are given as surface area ratio normalized with the values of the pure CeO2 (Cu mole
fraction x = 0) [128] or the Cu-Ce mixture with 12 mole-% Cu (x = 0.12) [112]. M-BET
is a mole-based surface area (m2/mole), i.e. BET normalized with the molar mass. The
densities used for the calculation of the surface-to-volume ratio (stv, m2/m3) are based
on a linear combination of tabulated values for CeO2 (7650 kg/m3) and CuO (6310
kg/m3) [186].
A TEM image of a calcined Cu-Ce-Zr mixed oxide sample (sample Bo) is shown in
Figure 17 indicating the crystalline nature of the nanoparticles. Figure 18 shows the
STEM-EDS elemental mapping of sample Bo after calcination. Cu, Ce and Zr are all
distributed over the whole mapping area (roughly 40 x 40 nm2), and the average metal
Results and discussion
40
composition is close to the overall composition (Cu:Ce:Zr = 0.23:0.54:0.23). Within the
mapping area, regions of a few nanometers in size with high Cu and Zr concentration
compared to the surroundings are found (Figure 18C, E).
Figure 17. TEM image of a representative Cu-Ce-Zr mixed metal oxide powder
(sample Bo) after calcination at 350 °C (size bar 10 nm).
Results and discussion
41
A)
B)
Atomic
%
Metal
frac.
Cu K 10.52 0.27
Zr L 7.73 0.20
Ce L 20.68 0.53
C)
D)
E)
F)
G)
Figure 18. STEM-EDS elemental mapping of a calcined Cu-Ce-Zr mixed oxide powder
(sample Bo). A) Mapping area, marked with white frame. B) Metal ratio in the mapping
area, according to EDS. C) Cu Kα1. D) Ce Lα1. E) Zr Kα1. F) O Kα1. G) EDS spectrum
of the mapping area. The size bar corresponds to 40 nm for all images.
Results and discussion
42
4.2.4. Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas
shift (Paper IV)
Summary:
Cu-Zn-Al and Cu-Ce-Zr mixed oxide catalysts were prepared by two different methods,
co-precipitation and flame spray pyrolysis. The performance of the catalysts was
evaluated using the water-gas shift reaction with and without CO2 and H2 added to the
feed. Cu-Ce-Zr catalysts are found not to be generally superior to Cu-Zn-Al catalysts in
terms of activity or short-term stability. Instead, the difference in activity seems to be
related to structural characteristics of the catalysts as well as the reaction conditions. In
general, increasing the surface area of the catalytic material results in increased activity.
The result of the comparison of different catalysts depends strongly on the structural
characteristics of each catalyst, i.e. the catalyst under study is only as good as the
reference catalyst allows it. Thus, the true potential of a catalyst system can only be
evaluated by using a reasonable reference catalyst, and preparation procedures that lead
to comparable structural characteristics.
The apparent activation energy of the Cu-Ce-Zr MMO catalysts appears to be less
affected by increased concentrations of CO2 and H2 than the Cu-Zn-Al MMO catalysts.
Cu-Ce-based catalysts do not display the same increase in apparent activation energy as
Cu-Zn-based catalysts upon increased concentration of hydrogen and CO2 in the feed.
Further studies that ensure the comparability and consider the impact of additional
factors such as temperature range, conversion level, initial deactivation, catalyst
reducibility and Ea estimation method are necessary. Another question that should be
considered is whether the Cu loading and preparation method affect the value of the
apparent activation energy. For example, different Cu species can be identified in Cu-
Ce- or Cu-Zr-based MMO catalyst materials [187,188] depending on Cu loading and
preparation method (i.e. co-precipitation or impregnation on the support). Assuming
different reactivity of the different Cu species, overall activity and apparent activation
energy might vary with the content of the Cu species in the catalyst.
Results and discussion
43
Additional data:
Figure 19 shows the TPR profile of CuZn-CP.
-0,05
-0,04
-0,03
-0,02
-0,01
00 25 50 75 100 125 150
time [ min ]
norm
alis
ed d
eriv
ativ
e w
eigh
t [ m
g/(m
gCu
min
) ]
90
140
190
240
290
340
390
tem
pera
ture
[ C
]
rep 1
rep 2
temp
Figure 19. TGA-TPR of catalyst CuZn-CP in 7 % H2/Ar. Heating rate: 2 °C/min, flow
rate: 80 ml/min at ambient conditions, catalyst amount: 56-60 mg. The weight
derivative of two parallels, normalized with the total weight loss during reduction, and
the temperature profile are plotted as a function of time.
4.2.5. Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for water-gas shift: Carbon
nanofibers as dispersing agent for the mixed oxide particles (Paper V)
Summary:
Carbon nanofiber (CNF) and Cu-Ce-Zr mixed metal oxide (MMO) containing
nanocomposite catalysts have been prepared by homogeneous co-precipitation with urea.
The water-gas shift reaction (WGS) has been used as test reaction. The CNF-containing
nanocomposite catalysts exhibit similar overall catalytic activity and stability as the
corresponding CNF-free catalyst. 13 wt% of the MMO could be replaced by CNF
without decreasing the overall activity and stability of the catalyst. The specific activity
of the nanocomposites based on the total metal oxide content is similar or higher than
the activity of the CNF-free material, depending on the CNF content. Similar activation
Results and discussion
44
energies are, however, obtained for the CNF-free and CNF-containing materials. We
can not exclude that the CNF material acts as reaction promoter under certain conditions,
but suggest that the impact of CNF addition on the precipitation of the mixed oxide
particles, and hence the catalytic activity relative to the CNF-free MMO, should also be
considered. This is supported by the similar activation energies obtained for the CNF-
free and CNF-containing materials, and CO chemisorption studies resulting in no
significant CO chemisorption on CNF. CNF may be regarded as inert dispersant
material improving the precipitation of the MMO under conditions where the co-
precipitation of the MMO precursors does not result in materials with high surface area.
A possible candidate would be for example the Cu-Zn-Al mixed oxide catalyst
described in Paper VI.
In this study, the CNF-free and the CNF-containing catalysts were prepared under
identical conditions. Under these preparation conditions, the CNF-free catalyst
exhibited a rather high surface area, while for the CNF-containing catalysts, the
precipitation did not result in a complete coverage of the CNF with the mixed oxide.
The CNF rather dispersed the MMO agglomerates. The degree of dispersion of the
MMO by the CNF was therefore limited. Improved preparation conditions, including
lower stirring rate, lower precipitation rate (e.g. via pH-control), good dispersion of the
CNF in the solvent prior to precipitation (e.g. via US treatment and reasonably large
solvent volume), right choice of CNF, high density of oxygen-containing surface groups
on the CNF (e.g. via strong oxidation agent such as a mixture of HNO3 and H2SO4 [189-
191]) and charge matching between support and precipitate should be applied in order
to improve the deposition of the precipitate on the CNF without extensive formation of
agglomerates. In this way, the CNF loading may be increased.
4.2.6. The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for
water-gas shift (Paper VI)
Summary:
Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts were prepared by homogeneous co-
precipitation with urea. Pt-impregnated Cu-Ce-Zr and Cu-Zn-Al MMO catalysts were
Results and discussion
45
prepared by wet impregnation of the precipitates with a Pt precursor salt prior to drying
and calcination.
While the Cu-Ce-Zr mixed oxide catalyst exhibited a high surface area and a good
distribution of all three metals, the Cu-Zn-Al mixed oxide catalyst exhibited a relatively
low surface area and an inhomogeneous distribution of the three metals under the
preparation conditions used. For the Cu-Ce-Zr mixed oxide catalyst, the reduction of Cu
could be achieved at lower temperatures than for the Cu-Zn-Al mixed oxide.
Improvement of the surface area, metal distribution and Cu reducibility in the Cu-Zn-Al
mixed oxide catalyst requires therefore different preparation conditions.
Cu-Ce-Zr mixed oxide catalysts showed WGS activity without pre-reduction under the
reaction conditions used. The not pre-reduced sample was somewhat less active but
more stable than the pre-reduced sample. Pre-reduction is therefore not absolutely
necessary, but an adequate pre-reduction procedure may be applied to optimize the CO
conversion. Pt impregnated on the Cu-Ce-Zr mixed oxide had no significant effect on
Cu reducibility as well as WGS activity and stability.
The WGS activity of the Cu-Zn-Al mixed oxide catalyst correlated with the Cu
reducibility. Below the temperatures required for Cu reduction, the not pre-reduced
sample did not show significant WGS activity. At temperatures where Cu is at least
partly reduced, the catalyst showed significant WGS activity. Above the temperatures
required for partial Cu reduction, the not pre-reduced sample exhibited a somewhat
lower CO conversion than the pre-reduced sample, under similar short-term stability.
Consequently, pre-reduction is not absolutely necessary at these reaction temperatures,
but an adequate pre-reduction procedure may improve the CO conversion.
Pt impregnated on the Cu-Zn-Al mixed oxide catalyst shifted the Cu reduction and
hence also the WGS activity of the unreduced sample to lower temperatures. The not
pre-reduced sample showed similar WGS activity and stability, as compared to the Cu-
Zn-Al mixed oxide catalyst. The pre-reduced sample, however, exhibited a lower CO
conversion and a better stability at 250 °C than the Cu-Zn-Al mixed oxide catalyst
Results and discussion
46
under CO/H2O/N2-feed conditions, indicating the existence of an interaction between
Cu and Pt in the bimetallic catalyst. Under CO/H2O/CO2/H2/N2-feed conditions at 300
°C, the impact of Pt diminished and the CO conversion of the Pt-impregnated catalyst
approached the one of the Cu-Zn-Al mixed oxide catalyst with time on stream,
indicating a destruction of the interaction between Cu and Pt.
Concluding remarks
47
5. Concluding remarks Individual conclusions on the topics of the single papers have been given in the
corresponding papers, and will therefore not be repeated in this section. Instead, more
general comments on WGS catalyst requirements for small-scale hydrogen production
systems will be given, based on Table 1 in chapter 2.
The low temperature activity of Cu-based MMO catalysts continues to be the
benchmark for new developments [5,19,34,35,53-55]. Engineered solutions are possible
that take advantage of the low temperature activity, low cost and proven performance
also for small-scale hydrogen production systems [13,26,56-58], especially for
stationary applications, for which the catalyst requirements are less critical than for
mobile applications (Table 1).
A reduction in catalyst weight and volume can be obtained by improving the dispersion
of the catalyst particles, thus increasing the effectiveness factors for pore-diffusional
regimes and avoiding extensive heat and mass transport limitations. This can be
achieved by depositing the active catalyst onto monolith or microreactors instead of
using classical particulate catalyst beds [19,27,192-194]. A further advantage of this
type of reactors is its lower pressure drop compared to conventional fixed bed reactors.
This approach has been successfully applied for production of synthesis gas by partial
oxidation or oxidative steam reforming in microstructured and monolithic reactors
[16,17,29,30] as well as for conversion of CO by water-gas shift and selective oxidation
of CO in a microreactor [24]. As a further advancement of the work on CNF as inert
support and dispersant material (Paper V), the active catalyst may also be supported on
CNF-covered, ceramic monolith reactors [195] or on structured carbon paper made of
CNF [196].
A problem of Cu-based catalysts is the pyrophoricity of reduced copper, i.e.
considerable temperature rise when exposed to air [19]. The extent of heat released
during oxidation in air depends on the amount of reducible Cu in the catalyst. The
reduction degree of Cu in Cu-Ce-Zr-based MMO catalyst is less than 1 (Paper III), in
contrast to conventional Cu-Zn-Al MMO catalysts, in which basically all the Cu is
Concluding remarks
48
reduced during pre-reduction (Paper I and II), including bulk copper which does not
contribute to the catalytic activity. The key factor for avoiding extensive pyrophoricity
is a high copper dispersion, reducing the total Cu content and the fraction of inactive,
but pyrophoric bulk Cu. The benefit of Cu-Ce-Zr mixed oxides is that Cu is in principle
highly dispersed within the mixed oxide crystals, and inactive Cu buried inside the
crystal particles, and hence not accessible to the reactant gas, does not contribute to the
pyrophoricity (Paper III). It should, however, be noted that the MMO crystals undergo
structural changes during heat treatment resulting in an increase in Cu reducibility.
Another class of Cu-containing mixed oxides with reduced pyrophoric behaviour are
spinel structures. The Engelhard company developed a non-pyrophoric base metal
catalyst [19], possibly based on a Cu-Al spinel structure [57,58]. A similar catalyst
formulation, but prepared in a different way and additionally containing ZnO, was also
patented by Norsk Hydro [56]. In principle, spinel-type Cu-Al mixed oxides give rise to
highly dispersed, small metallic Cu particles deposited on the alumina support after
reduction [84]. This also ensures efficient heat transfer from copper to alumina which
has a higher specific heat capacity than the copper component [57]. Different Cu-based
mixed oxides have been investigated, with Cu-Al- and Cu-Mn-based spinels being the
most promising materials [197-199]. Cu-Si-based mixed oxides may also be an
interesting class of materials to look at in this context (see for example [200]).
Condensation and oxidation tolerance as well as use without pre-reduction are important
issues for small-scale applications. Cu-based catalysts can be operated without pre-
reduction, as shown in Paper VI. Cu is then reduced by the reactant gas and at least
partially reoxidized during shutdown. Important for high activity and rapid response
during start-up is that Cu can be reduced already at low temperatures, i.e. WGS activity
and Cu reducibility are highly correlated. Cu-Ce-Zr MMO catalysts have been shown to
exhibit a low temperature reducibility of Cu (Paper VI). Ko et al. compared Cu-Zr
mixed oxide catalysts with Cu-Zn-Al mixed oxide catalysts for WGS and found that the
Cu-Zr mixed oxide catalysts exhibited lower Cu reduction temperatures as well as
higher CO conversions at low temperatures compared to Cu-Zn-Al [201]. This is,
however, not necessarily a unique feature of ceria/zirconia-based systems. If it is
possible to shift the Cu reduction temperature of Cu-Zn-Al mixed oxides to lower
Concluding remarks
49
temperatures, e.g. by decreasing the CuO particle size and hence increasing the surface-
to-volume ratio, then Cu-Zn-Al mixed oxide catalysts may exhibit an increased low-
temperature activity. The Cu reducibility and hence the low-temperature activity can be
improved by depositing noble metals onto the mixed oxide, as shown for Pt on a Cu-Zn-
Al MMO catalyst (Paper VI). Noble metals can also improve the deactivation behaviour,
depending on the preparation conditions (Paper VI). The improved performance must
then, however, be charged up against increased material costs. Appropriate operation
conditions for start-up and shutdown cycles have to be applied in order to improve the
life-time of the catalyst and hence reduce the economic costs [57,202]. Finally, the risk
of hot spots during reduction and reoxidation may be reduced by depositing the Cu-
based catalysts onto monolith or microreactors instead of using classical particulate
catalyst beds.
True individual commitment to a group effort –
that is what makes a team work, a company work,
a society work, a civilization work.
Vince Lombardi
References
50
6. Suggestions for further work Some suggestions for possible continuative studies on the work presented in this thesis
have already been given in the corresponding papers as well as in chapter 5. Further
work should be done on the operations conditions in small-scale systems. The
performance of Cu-based catalysts needs to be evaluated under frequent start-up and
shutdown conditions [57]. This includes further work on the deactivation behaviour.
The deactivation of Cu-Zn-Al MMO catalysts has been addressed in literature [56,203-
207]. Further studies should be performed on deactivation and regeneration of Cu-Ce-Zr
MMO catalysts (e.g. [82,98,208]). Cu-Ce-based mixed oxide catalysts are claimed to
exhibit good stability at high temperatures (up to 600 °C), which would make them
interesting candidates for membrane reactor applications [52]. Another aspect in respect
of deactivation in fuel processing units is the impact of volatile compounds, such as
metal carbonyls, that may be formed in the reforming unit by catalyst leaching [209,210]
and transported to the WGS unit causing deactivation of the Cu-based catalyst [211-
213], as possibly suggested by pre-studies performed in our group [214].
Further studies should be carried out on oxide-supported noble metal catalysts as
alternative to Cu-based catalysts, especially in view of mobile applications
[5,19,28,34,53]. This should also include bimetallic systems continuing the work started
in Paper VI, since this class of catalysts opens new possibilities for tailoring catalyst
activity and stability [215]. An appropriate comparison of Cu-based and noble metal-
based catalysts may then give information about the true potential of these two classes
of catalysts.
References
51
References [1] http://www.shell.com/renewables.
[2] RENERGI programme established by The Research Council of Norway:
http://www.forskningsradet.no/renergi.
[3] http://www.shell.com/hydrogen.
[4] http://www.hytrec.no.
[5] R. Farrauto, S. Hwang, L. Shore, W. Ruettinger, J. Lampert, T. Giroux, Y. Liu,
O. Ilinich, Annu. Rev. Mater. Res. 33 (2003) 1.
[6] C. Song, Catal. Today 77 (2002) 17.
[7] J.N. Armor, Catal. Lett. 101 (2005) 131.
[8] J.M. Ogden, M.M. Steinbugler, T.G. Kreutz, J. Power Sources 79 (1999) 143.
[9] B.F. Hagh, J. Power Sources 130 (2004) 85.
[10] A. Ersoz, H. Olgun, S. Ozdogan, C. Gungor, F. Akgun, M. Tırıs, J. Power
Sources 118 (2003) 384.
[11] J.R. Lattner, M.P. Harold, Appl. Catal. B 56 (2005) 149.
[12] D.L. Trimm, Z. Ilsen Önsan, Catal. Rev. 43 (2001) 31.
[13] B. Emonts, J. Bøgild Hansen, T. Grube, B. Höhlein, R. Peters, H. Schmidt, D.
Stolten, A. Tschauder, J. Power Sources 106 (2002) 333.
[14] R. Holland, G. Schubak, M. Bradley, K. O’Connor, B. Peppley, US Patent
US6572837, 2003.
[15] J.D. Holladay, Y. Wang, E. Jones, Chem. Rev. 104 (2004) 4767.
[16] I. Aartun, Microstructured reactors for hydrogen production, Ph.D. thesis,
Norwegian University of Science and Technology NTNU, Trondheim, Norway,
2005.
[17] I. Aartun, B. Silberova, H. Venvik, P. Pfeifer, O. Görke, K. Schubert, A.
Holmen, Catal. Today 105 (2005) 469.
[18] J.R. Lattner, M.P. Harold, Int. J. Hydrogen Energy 29 (2004) 393.
[19] W. Ruettinger, O. Ilinich, R.J. Farrauto, J. Power Sources 118 (2003) 61.
[20] D.G. Löffler, K. Taylor, D. Mason, J. Power Sources 117 (2003) 84.
[21] C.D. Dudfield, R. Chen, P.L. Adcock, Int. J. Hydrogen Energy 26 (2001) 763.
[22] G.-G. Park, S.-D. Yim, Y.-G. Yoon, W.-Y. Lee, C.-S. Kim, D.-J. Seo, K. Eguchi,
J. Power Sources 145 (2005) 702.
References
52
[23] J. Han, I.-S. Kim, K.-S. Choi, Int. J. Hydrogen Energy 27 (2002) 1043.
[24] O. Görke, P. Pfeifer, K. Schubert, Appl. Catal. A 263 (2004) 11.
[25] T. Giroux, S. Hwang, Y. Liu, W. Ruettinger, L. Shore, Appl. Catal. B 56 (2005)
95.
[26] I.R. Wheeldon, B.A. Peppley, H. Wieland, US Patent US2003215374, 2003.
[27] A. Karim, J. Bravo, D. Gorm, T. Conant, A. Datye, Catal. Today 110 (2005) 86.
[28] A.F. Ghenciu, Curr. Opin. Solid State Mater. Sci. 6 (2002) 389.
[29] B. Silberova, H.J. Venvik, J.C. Walmsley, A. Holmen, Catal. Today 100 (2005)
457.
[30] B. Silberova, H.J. Venvik, A. Holmen, Catal. Today 99 (2005) 69.
[31] B. Arstad, H. Venvik, H. Klette, J. Walmsley, W.M. Tucho, R. Holmestad, A.
Holmen, R. Bredesen, Catal. Today, accepted.
[32] D. Andreeva, Gold Bull. 35 (2002) 82.
[33] L. Mond and C. Langer, British Patent 12 (1888) 608.
[34] J.R. Ladebeck, J.P. Wagner, in: Handbook of Fuel Cells – Fundamentals,
Technology and Applications, Vol. 3: Fuel Cell Technology and Applications,
eds. W. Vielstich, H.A. Gasteiger, A. Lamm, Wiley and Sons, New York, 2003,
pp.190-201.
[35] N. Schumacher, A. Boisen, S. Dahl, A.A. Gokhale, S. Kandoi, L.C. Grabow, J.A.
Dumesic, M. Mavrikakis, I. Chorkendorff, J. Catal. 229 (2005) 265.
[36] T. Sueyama, III. Design and operation of carbon monoxide shift: Partial
oxidation, in: Ammonia vol. II, eds. A.V. Slack, G.R. James, Marcel Dekker,
New York, 1974, pp. 73-113.
[37] K. Atwood, M.R. Arnold, E.G. Appel, Ind. Eng. Chem. 42 (1950) 1600.
[38] I. Fishtik, R. Datta, Surf. Sci. 512 (2002) 229.
[39] O.A. Rokstad, Matbal, programme for calculation of equilibrium compositions
(by B. Jensen), Department of Chemical Engineering, Norwegian University of
Science and Technology, memorandum, Trondheim, 2002.
[40] C. Rhodes, G.J. Hutchings and A.M. Ward, Catal. Today 23 (1995) 43.
[41] K. Kochloefl, in: Handbook of Heterogeneous Catalysis, eds. G. Ertl, H.
Knözinger, J. Weitkamp, VCH, Ludwigshafen, Vol. 4, Chapter 3.3, 1997, pp.
1831-1843.
References
53
[42] J. Barbier, D. Duprez, Appl. Catal. B 4 (1994) 105.
[43] R.F. Mann, J.C. Amphlett, B. Peppley, C.P. Thurgood, Int. J. Chem. React. Eng.
2 (2004) A5.
[44] N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H. Ribeiro, J.
Catal. 217 (2003) 233.
[45] (a) G. Wang, L. Jiang, Y. Zhou, Z. Cai, Y. Pan, X. Zhao, Y. Li, Y. Sun, B.
Zhong, X. Pang, W. Huang, K. Xie, J. Mol. Struct. : THEOCHEM 634 (2003)
23; (b) G. Wang, L. Jiang, Z. Cai, Y. Pan, X. Zhao, W. Huang, K. Xie, Y. Li, Y.
Sun, B. Zhong, J. Phys. Chem. B 107 (2003) 557.
[46] R.L. Keiski, O. Desponds, Y.-F. Chang, G.A. Somorjai, Appl. Catal. A 101
(1993) 317.
[47] E. Fiolitakis, H. Hofmann, J. Catal. 80 (1983) 328.
[48] C. Callaghan, I. Fishtik, R. Datta, M. Carpenter, M. Chmielewski, A. Lugo, Surf.
Sci. 541 (2003) 21.
[49] C.V. Ovesen, B.S. Clausen, B.S. Hammershøi, G. Steffensen, T. Askgaard, I.
Chorkendorff, J.K. Nørskov, P.B. Rasmussen, P. Stoltze, P. Taylor, J. Catal. 158
(1996) 170.
[50] D.S. Newsome, Catal. Rev. Sci. Eng. 21 (1980) 275.
[51] Y. Li, Q. Fu and M. Flytzani-Stephanopoulos, Appl. Catal., B 27 (2000) 179.
[52] X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055.
[53] S.L. Swartz, M.M. Seabaugh, C.T. Holt, W.J. Dawson, Fuel Cells Bull. 4 (2001)
7.
[54] T. Utaka, T. Takeguchi, R. Kikuchi, K. Eguchi, Appl. Catal. A 246 (2003) 117.
[55] H. Meland, T. Johannessen, B. Arstad, H.J. Venvik, M. Rønning, A. Holmen,
Stud. Surf. Sci. Catal., accepted.
[56] L.A. Kristiansen, US Patent US4308176, 1981.
[57] W.F. Ruettinger, O.M. Ilinich, R.J. Farrauto, US Patent US2004082669, 2004.
[58] O.M. Ilinich, W.F. Ruettinger, R.T. Mentz, R.J. Farrauto, US Patent
US2004082471, 2004.
[59] P. Panagiotopoulou, D.I. Kondarides, Catal. Today 112 (2006) 49.
[60] F. Boccuzzi, A. Chiorino, M. Manzoli, D. Andreeva, T. Tabakova, J. Catal. 188
(1999) 176.
References
54
[61] A. Basinska, J. Stoch, F. Domka, Pol. J. Environ. Stud. 12 (2003) 395.
[62] H. Sakurai, A. Ueda, T. Kobayashi, M. Haruta, Chem. Commun. 3 (1997) 271.
[63] J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157.
[64] T. Utaka, T. Okanishi, T. Takeguchi, R. Kikuchi, K. Educhi, Appl. Catal. A 245
(2003) 343.
[65] A. Basińska, L. Kępiński, F. Domka, Appl. Catal. A 183 (1999) 143.
[66] J.C. Frost, Nature 334 (1988) 577.
[67] A. Basińska, W.K. Jóźwiak, J. Góralski, F. Domka, Appl. Catal. A 190 (2000)
107.
[68] Q. Fu, W. Deng, H. Saltsburg, M. Flytzani-Stephanopoulos, Appl. Catal. B 56
(2005) 57.
[69] T. Tabakova, F. Boccuzzi, M. Manzoli, J.W. Sobczak, V. Idakiev, D. Andreeva,
Appl. Catal. B 49 (2004) 73.
[70] D. Andreeva, V. Idakiev, T. Tabakova, R. Giovanoli, Bulg. Chem. Comm. 30
(1998) 59.
[71] F. Boccuzzi, A. Chiorino, M. Manzoli, D. Andreeva, T. Tabakova, L. Ilieva, V.
Idakiev, Catal. Today 75 (2002) 169.
[72] T. Tabakova, V. Idakiev, D. Andreeva, I. Mitov, Appl. Catal. A 202 (2000) 91.
[73] G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G. Thomas, B.H.
Davis, Appl. Catal. A 258 (2004) 203.
[74] P. Panagiotopoulou, A. Christodoulakis, D.I. Kondarides, S. Boghosian, J. Catal.
240 (2006) 114.
[75] S. Hilaire, X. Wang, T. Luo, R.J. Gorte, J. Wagner, Appl. Catal. A 258 (2004)
271.
[76] M. Saito, J. Wu, K. Tomoda, I. Takahara, K. Murata, Catal. Lett. 83 (2002) 1.
[77] M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003)
11.
[78] L.-T. Weng, B. Delmon, Appl. Catal. A 81 (1992) 141.
[79] R. Di Monte, J. Kašpar, J. Mater. Chem. 15 (2005) 633.
[80] S. Colussi, C. de Leitenburg, G. Dolcetti, A. Trovatelli, J. Alloys Comp. 374
(2004) 387.
References
55
[81] J.P. Breen, R. Burch, D. Tibiletti, in: Catalysis in Application, eds. S.D. Jackson,
J.S.J. Hargreaves, D. Lennon, Royal Society of Chemistry, Cambridge, 2003, pp.
227-232.
[82] W. Ruettinger, X. Liu, R.J. Farrauto, Appl. Catal. B 65 (2006) 135.
[83] A. Basińska, F. Domka, Appl. Catal. A 179 (1999) 241.
[84] K. Sekizawa, S.-i. Yano, K. Eguchi, H. Arai, Appl. Catal. A 169 (1998) 291.
[85] E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347.
[86] X. Wang, R.J. Gorte, Appl. Catal. A 247 (2003) 157.
[87] S. Zhao, R.J. Gorte, Catal. Lett. 92 (2004) 75.
[88] C.-S. Chen, W.-H. Cheng, S.-S. Lin, Appl. Catal. A 257 (2004) 97.
[89] S.Y. Choung, M. Ferrandon, T. Krause, Catal. Today 99 (2005) 257.
[90] A. Basińska, F. Domka, Catal. Lett. 43 (1997) 59.
[91] S.-i. Ito, H. Tanaka, Y. Minemura, S. Kameoka, K. Tomishige, K. Kunimori,
Appl. Catal. A 273 (2004) 295.
[92] J. Nakamura, J.M. Campbell, C.T. Campbell, J. Chem. Soc., Faraday Trans. 86
(1990) 2725.
[93] M. Maack, H. Friis-Jensen, S. Sckerl, J.H. Larsen, I. Chorkendorff, Top. Catal.
22 (2003) 151.
[94] J. Wold, L.A. Kristiansen, Patent WO 8504597 PCT NO8500016, 1985.
[95] N. Pavlenko, P.P. Kostrobij, Yu. Suchorski, R. Imbihl, Surf. Sci. 489 (2001) 29.
[96] A. Houteit, H. Mahzoul, P. Ehrburger, P. Bernhardt, P. Légaré, F. Garin, Appl.
Catal. A 306 (2006) 22.
[97] Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Catalytic Science
Series Vol. 2, ed. G.J. Hutchings, Imperial College Press, London, 2002.
[98] D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107.
[99] Y. Liu, T. Hayakawa, T. Tsunoda, K. Suzuki, S. Hamakawa, K. Murata, R.
Shiozaki, T. Ishii, M. Kumagai, Top. Catal. 22 (2003) 205.
[100] X. Zhang, P. Shi, J. Mol. Catal. A 194 (2003) 99.
[101] X.R. Zhang, P. Shi, J. Zhao, M. Zhao, C. Liu, Fuel Process. Tech. 1680 (2003) 1.
[102] A. Szizybalski, Ph.D. Thesis, Technical University Berlin, Germany, 2005.
[103] P. Bera, S. Mitra, S. Sampath, M.S. Hegde, Chem. Comm. 10 (2001) 927.
[104] T.-J. Huang, Y.-C. Kung, Catal. Lett. 85 (2003) 49.
References
56
[105] N.Y. Usachev, I.A. Gorevaya, E.P. Belanova, A.V. Kazakov, O.K. Atal´yan,
V.V. Kharlamov, Russ. Chem. Bull., Int. Ed. 53 (2004) 538.
[106] G. Avgouropoulos, T. Ioannides, H. Matralis, Appl. Catal. B 56 (2005) 87.
[107] P. Bera, S.T. Aruna, K.C. Patil, M.S. Hegde, J. Catal. 186 (1999) 36.
[108] Lj. Kundakovic, M. Flytzani-Stephanopoulos, J. Catal. 179 (1998) 203.
[109] S. Hocevar, J. Batista, J. Levec, J. Catal. 184 (1999) 39.
[110] C. de Leitenburg, D. Goi, A. Primavera, A. Trovarelli, G. Dolcetti, Appl. Catal.
B 11 (1996) 29.
[111] E.E. Ortelli, J.M. Weigel, A. Wokaun, Catal. Lett. 54 (1998) 41.
[112] W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 83 (2002) 33.
[113] C. Lu, W.L. Worrell, J.M. Vohs, R.J. Gorte, J. Electrochem. Soc. 150 (2003)
1357.
[114] L. Jalowiecki-Duhamel, A. Ponchel, C. Lamonier, Int. J. Hydrogen Energy 24
(1999) 1083.
[115] D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90.
[116] M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal.
Today 100 (2005) 249.
[117] J. Kašpar, P. Fornasiero, J. Solid State Chem. 171 (2003) 19.
[118] S.-P. Wang, X.-C. Zheng, X.-Y. Wang, S.-R. Wang, S.-M. Zhang, L.-H. Yu,
W.-P. Huang, S.-H. Wu, Catal. Lett. 105 (2005) 163.
[119] M. Ozawa, J. Alloy Compd. 275-277 (1998) 886.
[120] S. Bedrane, C. Descorme, D. Duprez, Catal. Today 75 (2002) 401.
[121] A. Norman, V. Perrichon, Phys. Chem. Chem. Phys. 5 (2003) 3557.
[122] A. Suzuki, T. Yamamoto, Y. Nagai, T. Tanabe, F. Dong, T. Sasaki, T. Taniike,
M. Nomura, Y. Iwasawa, Proc. of the 13th Int. Cong. on Catal., Paris, France,
July 2004.
[123] M.P. Kapoor, A. Raj, Y. Matsumura, Micropor. Mesopor. Mater. 44-45 (2001)
565.
[124] M. Hirano, T. Miwa, M. Inagaki, J. Solid State Chem. 158 (2001) 112.
[125] J.A. Wang, M.A. Valenzuela, S. Castillo, J. Salmones, M. Moran-Pineda, J. Sol-
Gel Sci. Tech. 26 (2003) 879.
[126] Y. Zhang, S. Andersson, M. Muhammed, Appl. Catal. B 6 (1995) 325.
References
57
[127] (a) G. Wrobel, C. Lamonier, A. Bennani, A. D`Huysser, A. Aboukais, J. Chem.
Soc., Faraday Trans. 92 (1996) 2001-2009; (b) C. Lamonier, A. Ponchel, A.
D`Huysser, L. Jalowiecki-Duhamel, Catal. Today 50 (1999) 247.
[128] C. Lamonier, A. Bennani, A. D`Huysser, A. Aboukais, G. Wrobel, J. Chem.
Soc., Faraday Trans. 92 (1996) 131.
[129] M. Kurtz, H. Wilmer, T. Genger, O. Hinrichsen, M. Muhler, Catal. Lett. 86
(2003) 77.
[130] R.A. Hadden, P.J. Lambert, C. Ranson, Appl. Catal. A 122 (1995) L1.
[131] S. Golunski, R. Rajaram, N. Hodge, G.J. Hutchings, C.J. Kiely, Catal. Today 72
(2002) 107.
[132] Y. Liu, T. Hayakawa, K. Suzuki, S. Hamakawa, T. Tsunoda, T. Ishii, M.
Kumagai, Appl. Catal. A 223 (2002) 137.
[133] Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki, K. Eguchi, Appl. Catal. A 238
(2002) 11.
[134] M.M. Mohamed, S.M.A. Katib, Appl. Catal. A 287 (2005) 236.
[135] X. Tang, B. Zhang, Y. Li, Y. Xu, Q. Xin, W. Shen, Catal. Today 93-95 (2004)
191.
[136] H. Meland, MSc Thesis, Department of Chemical Engineering, Norwegian
University of Science and Technology, Trondheim, 2003.
[137] C.B. Nilsen, MSc Thesis, Department of Chemical Engineering, Norwegian
University of Science and Technology, Trondheim, 2003.
[138] J. Lemaître, P.G. Menon, F. Delannay, in: Characterization of heterogeneous
catalysts, ed. F. Delannay, Marcel Dekker, New York, 1984, pp. 299-365.
[139] Diffracplus Profile, Profile fitting program, User’s Manual, Siemens.
[140] Diffracplus Win-crysize, Crystallite size and microstrain, User’s Manual, Bruker
Analytical X-Ray Systems.
[141] B.E. Warren, in: Progress in Metal Physics Vol. 8, eds. B. Chalmes, R. King,
Pergamon Press, London, 1959, pp. 147-202.
[142] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373.
[143] P.B. Balbuena, K.E. Gubbins, Fluid Phase Equilib. 76 (1992) 21.
[144] DFT Plus for Windows, version 3.00, User’s Manual, Micrometrics Instrument
Corporation.
References
58
[145] W.D. Harkins, G. Jura, J. Am. Chem. Soc. 66 (1944) 136.
[146] A. Dandekar, R.T.K. Baker, M.A. Vannice, J. Catal. 183 (1999) 131.
[147] G.C. Chinchen, C.M. Hay, M.D. Vandervell, K.C. Waugh, J. Catal. 103 (1987)
79.
[148] J. Ma, C. Park, N.M. Rodriguez, R.T.K. Baker, J. Phys. Chem. B 105 (2001)
11994.
[149] S. Sato, R. Takahashi, T. Sodesawa, K. Yuma, Y. Obata, J. Catal. 196 (2000)
195.
[150] J.R. Jensen, T. Johannessen, H. Livbjerg, Appl. Catal. A 266 (2004) 117.
[151] G.J.J. Bartley, R. Burch, R.J. Chappell, Appl. Catal. 43 (1988) 91.
[152] F.W. Lytle, R.B. Greegor, E.C. Marques, D.R. Sandstrom, G.H. Via, J.H. Sinfelt,
J. Catal. 95 (1985) 546.
[153] T. Ressler, J. Synch. Rad. 5 (1998) 118.
[154] T. Ressler, J. Wong, J. Roos, I.L. Smith, Environ. Sci. Tech. 34 (2000) 950.
[155] N. Binsted, EXCURV98: CCLRC Daresbury Laboratory computer program,
1998.
[156] S.J. Gurman, J. Phys. C 21 (1988) 3699.
[157] M. Rønning, MSc Thesis, Department of Chemistry, Norwegian University of
Science and Technology, Trondheim, 1994.
[158] D.C. Koningsberger, R. Prins, X-Ray Absorption: Principles, Applications,
Techniques of EXAFS, SEXAFS and XANES, Wiley and Sons, New York,
1987.
[159] D.C. Koningsberger, B.L. Mojet, G.E. van Dorssen, D.E. Ramaker, Top. Catal.
10 (2000) 143.
[160] D.E. Ramaker, G.E. van Dorssen, B.L. Mojet, D.C. Koningsberger, Top. Catal.
10 (2000) 157.
[161] P. Behrens, Trends Anal. Chem. 11(6) (1992) 218.
[162] P. Behrens, Trends Anal. Chem. 11(7) (1992) 237.
[163] A.I. Frenkel, C.W. Hills, R.G. Nuzzo, J. Phys. Chem. B 105 (2001) 12689.
[164] J.J. Rehr, R.C. Albers, Rev. Mod. Phys. 72 (2000) 621.
[165] M. Fernández-García, Catal. Rev. 44 (2002) 59.
[166] B.S. Clausen, J.K. Nørskov, Top. Catal. 10 (2000) 221.
References
59
[167] R.B. Greegor, F.W. Lytle, J. Catal. 63 (1980) 476.
[168] M. Borowski, J. Phys. IV 7 (1997) 259.
[169] R.L. Keiski, O. Desponds, Y.F. Chang, G.A. Somorjai, Appl. Catal. A 101
(1993) 317.
[170] R.J. Berger, J. Pérez-Ramírez, F. Kapteijn, J.A. Moulijn, Chem. Eng. Sci. 57
(2002) 4921.
[171] F. Kapteijn, J.A. Moulijn, Laboratory Reactors, in: Handbook of Heterogeneous
Catalysis Vol. 3, eds. G. Ertl, H. Knözinger, J. Weitkamp, VCH
Verlagsgesellschaft mbH, Weinheim, Germany, 1997, chapter 9, p. 1359.
[172] F. Rouquerol, J. Rouquerol, K. Sing, Adsorption by Powders and Porous Solids,
Academic Press, London, 1999.
[173] S. Chytil, W.R. Glomm, E. Vollebekk, H. Bergem, J. Walmsley, J. Sjöblom, E.A.
Blekkan, Micropor. Mesopor. Mater. 86 (2005) 198.
[174] R.J. Berger, J. Pérez-Ramírez, F. Kapteijn, J.A. Moulijn, Appl. Catal. A 227
(2002) 321.
[175] R.J. Berger, J. Pérez-Ramírez, F. Kapteijn, J.A. Moulijn, Chem. Eng. J. 90 (2002)
173.
[176] M. Baerns, H. Hofmann, A. Renken, Chemische Reaktionstechnik Vol. 1, Georg
Thieme Verlag, Stuttgart, Germany, 1999, chapter 6, p. 135.
[177] K.-D. Jung, O.-S. Joo, Catal. Lett. 84 (2002) 21.
[178] J. Wu, M. Saito, H. Mabuse, Catal. Lett. 68 (2000) 55.
[179] J. Wu, M. Saito, J. Catal. 195 (2000) 420.
[180] J. Wu, M. Saito, M. Takeuchi, T. Watanabe, Appl. Catal. A 218 (2001) 235.
[181] C.-S. Chen, W.-H. Cheng, S.-S. Lin, Appl. Catal. A 257 (2004) 97.
[182] Z. Yu, PhD Thesis, Norwegian University of Science and Technology,
Trondheim, 2005.
[183] A. Olafsen, Å. Slagtern, I.M. Dahl, U. Olsbye, Y. Schuurman, C. Mirodatos, J.
Catal. 229 (2005) 163.
[184] A. Suzuki, T. Yamamoto, Y. Nagai, T. Tanabe, F. Dong, T. Sasaki, T. Taniike,
M. Nomura, Y. Iwasawa, Proc. of the 13th Int. Cong. on Catal., Paris, France,
July 2004.
[185] H. Tanaka, M. Taniguchi, N. Kajita, M. Uenishi, I. Tan, N. Sato, K. Narita, M.
References
60
Kimura, Top. Catal. 30-31 (2004) 389.
[186] http://www.matweb.com; primary literature: CRC Handbook of Chemistry and
Physics, David R. Lide (Ed.), 79th ed., CRC Press, Boca Raton, FL, USA, 1998.
[187] C. Lamonier, A. Bennani, A. D`Huysser, A. Aboukais, G. Wrobel, J. Chem.
Soc., Faraday Trans. 92 (1996) 131.
[188] V. Ramaswamy, M. Bhagwat, D. Srinivas, A.V. Ramaswamy, Catal. Today 97
(2004) 63.
[189] T.G. Ros, A.J. van Dillen, J.W. Geus, D.C. Koningsberger, Chem. Eur. J. 8
(2002) 1151.
[190] J.H. Bitter, M.K. van der Lee, A.G.T. Slotboom, A.J. van Dillen, K.P. de Jong,
Catal. Lett. 89 (2003) 139.
[191] S. Kubota, H. Nishikiori, N. Tanaka, M. Endo, T. Fujii, J. Phys. Chem B 109
(2005) 23170.
[192] T. Giroux, S. Hwang, Y. Liu, W. Ruettinger, L. Shore, Appl. Catal. B 56 (2005)
95.
[193] J.T. Richardson, D. Remue, J.-K. Hung, Appl. Catal. A 250 (2003) 319.
[194] C.P. Thurgood, J.C. Amphlett, R.F. Mann, B.A. Peppley, Proc. of AIChE
Annual Meeting, Cincinnati, USA, Oct. 30-Nov. 4 2005, 501d/1.
[195] E. García-Bordejé, I. Kvande, D. Chen, M. Rønning, Adv. Mater. 18 (2006)
1589.
[196] O. Smiljanic, T. Dellero, A. Serventi, G. Lebrun, B.L. Stansfield, J.P. Dodelet,
M. Trudeau, S. Désilets, Chemical Physics Letters 342 (2001) 503.
[197] Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki, K. Eguchi, Appl. Catal. A 242
(2003) 287.
[198] Y. Tanaka, T. Utaka, R. Kikuchi, T. Takeguchi, K. Sasaki, K. Eguchi, J. Catal.
215 (2003) 271.
[199] Y. Tanaka, T. Takeguchi, R. Kikuchi, K. Eguchi, Appl. Catal. A 279 (2005) 59.
[200] E.K. Poels, D.S. Brands, Appl. Catal. A 191 (2000) 83.
[201] J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157.
[202] C.E. Quincoces, N.E. Amadeo, M.G. Gonzalez, Stud. Surf. Sci. Catal. 111 (1997)
535.
[203] M.V. Twigg, M.S. Spencer, Appl. Catal. A 212 (2001) 161.
References
61
[204] M.V. Twigg, M.S. Spencer, Top. Catal. 22 (2003) 191.
[205] C.P. Thurgood, J.C. Amphlett, R.F. Mann, B.A. Peppley, Top. Catal. 22 (2003)
253.
[206] J.T. Sun, I.S. Metcalfe, M. Sahibzada, Ind. Eng. Chem. Res. 38 (1999) 3868.
[207] M. Kurtz, H. Wilmer, T. Genger, O. Hinrichsen, M. Muhler, Catal. Lett. 86
(2003) 77.
[208] W. Daniell, N.C. Lloyd, C. Bailey, P.G. Harrison, J. Phys. IV 7 (1997) 963.
[209] P.M. Torniainen, X. Chu, L.D. Schmidt, J. Catal. 146 (1994) 1.
[210] C.H. Bartholomew, Appl. Catal. A 212 (2001) 17.
[211] T.-c. Xiao, H.-t. Wang, J.-x. Su, Y.-l. Lu, S.-w. Gu, Proc. of 12th Int. Symp. on
Alcohol Fuels, Beijing, Sept. 1998, 8.
[212] G.W. Roberts, D.M. Brown, T.H. Hsiung, J.J. Lewnard, Chem. Eng. Sci. 45
(1990) 2713.
[213] G.W. Roberts, D.M. Brown, T.H. Hsiung, J.J. Lewnard, Ind. Eng. Chem. Res.
32 (1993) 1610.
[214] K. Johnsen, MSc Thesis, Department of Chemical Engineering, Norwegian
University of Science and Technology, Trondheim, 2002.
[215] A. Groβ, Top. Catal. 37 (2006) 29.
List of appendices
62
List of appendices
I M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Relating
catalyst structure and composition to the water-gas shift activity of Cu-Zn-
based mixed-oxide catalysts, Catalysis Today 100 (2005), 249-254.
II F. Huber, Z. Yu, S. Lögdberg, M. Rønning, D. Chen, H. Venvik, A. Holmen,
Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts,
Catalysis Letters, in press.
III F. Huber, H. Venvik, M. Rønning, J. Walmsley, A. Holmen, Preparation and
characterization of nanocrystalline, high-surface area Cu-Ce-Zr mixed oxide
catalysts from homogeneous co-precipitation, Manuscript in preparation.
IV F. Huber, H. Meland, M. Rønning, H. Venvik, A. Holmen, Comparison of Cu-
Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift, submitted.
V F. Huber, Z. Yu, J. Walmsley, D. Chen, H. Venvik, A. Holmen,
Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for water-gas shift: Carbon
nanofibers as dispersing agent for the mixed oxide particles, Applied Catalysis
B: Environmental, accepted.
VI F. Huber, J. Walmsley, H. Venvik, A. Holmen, The effect of platinum in Cu-
Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift, Manuscript in
preparation.
Paper I
Relating catalyst structure and composition to the water-gas shift
activity of Cu-Zn-based mixed-oxide catalysts
Catalysis Today 100 (2005), 249-254.
www.elsevier.com/locate/cattod
Catalysis Today 100 (2005) 249–254
Relating catalyst structure and composition to the water–gas
shift activity of Cu–Zn-based mixed-oxide catalysts
Magnus Rønninga,*, Florian Hubera, Hilde Melanda, Hilde Venvikb,De Chena, Anders Holmena
aNorwegian University of Science and Technology (NTNU), Department of Chemical Engineering, N-7491 Trondheim, NorwaybSINTEF Materials and Chemistry, N-7465 Trondheim, Norway
Available online 28 December 2004
Abstract
In order to investigate the effect of cerium oxide on Cu–Zn-based mixed-oxide catalysts four catalyst samples were characterized by means
of XRD, in situ XANES and thermogravimetric analysis. The activity of the catalyst samples was tested for the forward water–gas shift
reaction. Cerium oxide was found to increase the crystallinity of the ZnO phase indicating a segregation of the Cu and ZnO phases. The TOF
of the water–gas shift reaction based on chemisorption data was found to be independent of composition and preparation conditions of the four
catalyst samples. In contrast, the catalyst stability depends on composition and preparation conditions. Cerium oxide impregnated before
calcination of the hydrotalcite-based Cu–Zn precursors leads to a more stable water–gas shift catalyst.
# 2004 Elsevier B.V. All rights reserved.
Keywords: Water–gas shift; Copper; Cerium oxide; XANES; XRD; Thermogravimetric analysis
1. Introduction
The water–gas shift (WGS) reaction (CO + H2O $CO2 + H2) is an important step in the production of H2
from hydrocarbons. Recently, the WGS reaction has received
renewed interest as a key step in fuel processing to reduce
the CO level in hydrogen produced for proton exchange
membrane fuel cell (PEMFC) applications [1]. For low
temperature WGS, Cu is usually preferred as the active com-
ponent because of its proven activity [2,3]. However, there is a
need for catalysts with even higher activity and stability
compared to the traditional CuO–ZnO–Al2O3 system.
It has been shown that oxides with high oxygen storage
capacity such as CeO2 can exhibit high WGS activity in
conjunction with various metal promoters [4–7]. The role of
ceria in such systems is proposed to be via a ceria-mediated
redox process where the oxygen storage capacity of ceria
and the metal are the active elements [4,5,8]. Others have
found evidence of a mechanism involving the reaction
between CO and active OH groups to form surface formates
* Corresponding author. Tel.: +47 73594121; fax: +47 73595047.
E-mail address: [email protected] (M. Rønning).
0920-5861/$ – see front matter # 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.cattod.2004.09.059
[7,9–11]. The promoting metal in the form of e.g. Pt, Au or
Cu significantly increases the WGS activity of ceria and
enhances the reducibility of ceria [4,6]. In the presence of
Cu, ceria reduction is proposed to start at low temperatures
(<200 8C) [4]. Ceria is also believed to enhance reducibility
and stabilise the Cu particles towards sintering [12]. For both
cases, ceria as an active component and as a promoter, the
results indicate that ceria is partially reduced in the active
catalyst, most likely by surface reduction. However, the
effect of ceria on the Cu metal particles is not clear.
In this study, we examine the effect of ceria addition on
the Cu metal particles in Cu–ZnO–Al2O3 catalysts derived
from layered hydroxide materials of the hydrotalcite
structure [13]. Mixed-oxide catalysts derived from hydro-
talcite-like materials are shown to possess high activity
and thermal stability together with relatively high metal
dispersion and homogeneous distribution of the components
[14]. WGS activity measurements and deactivation studies
are examined together with structural information from
XRD, XANES and other techniques. However, reaction
mechanisms are not being considered in this report. Kinetic
considerations of the activity measurements will be treated
elsewhere [15].
M. Rønning et al. / Catalysis Today 100 (2005) 249–254250
2. Experimental
A series of Cu- and Zn-based catalysts derived from
hydrotalcite structures were synthesised by co-precipitation
of the precursor salts followed by subsequent aging, filtra-
tion, washing, drying and calcination. The aqueous anionic
solution was made by dissolving Na2CO3 and NaOH.
Nitrates of Cu, Zn and Al were added at 80 8C under
continuous stirring and pH adjustment. The solution was
aged for 15 h before being filtered and washed and dried in
vacuum. The catalysts were impregnated with Ce(NO3)3
either before or after calcination. The sample labels (see
Table 1) are related to the preparation conditions: Cu-350
means a CuO–ZnO–Al2O3 catalyst that was calcined for
6 h at 350 8C. A portion of this sample was impregnated
with cerium nitrate after calcination before undergoing a
second calcination for 6 h at 250 8C, resulting in the sample
Ce-aC-350. Cu-400 is CuO–ZnO–Al2O3 with the same
nominal composition as Cu-350 calcined for 6 h at 400 8C.
Ce-bC-400 is based on Cu-400, but was impregnated with
cerium nitrate before calcination for 6 h at 400 8C. Ce-aC-
350 and Ce-bC-400 have the same nominal composition.
Copper dispersion measurements by means of selective
oxidation via N2O surface titration [16] were performed in a
Thermogravimetric Analyser (Perkin Elmer TGA 7) using
10 vol.% N2O in Argon (temperature: 75 8C, flow rate
80 ml/min at ambient temperature). Temperature-pro-
grammed reduction (TPR) using 7 vol.% H2 or CO in
Argon (heating rate: 2 K/min, flow rate: 80 ml/min at
ambient temperature), as well as the estimation of the weight
loss of the calcined samples assigned to adsorbed water and/
or surface carbonates [6] was carried out in the same
apparatus. The weight loss (at 260 8C for 1 h in Ar) was
determined prior to reduction (at 260 8C for 2 h in 7 vol.%
H2 in Ar) and dispersion measurements. Dispersion
calculations based on oxygen chemisorption from N2O-
decomposition of the reduced catalyst samples were
Table 1
Chemical composition and physical properties for the catalyst samples
Sample ICP-AESa catalyst composition
mass fractionb
BETc (m2/g) H
Cu Zn Al Ce O
Cu-350 0.29 0.39 0.08 0.00 0.24 72.4 5
Ce-aC-350 0.26 0.35 0.08 0.07 0.24 – 3
Cu-400 0.27 0.40 0.09 0.00 0.24 76.2 1
Ce-bC-400 0.25 0.37 0.08 0.06 0.24 – 1
a ICP-AES: inductively coupled plasma-atomic emission spectroscopy with an
performed on calcined samples.b Normalised mass fractions, i.e. only CuO, ZnO, Al2O3 and CeO2 taken intoc Performed on the calcined samples, prior to reduction.d Thermogravimetric measurements performed in Ar on the calcined samples u
adsorbed water and/or surface carbonates [6].e Dispersion based on oxygen chemisorption (from N2O) of the reduced cataf Performed on the reduced and passivated samples. The copper crystallite size
into account the thickness of the passivation layer. The copper dispersion is calcula
contact with the support [18].
corrected for bulk oxidation of copper according to a
method proposed by Sato et al. [17].
XRD spectra for crystallite size estimation and phase
identification were recorded on Siemens diffractometers D-
5000 (monochromatic radiation) and D-5005 (dichromatic
radiation), respectively. Particle size estimates for the
reduced samples were calculated from experimental line
broadening (linewidth at half maximum) of the Cu(1 1 1)
reflection by using the Scherrer equation and LaB6 as a
standard for correction of instrumental line broadening. To
do this, the catalysts were dried (at 260 8C for 1 h in Ar),
reduced (at 260 8C for 2 h in 7 vol.% H2 in Ar) and
passivated (at ambient temperature for 2 h in 0.5 vol.% O2
in Ar) in the TGA before transportation to the X-ray
diffractometer. The crystallite sizes reported in Table 1 are
corrected for the thickness of the passivation layer assuming
a cubic model for the copper particles [18].
Transmission X-ray absorption spectroscopy (XAS) data
were collected at the Swiss–Norwegian Beam Lines (SNBL)
at the European Synchrotron Radiation Facility (ESRF),
France. Spectra were obtained at the Cu K-edge (8.979 keV)
and Zn K-edge (9.659 keV) using a channel-cut Si(1 1 1)
monochromator. Ce K-edge data (40.444 keV) were
recorded using a Si(3 1 1) monochromator. For the
Si(1 1 1) monochromator higher order harmonics were
rejected by means of a chromium-coated mirror aligned
with respect to the beam to give a cut-off energy of
approximately 15 keV. The beam currents ranged from 130
to 200 mA at 6.0 GeV. The maximum resolution (DE/E) of
the Si(1 1 1) bandpass is 1.4 � 10�4 using a beam of size
0.6 mm � 7.2 mm. Ion chamber detectors with their gases at
ambient temperature and pressure were used for measuring
the intensities of the incident (I0) and transmitted (It) X-rays.
The XANES (X-ray absorption near edge structure)
measurements were performed on two of the four catalysts,
Cu-400 and Ce-bC-400. The amounts of material in the
samples were calculated to give an absorber optical
2O/CO2 lossd (%) Cu dispersione (%) XRDf
Cu crystal
size (nm)
Dispersion
(%)
.7 6.5 23 3.8
.2 4.8 16 5.6
.7 5.4 25 3.5
.7 8.2 23 3.8
estimated detection limit of 0.01–0.03 mg/g. The elementary analysis was
account.
p to the reduction temperature (260 8C) resulted in a weight loss assigned to
lyst samples taking into account copper bulk oxidation [17].
is estimated using the Scherrer equation for the Cu(1 1 1) reflection taking
ted from the crystallite size assuming a cubic particle shape with one face in
M. Rønning et al. / Catalysis Today 100 (2005) 249–254 251
Fig. 1. XRD spectra of the (a) calcined and (b) reduced–passivated samples.
The spectra of the calcined samples were recorded with D-5005. The
reduced samples were recorded with D-5000 except the shown spectra
for Ce-aC-350 which was also recorded with D-5005. The symbols on top of
the spectra are related to the different phases: (&) ZnO, (~) CuO, (!) Cu,
(*) CeO2.
thickness close to 2 absorption lengths. The samples were
ground and mixed with the requisite amount of boron nitride
to achieve the desired absorber thickness. The samples were
then loaded into an in situ reactor-cell [19] and reduced in a
mixture of 5 vol.% H2 or CO in He (purity: 99.995%; flow
rate 30 ml/min at ambient temperature) by heating at a rate
of 6 K/min from room temperature to 350 8C. XANES
profiles were collected during heating of the samples to
follow the reduction progression of CuO in H2 or CO. Cu
and Zn metal foils, Cu2O, CuO and ZnO were used as
reference materials.
The software package WINXAS v3.0 [20] was used for
XANES analysis to obtain qualitative and quantitative
information on copper, zinc oxide and cerium oxide bulk
phases under temperature programmed reduction condi-
tions. The XANES data were energy calibrated, pre-edge
background subtracted (linear fit) and normalised. Principal
component analysis (PCA) was applied for identification of
the number and type of phases in the experimental XANES
spectra. The reference spectra of these phases were then
used in a least-square fitting procedure to determine the
fraction of each phase present [21].
The initial water–gas shift activity was tested in an
externally heated tubular fixed-bed reactor with on-line GC
analysis over a wide temperature range (200–300 8C) [15].
In order to minimize temperature gradients the catalyst
samples were diluted with inert SiC (300–600 mm) taking
into account the falsifying effect of dilution on the measured
catalyst conversion [22]. Parameters like catalyst amount,
particle size, feed composition and gas flow rate were chosen
to perform the activity measurements within the kinetic
regime and away from the thermodynamic equilibrium of
the water–gas shift reaction [23]. The catalysts (0.15–
0.25 mg, 200–400 mm) were pre-reduced in situ at 260 8Cfor 3 h in 7 vol.% H2 in nitrogen, and the activity tests were
performed at a total pressure of 3 bar and 1.3 nl/min total
flow (0.7 nl/min N2, 0.3 nl/min CO, 0.3 nl/min H2O). The
deactivation measurements (feed: 1.0 nl/min N2, 0.3 nl/min
CO, 0.3 nl/min H2O) were performed at 300 8C in order to
accelerate the copper sintering process.
3. Results and discussion
The chemical composition and some physical properties of
the catalyst samples are presented in Table 1. Phase identi-
fication was performed by XRD both before and after cal-
cination. The nominal catalyst loadings were confirmed by
inductively coupled plasma (ICP-AES). The copper content
was also confirmed by the TPR-TGA analysis except for Ce-
aC-350. For this catalyst, the extent of reduction exceeded the
maximum copper equivalent indicating a reduction of com-
ponents other than copper oxide. This effect is object for more
studies and will not be discussed here in further detail.
XRD results confirm that the layered hydrotalcite phase
was obtained and that it breaks down during calcination to
produce mixed oxides (see Fig. 1(a)). The diffraction peaks
associated with ZnO are significantly more intense in the
samples containing ceria, for the calcined (Fig. 1(a)) as well
as the reduced–passivated samples (Fig. 1(b)). This can be
due to low crystallinity of ZnO in absence of ceria or simply
arising from smaller ZnO crystallites. The enhanced ZnO
crystallinity in the samples containing CeO2 may indicate
less interaction between ZnO and Cu in these samples. The
shape of the main diffraction peak for CeO2 at around 298indicates either low CeO2 crystallinity or small crystallites.
It has been stated in the literature that ZnO is not an active
reaction site in the water–gas shift reaction [2,3]. The XRD
data suggest that addition of CeO2 to the copper–zinc system
induces segregation of Cu and Zn phases, whereas CeO2
seems to be well dispersed. The Cu crystallite sizes are
calculated from XRD line broadening analysis (XLBA)
using the Scherrer equation and indicate particle sizes
around 20 nm for the Cu(1 1 1) reflection (see Table 1).
The Cu dispersion obtained from oxygen chemisorption
show Cu dispersions of 4–8% (see Table 1). The sample
impregnated with Ce before calcination shows highest
dispersion (8%) whereas the Ce impregnation after
calcination gives lower Cu surface area. These data are in
good agreement with the dispersions calculated from XRD
data (3.5–5.5%) in terms of the order of magnitude of copper
dispersion in the catalyst samples. However, the dispersions
derived from XLBA show a different trend. For turnover
frequency (TOF) calculations, data from oxygen chemisorp-
tion have been used since this method gives a more direct,
and thus presumably more accurate, estimate of the active
M. Rønning et al. / Catalysis Today 100 (2005) 249–254252
surface than the XLBA results from the reduced–passivated
samples. The XLBA method is a bulk technique and derives
the copper surface area indirectly from the crystallite size.
The XLBA-derived dispersions depend on a structural
model used for converting crystallite size into dispersion
and taking into account the passivation layer (here: cubic
model) and might therefore be a less realistic measure of the
copper surface area. The estimates are useful, however, for
validating the range of the chemisorption-based dispersions.
Currently, selective oxygen chemisorption via the
decomposition of N2O on the Cu surface is a widely used
technique for measuring metallic Cu surface area of copper-
based catalysts [16,17]. Nevertheless, uncertainties remain
concerning N2O decomposition on ceria and whether or not
partially reduced cerium oxide corrupts the measured
dispersion [3,18]. It still remains a matter of discussion
whether chemisorption can be applied, or if this standard
procedure needs some corrections.
In situ XANES measurements of the reduction behaviour
of the catalysts at the Cu K-edge show that the bulk copper in
the catalyst is reduced to the metallic state (see Fig. 2(a)).
The reduction proceeds more easily in CO than in H2 (results
not shown here), since CO has a higher reduction potential
than H2 [24,25]. The transition goes directly from CuO to
Cu, and Cu2O is not detected (see Fig. 2(b)) at the given time
resolution of the recorded XANES spectra. This is in
Fig. 2. TPR-XANES of Cu-400 at the Cu K-edge in 5 vol.% H2 in helium.
Heating rate: 6 K/min, flow rate: 30 ml/min at ambient temperature. (a) Cu
K-edge profiles recorded during the reduction procedure. (b) Change in
phase composition as a function of temperature during the reduction
procedure showing the direct transition from CuO to Cu metal. Cu2O
was not detected during the transition process.
agreement with recent studies [26] and suggests that Cu2O is
not a stable component under the applied reduction
conditions. Agreement also exists between the XANES
(Fig. 2(b)) and the TGA results (Fig. 3) concerning the
transition temperature. Slight differences arise from the
limited time resolution in the XANES experiments.
Furthermore, the TGA experiments where carried out using
a slightly higher H2 concentration than the XANES
experiments. According to Fig. 3, the catalyst Ce-aC-350
shows a notably higher reduction temperature than the other
catalysts. A reason for this might be that copper is partially
covered by cerium oxide. This effect will be addressed in
further studies. Ceria does not enhance the reducibility of
copper in any of the investigated catalysts. This is most
likely because the co-operational effect of Cu and ceria in
the WGS reaction is not present when ceria is in a state of
low oxygen storage capacity [5].
The effect of catalysts composition on the metal particles
and the role of Ce will be more closely investigated by the
proceeding EXAFS analysis [15]. However, both XRD and
the EXAFS analysis support the conclusion drawn by
Grunwaldt et al. [24], stating that the Cu metal particles take
a disk-like shape under reducing conditions. This is reflected
in a pronounced anisotropy in particle shape seen when
analysing the particle size in different crystallographic
directions [15].
Zn and Ce K-edge XANES profiles (not presented here)
show that the bulk of ZnO and CeO2 is not reduced under the
applied conditions up to 350 8C, neither in H2 nor in CO
atmosphere. Since XANES is a bulk technique, surface
oxidation of ZnO or CeO2 is not detected but cannot be
excluded.
The initial WGS activity of the four catalysts is shown in
Fig. 4(a). The selectivity of the reaction was 100% towards
CO2 formation. The chemisorption-based turnover frequen-
cies (TOFs) of the different catalysts are similar within the
experimental accuracy, indicating that the WGS reaction
depends on the exposed Cu surface area only [2,3]. Ceria
seems to enhance copper dispersion (N2O-based dispersion,
Table 1) and catalyst stability (see Fig. 4(b)) when
introduced before calcination. This indicates a promoting
effect of ceria on the active copper sites and implies
sufficient contact between both phases. The XRD spectra of
the reduced–passivated catalysts (Fig. 2(b)) with broad ceria
peaks with low intensity leads to the conclusion of ceria
being present in the form of small crystallites, a prerequisite
for a sufficient interface between copper and ceria. Ceria
included after calcination seems to reduce the chemisorp-
tion-based Cu surface area and slightly increase the catalyst
stability. Assuming coverage of Cu by ceria, a reduced
amount of accessible copper surface atoms would result in a
lower catalytic activity of Ce-aC-350 in terms of CO
reaction rate normalised with metallic copper. Since the
chemisorption-based TOF is similar to the other samples
(Fig. 4(a)), the nature of the active site seems to be
unchanged. This also supports the idea of ceria partially
M. Rønning et al. / Catalysis Today 100 (2005) 249–254 253
Fig. 3. Temperature-programmed reduction with 7% H2 in Argon using a thermogravimetric device. Heating rate: 2 K/min, flow rate: 80 ml/min at ambient
temperature. The weight derivative of the four catalyst samples normalised with the total weight loss during the reduction procedure and the temperature profile
are plotted as a function of time.
covering the copper surface. Thus, diffusion limitations as a
result of ceria coverage may explain the higher reduction
temperature of Cu observed in the TPR profile of Ce-aC-350
(Fig. 3).
It is evident that the preparation procedure for introdu-
cing CeO2 is important for the effect of CeO2. As mentioned
above, uncertainties still exist concerning the contribution
Fig. 4. (a) Turnover frequency (TOF) and (b) short-term deactivation behaviour o
N2O) and corrected for loss of H2O/CO2. The deactivation measurements are pe
disappearance of CO vs. time on stream (TOS). The normalised reaction rates are
corresponding catalysts exhibit a quite similar weight loss, this effect is negligible.
out of nitrogen. This has no further effect on the deactivation behaviour of the c
of ceria to the measured dispersion based on oxygen
chemisorption by means of N2O surface titration. Compar-
ing the N2O-based dispersions of Cu-350 and Cu-400 also
reveals a slight effect of the calcination temperature on the
copper dispersion.
Sintering of the copper particles at elevated temperature
close to the Huettig temperature (325 8C for copper), where
f the catalyst samples. TOF is calculated from oxygen chemisorption (from
rformed at 300 8C and are presented as normalised reaction rates for the
not corrected for the H2O/CO2 uptake of the calcined samples, but as the
The step in the deactivation curve for Cu-400 after 20 h stems from running
atalyst.
M. Rønning et al. / Catalysis Today 100 (2005) 249–254254
copper atoms become mobile, is known to be a severe reason
for catalyst deactivation under reaction conditions [27].
Steam enhances the sintering process. Experiments not
shown here, however, indicate that the slight variations
concerning the water vapour pressure between the four
deactivation experiments have no visible influence on the
deactivation behaviour.
4. Conclusions
A comparative study on the effect of cerium oxide on the
water–gas shift activity of Cu–Zn-based mixed-oxide
catalysts has been carried out using mainly XRD, in situ
XANES and thermogravimetric analysis.
According to XRD data, the crystallinity of the ZnO
phase increases when CeO2 is included into the system. This
might indicate a segregation of Cu- and Zn-containing
phases resulting in a reduced interaction between Cu and
ZnO. On the other hand, cerium oxide seems to be well
dispersed. Activity measurements reveal that the chemi-
sorption-based TOF of the investigated catalyst samples is
independent of composition and preparation conditions.
Together with the in situ XANES results, this leads to the
conclusion that metallic copper is the active component. In
contrast, the stability of the catalyst samples under reaction
conditions depends on composition and preparation condi-
tions. Thus, the stability can be increased by adding cerium
oxide. In order to achieve this effect CeO2 should be added
before calcination of the hydrotalcite-based precursor.
Acknowledgements
This work was supported by the Research Council of
Norway. We gratefully acknowledge the project team at the
Swiss–Norwegian Beam Lines (SNBL) at the ESRF for their
assistance. Elin Nilsen (Department of Materials Technol-
ogy, NTNU) and Egil Haanæs (Department of Chemical
Engineering, NTNU) are gratefully acknowledged for their
assistance with XRD and TGA, respectively.
References
[1] Fuel Cell Handbook, 5th ed., US Department of Energy, NETL, 2000.
[2] C.V. Ovesen, B.S. Clausen, B.S. Hammershøi, G. Steffensen, T.
Askgaard, I. Chorkendorff, J.K. Nørskov, P.B. Rasmussen, P. Stoltze,
P. Taylor, J. Catal. 158 (1996) 170.
[3] N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H.
Ribeiro, J. Catal. 217 (2003) 233.
[4] Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000)
179.
[5] Q. Fy, A. Weber, M. Flytzani-Stephanopoulos, Catal. Lett. 77 (1–3)
(2001) 87.
[6] G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G.
Thomas, B.H. Davis, Appl. Catal. A 258 (2004) 203.
[7] S. Hilaire, X. Wang, T. Luo, R.J. Gorte, J. Wagner, Appl. Catal. A 258
(2004) 271.
[8] T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 15 (1998) 107.
[9] T. Shido, Y. Iwasawa, J. Catal. 136 (1992) 493.
[10] T. Shido, Y. Iwasawa, J. Catal. 141 (1993) 70.
[11] G. Jacobs, A. Crawford, L. Williams, P.M. Patterson, B.H. Davis,
Appl. Catal. A 267 (2004) 27–33.
[12] M. Fernandez-Garcıa, E. Gomez Rebollo, A. Guerrero Ruiz, J.C.
Conesa, J. Soria, J. Catal. 172 (1997) 146.
[13] F. Cavani, F. Trifiro, A. Vaccari, Catal. Today 11 (1991) 173.
[14] M.J.L. Gines, N. Amadeo, M. Laborde, C.R. Apesteguıa, Appl. Catal.
A 131 (1995) 283.
[15] F. Huber, M. Ronning, H. Meland, H. Venvik, D. Chen, A. Holmen, in
preparation.
[16] A. Dandekar, R.T.K. Baker, M.A. Vannice, J. Catal. 183 (1999) 131.
[17] S. Sato, R. Takahashi, T. Sodesawa, K. Yuma, Y. Obata, J. Catal. 196
(2000) 195.
[18] N. Pernicone, T. Fantinel, C. Baldan, P. Riello, F. Pinna, Appl. Catal. A
240 (2003) 199.
[19] F.W. Lytle, R.B. Greegor, E.C. Marques, D.R. Sandstrom, G.H. Via,
J.H. Sinfelt, J. Catal. 95 (1985) 546.
[20] T. Ressler, J. Synch. Rad. 5 (1998) 118–122.
[21] T. Ressler, J. Wong, J. Roos, I.L. Smith, Environ. Sci. Technol. 34
(2000) 950.
[22] R.J. Berger, J. Perez-Ramırez, F. Kapteijn, J.A. Moulijn, Chem. Eng.
Sci. 57 (2002) 4921.
[23] F. Kapteijn, J.A. Moulijn, in: G. Ertl, H. Knozinger, J. Weitkamp
(Eds.), Handbook of Heterogeneous Catalysis, VCH Verlagsge-
sellschaft mbH, 1997(Chapter 9).
[24] J.-D. Grunwaldt, A.M. Molenbroek, N.-Y. Topsøe, H. Topsøe, B.S.
Clausen, J. Catal. 194 (2000) 452.
[25] H. Wilmer, O. Hinrichsen, Catal. Lett. 82 (2002) 117.
[26] J.A. Rodriguez, J.Y. Kim, J.C. Hanson, M. Perez, A.I. Frenkel, Catal.
Lett. 85 (2003) 247.
[27] M.V. Twigg, M.S. Spencer, Appl. Catal. A 212 (2001) 161.
Paper II
Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based
catalysts
Catalysis Letters, in press.
1
Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts
Florian Huber1, Zhixin Yu1, Sara Lögdberg2, Magnus Rønning1, De Chen1, Hilde
Venvik1,*, Anders Holmen1
1Department of Chemical Engineering, Norwegian University of Science and
Technology (NTNU), N-7491 Trondheim, Norway 2Department of Chemical Engineering and Technology, Royal Institute of Technology
(KTH), SE-100 44 Stockholm, Sweden
* Corresponding author; E-mail: [email protected], Tel: +47-73592831,
Fax: +47-73595047
Catalysts containing metals such as Cu, Ni, Fe, Co in their reduced state are often
subjected to passivation procedures prior to characterization. Passivation with N2O or
O2 to create a protective oxide layer also results in a certain degree of sub-surface
oxidation. The heat released during oxidation is a critical parameter. The extent of bulk
oxidation depends on the type of oxidant as well as on the size of the metal particles, as
shown for copper catalysts. The final, meta-stable passivation layer requires a certain
thickness to sustain exposure to ambient atmosphere. The encapsulation of metal
particles in carbon is an efficient method for preserving the metallic state, as
demonstrated for metallic nickel and iron with carbon nanofibers. The use of passivated
samples for characterization of the active, i.e. reduced, catalyst has limited value.
KEY WORDS: passivation; encapsulation; reduced metals; Cu; Ni; Fe; Co; surface and
bulk oxidation; O2; N2O; CO; CO2; carbon.
1. Introduction
Catalysts comprising the transition metals Cu, Ni, Fe and Co are widely applied in
heterogeneous catalysis. In most cases, the active catalysts contain these transition
metals in their reduced, i.e. metallic, state rather than as an oxide. Under oxidizing
conditions, such as in air, the metallic state of Cu, Ni, Fe and Co is unstable. Depending
2
on the conditions, the metals will be partly or completely reoxidized to their stable
oxides: Cu → Cu2O/CuO [1], Ni → NiO, Fe → Fe3O4/Fe2O3 [2], Co → CoO/Co3O4 [3-
4].
In order to characterize material properties relevant to the catalytic performance of the
catalyst, the metal particles should be studied in the activated state, preferably during
the catalytic reaction, i.e. in situ. Several in situ studies have shown metal catalysts to
undergo dynamic structural changes depending on the reaction atmosphere (oxidizing or
reducing) [5-9]. However, in situ characterization can be time-consuming or require
equipment not readily available. To be able to characterize reduced metal catalysts, a
range of protection methods are applied to retain the reduced state of metal catalysts
during characterization or transfer to the characterization chamber:
1. Sealing in a container under reducing/inert atmosphere after reduction [10]
2. Encapsulation of the reduced metal particles
a. by deposition of polyethylene [11-13] or 1-butene films [14] on the
surface, analogous to the protection of bulk metals against corrosion
by coating with polymer films
b. by growing a layer of carbon (such as carbon nanofibers or paraffinic
wax) around the reduced metal particles [15-17]
3. Surface passivation of reduced metal particles by controlled reoxidation to
create a thin, protective oxide layer. Further oxidation of the bulk is inhibited
by diffusion limitation [18]. A quasi- or meta-stable reduced metal phase is
maintained below the protection layer. Oxidant gases that have been applied
include:
a. O2/air in inert gas [1,18-23]
b. N2O in inert gas [9,24-27]
4. Formation of combined carbonaceous and oxidic layers, applying
a. CO2/O2, CO2/H2O or CO2/H2O/O2, optionally in inert gas [28-31]
b. CO/H2 and O2/air (consecutive) [32]
3
Our group has previously reported the successful application of sealed quartz capillaries
for preserving activated catalysts [10]. In this study, we evaluate and share our
experience with other protection methods, namely O2 and N2O passivation, and relate
our findings to reports on passivation in the literature. In addition, the encapsulation in
carbon nanofibers (CNF) for a range of catalysts used in the ongoing research is
reported and proposed as a possible method.
2. Experimental
2.1. Catalyst samples
The copper catalysts are mixed metal oxides derived from hydrotalcite precursors. Cu-
350 and Cu-400 contain CuO, ZnO and Al2O3 after the final calcination at 350 and 400
°C, respectively, and their hydrotalcite precursors were prepared by coprecipitation. Ce-
bC-400 was made from Cu-400 by incipient wetness impregnation of the dried sample
with a cerium nitrate solution. Details on preparation, sample notation and catalyst
properties can be found elsewhere [33].
The Ni-Fe catalyst is designed for carbon nanofiber (CNF) production and contains both
metals in a molar ratio Ni:Fe = 8:2. The catalyst, composed of NiO, Fe oxides (Fe2O3
and Fe3O4) and Al2O3 after calcination, was prepared by coprecipitation to produce a
hydrotalcite precursor. The CNF were synthesized by catalytic chemical vapour
deposition from C2H4/CO/H2 (30/10/10 ml/min) at 600 °C for 1 h. Further details on
catalyst preparation and properties have been previously reported [15].
The Fischer-Tropsch (FT) catalyst containing 12 wt% cobalt was prepared by incipient
wetness impregnation of silica (PQ corp. CS-2133) with an aqueous solution of
Co(NO3)2·6H2O. The impregnated powder was dried in air at 120 °C for 3 h and
calcined at 300 °C for 16 h, increasing the temperature from 120 °C to 300 °C at a rate
of 1 °C /min. Further details on this type of catalyst have been published elsewhere
[3,34].
4
2.2. Characterization
Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) was used to
determine the actual amount of copper in the catalyst samples after calcination. The
samples were dissolved in hydrochloric acid prior to this analysis without any visible
residues.
Copper dispersion and passivation measurements by means of N2O decomposition
[22,27,35,36] according to the exothermic reaction
2 Cu + N2O → Cu2O + N2 (1)
were performed in a Thermogravimetric Analyser (Perkin Elmer TGA 7) using approx.
50 mg sample and 10 vol-% N2O in argon (reaction temperature: 75 ºC, flow rate: 80
ml/min at ambient pressure/temperature, 1 h under N2O atmosphere until the baseline
was stabilized). Temperature-programmed reduction (TPR) using 7 vol-% H2 in argon
(260 ºC, 2 h, heating rate 2 K/min, flow rate 80 ml/min at ambient pressure/temperature)
as well as the estimation of weight loss assigned to adsorbed water and/or surface
carbonates of the calcined samples [37], were carried out in the same apparatus. TPR
was also carried out with 7 vol-% CO in argon over Ce-bC-400, with the reduced
catalyst being cooled down in CO/Ar prior to N2O titration. The weight loss (at 260 ºC
after 1 h in Ar) was determined prior to reduction (at 260 ºC for 2 h) and dispersion
measurements. Dispersion calculations based on oxygen chemisorption from N2O
decomposition with a simplified reaction stoichiometry of Cu:O = 2 corresponding to
equation (1) were corrected for bulk oxidation of copper according to the method
proposed by Sato et al. [38].
The passivation of the copper catalysts with O2 was also performed in the TGA. In a
typical experiment, the catalyst sample was dried (260 ºC, 1 h, Ar), reduced (260 ºC, 2
h, 7 vol-% H2 in Ar) and passivated (ambient temperature, 2 h, 1 vol-% oxygen in Ar,
100 ml/min). The exothermic reoxidation can be written as follows [1,20]:
2 Cu + ½ O2 → Cu2O (2)
5
Hence, a simplified reaction stoichiometry of Cu:O = 2 is assumed. Pernicone et al. [20]
have discussed the reaction stoichiometry and the possibility of non-integer
stoichiometry. For the erroneous impact of the support (especially reducible metal
oxides like ceria or zirconia, and partly ZnO) on the quantification of dispersion and
passivation extent, we refer to Bartley et al. [27]. Concerning the stability of Cu2O
phases, Palkar et al. [40] report that cubic Cu2O is more stable than monoclinic CuO at
small crystallite sizes (< 25 nm). This is related to the increasing ionic character of
solids with decreasing particle size and is also dependent on calcination conditions.
XRD spectra for crystallite size estimation and phase identification of the (O2-
passivated) copper catalysts were recorded on Siemens diffractometers D-5000
(monochromatic CuKα-radiation) and D-5005 (dichromatic CuKα+β-radiation),
respectively. Particle size estimates (dP) for the reduced-(O2)passivated copper samples
were calculated by X-ray line broadening analysis (XLBA, linewidth at half maximum,
B) of the Cu(111) reflection at the Bragg angle Θ = 21.67° using the Scherrer equation
and LaB6 as a standard for correction of instrumental line broadening (Be)
2 2 cos( )
P
e
KdB B
⋅ λ=− ⋅ Θ
(3)
with λ = 1.5418 Å, the wavelength for Cu Kα, and K = 0.89, the Scherrer constant. The
value of the constant factor K depends on the definition of B, being set equal to 1.00
when using the integral breadth and 0.89 when using the full width at half maximum
[39]. The crystallite size of the reduced-passivated samples reported in Table 1 takes
into account the thickness of the passivation layer assuming a cubic shape for the copper
particles.
The passivation of the Co/SiO2 catalyst was characterized by XRD using the Siemens
D-5005 diffractometer. The sample (1 g) was reduced in flowing hydrogen at
atmospheric pressure and 350 °C for 16 h (heating rate: 5 °C/min up to 70 °C and then 1
°C/min up to 350 °C). According to Storsæter et al. [34] and references therein, most of
the reducible cobalt oxide is reduced after this procedure. Subsequently, the sample was
6
cooled down to ambient temperature, and the quartz reactor was flushed with N2 5.0 for
2 h. The consecutive passivation was carried out in either of two ways: In one
experiment, a gas mixture of 0.5 vol-% of O2 in nitrogen was used (first, 1 h at 50
ml/min and then 1 h at 140 ml/min). In a second experiment, 5 vol-% of N2O in
nitrogen was applied (2 h at 140 ml/min). The XRD analysis was started within 2 h after
passivation.
Transmission X-ray absorption spectroscopy (XAS) data were recorded for calcined,
reduced-O2-passivated and CNF-encapsulated Ni-Fe samples at the Swiss-Norwegian
Beamline (SNBL) at the European Synchrotron Radiation Facility (ESRF) in France.
Spectra were obtained at the Fe K-edge (7.112 keV) and Ni K-edge (8.333 keV) using a
channel-cut Si(111) monochromator. Higher order harmonics were rejected by means of
a chromium-coated mirror aligned with respect to the beam to give a cut-off energy of
approximately 15 keV. The software package WINXAS v3.1 [41] was used for XANES
analysis (X-ray absorption near edge structure) to obtain qualitative and quantitative
information on Ni/Fe oxide and metallic phases. The XAS data were calibrated, pre-
edge background subtracted (linear fit) and normalized. XAS spectra were collected
both for the calcined material, in the reduced-passivated state and after use in CNF
production. The passivation was conducted with about 1 vol-% of O2 in nitrogen. Ni and
Fe metal foils, NiO, α-Fe2O3 and Fe3O4 were used as reference materials. The phase
composition of the catalyst samples under study was determined by linear combination
of the reference XANES profiles applying a least-square fitting procedure in WINXAS.
3. Results
3.1. Copper catalysts
A representative experimental curve from TGA measurements of the copper-containing
samples is shown in Figure 1. Initially, water and/or carbonate species adsorbed on the
surface [37] are removed by increasing the temperature to the final reduction
temperature under Ar atmosphere. In this way, the weight change resulting from
desorption of adsorbed species does not interfere with the weight change caused by the
subsequent reduction. CuO is reduced to metallic copper and oxygen is released as
water molecules during TPR, decreasing the sample weight thereby. After reduction and
7
flushing with Ar, N2O decomposition was carried out to reoxidize the copper atoms in
or close to the surface. By plotting oxygen uptake versus time and scaling the abscissa
with the square root of time, surface and bulk oxidation contributions can be separated.
Since bulk oxidation is diffusion limited, it will appear as a linear segment in such a
diagram. Further details about this approach can be found elsewhere [38,42].
Weight losses caused both by adsorbates and copper reduction for the three copper
samples are given in Table 1. The copper content as determined by TGA is in good
agreement with the ICP-AES results. The copper dispersion of the samples varies
between 5 and 8 % assuming the Cu/N2O = 2 stoichiometry according to equation (1).
The oxygen uptake during N2O decomposition is correspondingly 7 – 11 %. This
indicates that the passivation layer is relatively thin and that the reoxidation of copper
does not reach far into the bulk. In contrast, Cu passivation with O2 results in a
considerably thicker passivation layer, extending to about 40 % of the Cu atoms.
The crystallite size of the Cu core embedded in the oxide layer is estimated from XRD
spectra recorded for reduced-(O2)passivated samples (Figure 2a). XRD spectra of
calcined samples before reduction/passivation are included for comparison (Figure 2b).
For all three catalyst samples, the size of the Cu core (dC) is estimated to about 20 nm
(Table 1). Using a cubic model the average size of the copper particles (dP), including
core and passivation layer, can be estimated by the following equation:
31
(1 )P Cp
d dx
= ⋅−
(4)
with xp being the fraction of the passivated Cu atoms. According to this estimation, the
thickness of the oxide layer lies in the range of a few nanometers (Table 1), which is in
agreement with literature ([20] and references therein). The structural model applied
does not, however, take into account possible structural changes during passivation.
Several studies have shown that both surface structure as well as particle shape may
change (dynamically) upon changes in the gas phase conditions [5-9]. Despite the
significant degree of reoxidation, the oxide layer does not appear in the XRD spectra
8
(Figure 2a). The passivation layer is only a few nanometers thick and will therefore
appear amorphous to XRD [5,43]. Thus, from XRD alone, the passivation layer appears
insignificant. Kvande et al. [44] identified a crystalline Cu2O phase in addition to
metallic copper by XRD after passivation of a reduced Cu/CNF catalyst in O2/He,
indicating a significant degree of bulk oxidation.
Pernicone et al. [20] report the passivation of a copper-containing catalyst with 60 % of
the Cu atoms being reoxidized. This copper catalyst contained smaller copper particles
(Cu core: 10.2 nm, with passivation layer: 13.8 nm) than the samples in Table 1. If we
assume that the passivation layer must reach a certain thickness in order to protect the
bulk against further oxidation, the fraction of reoxidized metal atoms should increase
with decreasing particle diameter (equation (4)). This trend was also observed within
our passivation studies. Another Cu-Zn-Al mixed oxide catalyst, prepared by
coprecipitation under different conditions but with a similar composition as the copper
catalysts described in chapter 2.1, showed a passivation degree of about 58 % (with
oxygen). The size of the copper core was estimated to 10.6 nm, and the particle size,
including the passivation layer, to 14.2 nm, in good agreement with the results obtained
by Pernicone et al.
Figure 3 shows the normalized weight increase of Ce-bC-400 during N2O titration as a
function of the square root of time, based on the total weight loss upon reduction. The
pre-reduction was carried out with H2/Ar and CO/Ar. In both cases, the reduced catalyst
was cooled down in the reduction gas prior to N2O titration. The sample pretreated with
CO/Ar exhibits a slower oxidation kinetics and a lower total oxygen uptake than the
sample pretreated with H2/Ar. The dispersions determined from the dispersion-square
root of time plot correspond to approx. 8.2 % and 6.2 % for pre-reduction with H2 and
CO, respectively. We assume that carbonaceous species on the surface of the reduced
Cu particles slow down and confine the re-oxidation of metallic Cu in the case of pre-
reduction with CO.
9
3.2. Nickel-iron catalyst
The Ni-Fe catalyst was investigated by XAS, in particular in the XANES region. By
recording data at both the Ni K-edge and the Fe K-edge, the oxidation state of both
metals can be determined. Figure 4 and 5 show the XAS spectra of Ni and Fe,
respectively, for the calcined and reduced-passivated samples and for the sample used in
the synthesis of CNFs. The spectra of NiO, a Ni foil, α-Fe2O3, Fe3O4 and a Fe foil are
used as references.
Figure 6 shows a least squares fit of a linear combination of the XANES spectra of
model compounds, reconstructing the profile of a catalyst sample. NiO and Ni foil are
combined in order to model the XANES profile of the reduced-passivated Ni-Fe catalyst
at the Ni K-edge. According to this fit, 70 % of the Ni atoms in the sample are in the
metallic state after O2-passivation while 30 % are reoxidized to NiO. Table 2 contains
the phase composition for the three Ni-Fe catalyst samples recorded at both the Ni and
the Fe K-edge. The Ni sample profiles are modelled with two reference materials, since
Ni(0) and Ni(2+) are the most common oxidation states of nickel. The Fe sample
profiles are reconstructed with three reference materials, since Fe(0) under low partial
pressure of oxygen first oxidizes to Fe3O4 (= FeO·Fe2O3) and then further to Fe2O3 [2].
The XANES analysis of the Ni K-egde of the sample used for CNF production shows
that essentially all Ni is preserved as Ni(0). Reduced Ni particles encapsulated in carbon
nanofibers (see also [15]) thus preserve their metallic state also upon exposure to air.
The encapsulation in carbon is better than passivation by O2, where about 30 % of the
Ni atoms are reoxidized to NiO, assuming that both H2 reduction and reduction under
CNF production conditions result in complete conversion of NiO to metallic Ni prior to
oxygen exposure. About 80 % of the iron atoms in the CNF-encapsulated sample retain
the metallic state, as compared to about 60 % with oxygen passivation. The remaining
20 % or 40 % are present as a mixture of Fe3O4 and Fe2O3. The 20 % non-metallic Fe
atoms in the CNF-encapsulated sample have either been reoxidized under air exposure
or they represent a fraction of iron atoms not reducible under the given pre-reduction
and CNF-production conditions (see also [45]). Correspondingly, the oxide fraction in
10
the O2-passivated sample may also contain the non-reducible part, in addition to the
oxide formed under passivation conditions.
3.3. Cobalt catalyst
Figure 7 shows XRD spectra of the calcined Co/SiO2 catalyst, as well as after reduction
and passivation with oxygen. The XRD profile recorded after passivation with N2O is
identical to the one after passivation with oxygen, and is hence not included in the
figure. The nine main peaks in the spectrum for the calcined sample can be assigned to
Co3O4. In addition, there are two broad peaks between 10 and 30° stemming from the
silica support. For the passivated sample, the characteristic Co3O4 peaks have vanished
and the apparent peaks can be assigned to CoO and metallic cobalt. The broad peak at
36 – 37° is assigned mainly to CoO, but it can not be excluded that Co3O4 contributes to
the right shoulder of this peak. The broad shoulder between 46 – 54° can be assigned to
different metallic cobalt phases.
4. Discussion
From dispersion measurements of copper catalysts [20,22,27,38,42], it is well-known
that the reoxidation of the reduced metal with O2 or N2O is not limited to the surface,
but also reaches sub-surface layers. As a consequence, these measurements are either
conducted at low temperatures (for O2 [20]) or the contribution of the bulk oxidation is
taken into account applying a diffusion model (for N2O [38,42]).
4.1. Thermodynamics and kinetics
Standard reaction enthalpies for the oxidation of Cu, Ni, Fe and Co by O2 and N2O are
shown in Table 3. They are estimated from the standard enthalpies of formation of the
compounds participating in the reactions, based on bulk data. The oxidation processes
investigated in this study are limited to a few nanometers into the bulk of the metals.
Thus, pure bulk data may not be accurate for quantifying the energies involved in these
reactions, but can be used for qualitative comparisons.
The values for the O2 oxidation are calculated for 1 mole O2 and not for 1 mole O,
which one might prefer when comparing with the values for N2O oxidation. We prefer
11
the O2-based enthalpies because each time an O2 molecule reacts with the metal, two
oxygen atoms are involved simultaneously in the oxidation reaction releasing the double
amount of energy of one oxygen atom. Thus, we prefer to compare N2O and O2 based
on the amount of energy these two molecules contain in total.
The following trends can be deduced from the thermodynamic data in Table 3:
1. The oxidation reactions are all exothermic. This is reflected in the well-
known pyrophoric behaviour of reduced transition metal catalysts in
contact with air.
2. The oxidation reaction with N2O is less exothermic than with O2, i.e.
during the consumption of one O2 molecule more energy is released as
per N2O molecule.
3. For N2O, the energy release during oxidation of the different metals
increases in the following order: Cu < (Ni, Co) < Fe. The same order is
obtained for oxidation with O2, if Fe2O3 is the main phase formed. The
stability of the metallic phase decreases in the same order.
4. Copper may be considered the thermodynamically most stable among the
four metals. Ni and Co behave relatively similar, while for Fe the energy
released depends strongly on the type of oxide formed. These
thermodynamic properties are to some extent responsible for Ni, Co and
Fe being more difficult to reduce than Cu, as reflected in the typical
reduction temperatures for these metals.
Apart from thermodynamics, there are also kinetic effects to be considered. According
to Pernicone et al. [20], temperature is a main parameter influencing the extent of bulk
oxidation using O2. Because the passivation process is a non-catalytic gas-solid type
reaction [46], the passivation rate and thus the extent of bulk oxidation depends on the
diffusion rate in the solid material with the diffusion coefficient being a function of
temperature. The kinetic rate constant of the oxidation reaction itself is also enhanced
by increasing temperature, although we assume that the diffusion process is the rate-
limiting step during bulk oxidation with O2. In principle, this is also the case for
passivation with N2O [27,38,42].
12
Via the interrelation between temperature and passivation kinetics, the thermodynamics
affects the kinetics of the process. Because of the exothermic nature of the oxidation
reactions, energy is released and the local temperature increases. The energy release
may thus enhance the diffusion rate and increase the extent of bulk oxidation.
According to this, the extent of bulk oxidation should be higher for passivation with O2
than with N2O, not taking into account the effect of different oxidation kinetics for O2
and N2O. Bartley et al. [27] identify local temperature increases as a significant cause
for bulk oxidation of Cu in dispersion measurements with O2 and N2O. They stress the
need for low oxidant concentrations, large sample amounts and characterization setups
with efficient heat dissipation, resulting in a more controlled release of the heat of the
oxidation reactions.
4.2. Copper catalysts
The higher extent of bulk oxidation with O2 as compared to N2O (Table 1) we assume
does not only stem from the effect discussed above, especially since the passivation
with N2O is conducted at higher temperatures and higher gas concentration than the
passivation with O2. Other factors could be the somewhat higher sample amount used
for the O2 passivation (because of requirements for the subsequent XRD analysis) and
the different reaction rate constants for N2O and O2, with N2O enabling a more
controlled passivation. This might also be reflected in that Cu dispersion measurements
conducted with O2 are performed at very low temperatures while the N2O method can
be performed at ambient temperatures and even up to 90 °C.
The question remains whether the thin oxide layer produced with N2O is sufficient to
protect the metallic copper core against further bulk oxidation in air. The observed
dependence of the passivation degree on the size of the copper particles (Chapter 3.1,
and similar discussion about higher reactivity of small metal particles in [27] and
references therein) implies that a certain, stable passivation layer thickness is required.
Furthermore, the similar XRD profiles for the Co catalyst sample passivated in O2 and
N2O (Figure 7) point in the same direction. Bartley et al. [27] report that they could not
prevent some continued reoxidation of copper after applying a N2O passivation
procedure that resulted in a thin oxide layer around the metallic copper core. If N2O
13
initially creates a passivation layer thinner than O2 does, yet not stable enough to
survive in ambient air atmosphere, further bulk oxidation leads to a thicker, meta-stable
passivation layer of similar thickness as created by O2 passivation. Hence, even though
N2O does not lead to a stable passivation layer, it can be used in a first passivation step
to grow an oxide layer that will slow down further reoxidation with O2 and prevent local
temperature increase, thereby keeping the final thickness of the passivation layer at a
minimum.
Moreover, in analogy with electrochemical processes [47], it is possible that Cu(2+)
might be formed in ambient atmosphere at the surface of the passivation layer by
adsorbing oxygen and/or H2O (from air moisture) leading to formation of Cu(OH)2
and/or CuO at the surface [2].
4.3. Nickel-iron catalyst
The Ni(2+) detected by XAS in the Ni-Fe-Al sample is identified as NiO. This
oxidation state could also be assigned to a Ni-Al spinel structure known to form at high
calcination temperatures [48]. However, Ni incorporated in such a structure is less
reducible, and would not be reduced under the conditions applied here [15]. The spinel
structure is detectable in XRD if formed to a significant extent [48], but was not found
for the Ni-Fe-Al sample calcined at 480 °C [15]. Ni K-edge data of the CNF-
encapsulated sample also show that almost 100 % of Ni remains in the metallic state
after CNF production at 600 °C. Consequently, the sample probably does not contain
XRD amorphous Ni-Al-spinel, since basically all Ni can be reduced. The reduced-
passivated sample is treated up to the same temperature. It is thus likely that the
detected Ni(2+) stems from NiO formed during passivation.
The reliability of the phase composition data obtained by the least-squares fitting
procedure depends on a reasonable choice of reference materials, i.e. fitting of the
experimental curve might still be possible even when using unreasonable model
compounds. The quality of the recorded spectra is also vital to the fitting procedure.
According to the XANES fit, the calcined sample contains around 7 % Ni metal. It
appears unlikely that Ni particles oxidized in air at 480 °C contains significant amounts
14
of metallic nickel. The XANES curves of the calcined sample and the NiO reference
(not shown) display similar profiles. The difference between these curves is a certain
deviation in the high energy part of the recorded spectra, believed to emerge from
measurement (e.g. thickness effects) and/or data treatment rather than being related to
the material itself. As a consequence, the 7 % of Ni metal obtained for the calcined
sample should be treated as an error, indicating the accuracy limit of the linear
combination approach for this system. The values presented in Table 2 should thus be
viewed as average values with an accuracy not better than ± 5 % composition
percentage. For comparison, the standard deviations given by Overbury et al. [49]
represent a qualitative assessment of the accuracy limit of a linear combination of
XANES data.
Yu et al. [15] performed an EXAFS analysis for the Ni-Fe sample investigated here. In
their structural model, the extent of reoxidation during passivation was estimated to
about 26 % for Ni and 35 % for Fe, when comparing the coordination numbers (N) of
the oxygen shells of the calcined and reduced-passivated samples (Iron: first Fe-O shell
N = 1.1 and second Fe-O shell N = 3.5 for the calcined sample, first Fe-O shell N = 0.3
and second Fe-O shell N = 1.3 for the passivated sample; Nickel: first Ni-O shell N =
5.4 for the calcined sample, first Ni-O shell N = 1.4 for the passivated sample). These
values are in the same range as the values in Table 2.
The thermodynamic prediction that Ni is more stable than Fe (when forming Fe2O3) can
be supported by the XAS data. The extent of bulk oxidation upon O2-passivation of Cu
in the TGA (Table 1) is, however, slightly higher than the one obtained for Ni with
XAS. The stability of the metal phase might depend on particle size and co-additives in
the catalytic material, as can be seen from the comparison of the three copper catalysts,
thus making interrelationships more complex than comprised in a mere discussion of
thermodynamic trends. As previously mentioned, a practical aspect with regard to hot
spots is for example the choice of equipment used during passivation [27]. The Cu
catalysts were passivated in a TGA sample pan with small dimensions, i.e. each sample
was concentrated in a small volume implying a certain risk for local temperature
increase. In contrast, the Ni-Fe catalyst was passivated in a conventional calcination
15
reactor with large dimensions, the sample being distributed over a larger area compared
with the TGA pan and a lower risk for hot spots.
4.4. Cobalt catalyst
XRD was used to qualitatively identify the main phases present in the samples after
passivation (Figure 7). The oxidation of cobalt is similar to the oxidation of nickel from
an energetic point of view (Table 3). Not taking into account kinetics, it should
therefore be possible to passivate cobalt in the same manner as nickel. The XRD profile
of the O2-passivated sample in Figure 7 confirms the presence of metallic cobalt
encapsulated in a CoO shell. The reduction of Co3O4 to metallic cobalt proceeds via
CoO [3,4]. Under mild conditions (low temperature and low oxygen concentration) the
reoxidation may only proceed to the intermediate CoO (in analogy with the behaviour of
copper), since CoO is stable under certain conditions [2]. Yet, the existence of a Co3O4
phase (a mixture of Co(2+)O and Co(3+)2O3 [2]) in the passivated samples can not be
ruled out. The CoO detected by XRD may originate either from the passivation process
or from CoO not reduced during the reduction step. Cobalt oxide supported on silica is
easily reduced from Co3O4 to CoO independent of particle size, morphology and
support properties, but the reduction of CoO to Co is more difficult and depends on
particle size and support properties [4]. According to Storsæter et al. [34], about two-
thirds of the silica-supported cobalt is reducible (determined by oxygen titration) upon
reduction in hydrogen at 350 °C for 16 h. About one-third remains unreduced even upon
a subsequent TPR up to 900 °C, and might be assigned to cobalt silicate or cobalt oxide
encapsulated in silica [4,21]. An in situ XAS analysis suggested that the degree of
reduction is around 80 % which is somewhat higher than determined by oxygen titration
[3].
4.5. Encapsulation in carbon
Encapsulation of reduced metals in carbon appears to be an efficient way to protect
metallic phases against reoxidation in air. The encapsulation in CNF is, however, not a
generally applicable technique, since it applies only to catalysts active for carbon
formation. Jacobs et al. [17] used the solidified, paraffinic Fischer-Tropsch wax product
as an encapsulation matrix to preserve the reduced state of cobalt in a used Fischer-
16
Tropsch catalyst. A more generally applicable encapsulation procedure mentioned in
literature is the deposition of polyethylene [11-13] or 1-butene [14] films at the catalyst
surface, in analogy with the protection of bulk metals against electrochemical corrosion
by coating with polymer films.
4.6. Methods combining carbon and oxide formation
The treatment in a mixture of CO and H2 to form a layer of carbonaceous species on the
catalyst surface was reported to be superior to passivation with oxygen for reduced
cobalt particles [32]. However, a measured temperature increase upon air exposure
(lower than for passivation with oxygen [32]) can be interpreted as an indication for the
formation of an oxide layer. The conditioning of reduced catalysts with CO2 at elevated
temperature (e.g. 200 °C) has been applied to the passivation of Ni-Cr/Al2O3 steam-
reforming-methanation catalysts [50]. Furthermore, the XPS and XRD data of Cu-based
[51] and Cu-Co-based [52] catalysts that had been exposed to CO/CO2-containing gas
mixtures during the synthesis of alcohols are relevant also to passivation. Metallic
copper and Cu2O were identified by XRD for a Cu-ZnO-Al2O3 catalyst after use in
methanol synthesis [51]. However, we suggest to assign the corresponding XPS spectra
to Cu(1+) rather than to Cu(0) or a mixture of both species, in agreement with literature
[47,53] and the well-known instability of Cu(0) in oxidizing atmosphere (i.e. air). For
the Cu-Co-based catalysts used in the synthesis of higher alcohols, metallic Co as well
as oxidized Co species were identified by XPS [52].
Passivation strategies that involve CO and/or CO2 without the formation of significant
amounts of coke or wax around the metal particles may comprise both formation of
carbonaceous species and an oxide layer (at the latest when exposed to air), depending
on the conditions used and the metal to be passivated. Deactivation studies carried out
on ceria-supported precious metal catalysts indicate that CO and CO2 can adsorb at
catalyst surfaces during reactor shutdown forming stable surface carbonates [54].
Vissokov [29] claims that the rate of oxidation could be lowered, hence the passivation
improved, by the ability of certain metals to form surface complexes (e.g. metal
carbonyl), but oxidation could not be prevented completely under the conditions used.
These findings are in line with the results shown in Figure 3. Pretreatment with CO
17
instead of H2 resulted in a slower oxidation kinetics and a lower oxygen uptake during
N2O titration, which might be related to the formation of carbonaceous surface species
limiting the access of N2O. Such methods may therefore be considered intermediate
between encapsulation with carbon and formation of a protective oxide layer, with the
carbonaceous surface layer suppressing the extent of bulk oxidation.
5. Conclusions
Studies of reduced metal catalysts based on Cu, Ni, Fe and Co show that the result of
passivation procedures should be monitored, in order to quantify the extent of bulk
oxidation. Heat released during exothermic oxidation reactions appear to be a critical
parameter, since the local temperature and hence bulk diffusion and the final extent of
bulk oxidation may be increased. Furthermore, the (oxidic) passivation layer requires a
certain thickness in order to be stable and prevent further bulk oxidation in ambient air
atmosphere.
Passivation of reduced catalysts is not an ideal strategy for characterization of reduced
systems that are unstable in air. In situ measurements are prefered, but passivated
samples can be used to some extent, e.g. for estimating particle size with XRD [20,33],
keeping in mind the limited relevance and accuracy of data derived from this approach.
The use of passivated samples for a detailed X-ray line broadening analysis, in
correlation with catalytic activity, distinguishing between particle size and strain effects
is questionable and requires evaluation by a parallel in situ approach. Through the
passivation procedure, strain may be introduced into the crystal lattice by the oxidation
of the outer metal layers. Possible morphological changes depending on the
reduction/oxidation potential of the surrounding atmosphere may also complicate the
interpretation of the characterization results. Finally, the discussion of active reaction
sites based on the characterization of passivated samples is disputable.
Encapsulation of reduced metal particles by a protective layer of carbon is found to
efficiently protect Ni particles. For certain catalysts, CNF encapsulation could be
preferred instead of passivation by a protective oxide layer. Provided that the carbon
layer is impermeable to oxygen when exposed to air, the reduced metal particles may be
18
preserved in their metallic, hence active, state. Possible morphological changes as a
result of a change in the reduction/oxidation potential of the surrounding atmosphere
should be of less concern than for passivation with oxygen. However, the particle
morphology might be affected by the CNF growth process [55].
With this paper, we want to call more attention to the passivation procedure, its effect
and limited usability as sample treatment prior to characterization.
Acknowledgements
This work was supported by the Research Council of Norway through Grant No.
140022/V30 (RENERGI). Statoil ASA through the Gas Technology Center NTNU-
SINTEF is also acknowledged for their support. We gratefully acknowledge the project
team at the Swiss-Norwegian Beam Lines (SNBL) at the ESRF for their assistance. Elin
Nilsen (Department of Materials Technology, NTNU) and Egil Haanæs (Department of
Chemical Engineering, NTNU) are gratefully acknowledged for their assistance with the
XRD devices and the TGA, respectively. Hilde Meland and Cathrine Bræin Nilsen
(Department of Chemical Engineering, NTNU) are gratefully acknowledged for
preparing the copper catalysts.
References
[1] J.A. Rodriguez, J.Y. Kim, J.C. Hanson, M. Perez, A.I. Frenkel, Catal. Lett. 85
(2003) 247.
[2] A.F. Hollemann, E.Wiberg, Lehrbuch der Anorganischen Chemie, 101st ed.
(Walter de Gruyter, Berlin, 1995).
[3] Ø. Borg, M. Rønning, S. Storsæter, W. van Beek, A. Holmen, Appl. Catal. A
accepted.
[4] D.G. Castner, P.R. Watson, I.Y. Chan, J. Phys. Chem. 94 (1990) 819.
[5] R.B. Greegor, F.W. Lytle, J. Catal. 63 (1980) 476.
[6] J.D. Grunwaldt, A.M. Molenbroek, N.Y. Topsøe, H. Topsøe, B.S. Clausen, J.
Catal. 194 (2000) 452.
[7] M.M. Günter, B. Bems, R. Schlögl, T. Ressler, J. Synch. Rad. 8 (2001) 619.
19
[8] P.L. Hansen, J.B. Wagner, S. Helveg, J.R. Rostrup-Nielsen, B.S. Clausen, H.
Topsøe, Science 295 (2002) 2053.
[9] W.P.A. Jansen, J. Beckers, J.C. van den Heuvel, A.W. Denier van den Gon, A.
Bliek, H.H. Brongersma, J. Catal. 210 (2002) 229.
[10] M. Rønning, T. Gjervan, R. Prestvik, D.G. Nicholson, A. Holmen, J. Catal. 204
(2001) 292.
[11] M.P. Kapoor, Y. Ichihashi, K. Kuraoka, W.-J. Shen, Y. Matsumura, Catal. Lett. 88
(2003) 83.
[12] W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 79 (2002) 125.
[13] W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 83 (2002) 33.
[14] S. Zhao, R.J. Gorte, Catal. Lett. 92 (2004) 75.
[15] (a) Z. Yu, T. Vrålstad, M. Rønning, E. Ochoa-Fernández, D. Chen, A. Holmen,
Manuscripts, part 1 and 2, in preparation; (b) Z. Yu, PhD Thesis, Norwegian
University of Science and Technology, Trondheim, 2005.
[16] M.L. Toebes, J.H. Bitter, A.J. van Dillen, K.P. de Jong, Catal Today, 76 (2002) 33.
[17] G. Jacobs, T.K. Das, P.M. Patterson, J. Li, L. Sanchez, B.H. Davis, Appl. Catal. A
247 (2003) 335.
[18] T. Huizinga, J. van Grondelle, R. Prins, Appl. Catal. 10 (1984) 199.
[19] E. Crezee, P.J. Kooyman, J. Kiersch, W.G. Sloof, G. Mul, F. Kapteijn, J.A.
Moulijn, Catal. Lett. 90 (2003) 181.
[20] N. Pernicone, T. Fantinel, C. Baldan, P. Riello, F. Pinna, Appl. Catal. A 240
(2003) 199.
[21] R. Riva, H. Miessner, R. Vitali, G. Del Piero, Appl. Catal. A 196 (2000) 111.
[22] J. Ma, C. Park, N.M. Rodriguez, R.T.K. Baker, J. Phys. Chem. B 105 (2001)
11994.
[23] R.M. Rioux, M.A. Vannice, J. Catal. 216 (2003) 362.
[24] E.K. Poels, D.S. Brands, Appl. Catal. A 191 (2000) 83.
[25] H. Wilmer, M. Kurtz, V. Klementiev, O.P. Tkachenko, W. Grünert, O. Hinrichsen,
A. Birkner, S. Rabe, K. Merz, M. Driess, C. Wöll, M. Muhler, Phys. Chem. Chem.
Phys. 5 (2003) 4736.
[26] X. Huang, L. Ma, M.S. Wainwright, Appl. Catal. A 257 (2004) 235.
[27] G.J.J. Bartley, R. Burch, R.J. Chappell, Appl. Catal. 43 (1988) 91.
20
[28] A.M. Alekseev, A.V. Mikhailova, V.S. Beskov, G.S. Shitikova, B.N. Kuznetsov,
Khimicheskaya Promyshlennost (russ.) 2 (1995) 99.
[29] G.P. Vissokov, J. Mater. Sci. 28 (1993) 6457.
[30] V.V. Dergachev, A.V. Krylova, SU Patent 1,625,520 (1991).
[31] A.V. Krylova, N.V. Nefedova, L.D. Kuznetsov, P.D. Rabina, G.A. Ustimenko,
T.L. Koroleva, P.T. Mikhailovich, N.S. Torocheshnikov, SU Patent 1,250,319
(1986).
[32] S. Hammache, J.G. Goodwin, R. Oukaci, Catal. Today 71 (2002) 361.
[33] M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal. Today
100 (2005), 249.
[34] (a) S. Storsæter, Ø. Borg, E.A. Blekkan, A. Holmen, J. Catal. 231 (2005) 405; (b)
S. Storsæter, B. Tøtdal, J.C. Walmsley, B.S. Tanem, A. Holmen, J. Catal. 236
(2005) 139-152.
[35] A. Dandekar, R.T.K. Baker, M.A. Vannice, J. Catal. 183 (1999) 131.
[36] G.C. Chinchen, C.M. Hay, M.D. Vandervell, K.C. Waugh, J. Catal. 103 (1987) 79.
[37] G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G. Thomas, B.H.
Davis, Appl. Catal. A 258 (2004) 203.
[38] S. Sato, R. Takahashi, T. Sodesawa, K. Yuma, Y. Obata, J. Catal. 196 (2000) 195.
[39] J. Lemaître, P.G. Menon, F. Delannay, in: Characterization of heterogeneous
catalysts, ed. F. Delannay (Marcel Dekker, New York, 1984) pp. 299.
[40] V.R. Palkar, P. Ayyub, S. Chattopadhyay, M. Multani, Phys. Rev. B 53 (1996)
2167.
[41] T. Ressler, J. Synch. Rad. 5 (2004) 118.
[42] J.R. Jensen, T. Johannessen, H. Livbjerg, Appl. Catal. A 266 (2004) 117.
[43] B.S. Clausen, J.K. Nørskov, Top. Catal. 10 (2000) 221.
[44] I. Kvande, D. Chen, M. Rønning, H.J. Venvik, A. Holmen, Catal. Today 100
(2005) 391-395.
[45] J.W. Niemantsverdriet, J.A.C. van Kaam, C.F.J. Flipse, A.M. van der Kraan, J.
Catal. 96 (1985) 58.
[46] O. Levenspiel, Chemical Reaction Engineering, third ed. (John Wiley & Sons,
New York, 1999).
[47] U. Collisi, H.-H. Strehblow, J. Electroanal. Chem. 284 (1990) 385.
21
[48] K.M. Lee, W.Y. Lee, Catal. Lett. 83 (2002) 65.
[49] S.H. Overbury, D.R. Huntley, D.R. Mullins, G.N. Glavee, Catal. Lett. 51 (1998)
133.
[50] I.J. Kitchener, EP Patent 89,761 (1983).
[51] X. Dong, H.-B. Zhang, G.-D. Lin, Y.-Z. Yuan, K.R. Tsai, Catal. Lett. 85 (2003)
237.
[52] H.-B. Zhang, X. Dong, G.-D. Lin, X.-L. Liang, H.-Y. LI, Chem. Commun. 40
(2005) 5094.
[53] (a) J. P. Espinos, J. Morales, A. Barranco, A. Caballero, J.P. Holgado, A.R.
Gonzalez-Elipe, J. Phys. Chem. B 106 (2002) 6921; (b) J. Morales, J.P. Espinos,
A. Caballero, A.R. Gonzalez-Elipe, J.A. Mejias, J. Phys. Chem. B 109 (2005)
7758.
[54] X. Liu, W. Ruettinger, X. Xu, R. Farrauto, Appl. Catal. B 56 (2005) 69.
[55] Z. Yu, D. Chen, B. Tøtdal, A. Holmen, Catal. Today 100 (2005) 261.
[56] P.C. Hayes, Process principles in minerals and materials production, third ed.
(Hayes Publishing Co., Brisbane, 2003).
22
Figure captions
Figure 1. A typical TGA experimental curve, obtained over Cu-350, including removal
of water/carbonate species adsorbed on the surface, TPR of Cu and N2O decomposition
for determination of Cu dispersion and passivation.
Figure 2. XRD spectra of (a) reduced-(O2)passivated (D-5000 diffractometer and (b)
calcined (D-5005 diffractometer) Cu catalysts. The symbols refer to different crystalline
phases: █ = ZnO, ▲ = CuO, ▼ = Cu, ○ = CeO2
Figure 3. Weight increase during N2O titration (10 vol-% N2O/Ar, 75 ºC, flow rate 80
ml/min) normalized with the total weight loss upon reduction on Ce-bC-400 as a
function of square root of time. Pre-reduction was performed either with 7 vol-% H2 or
with CO in argon (260 ºC, 2 h, heating rate 2 K/min, flow rate 80 ml/min). The reduced
sample was cooled in the reduction gas prior to N2O titration.
Figure 4. XAS spectra of the Ni-Fe mixed oxide catalyst recorded at the Ni K-edge.
XANES profiles are shown for the calcined, reduced-(O2)passivated and used samples
together with the reference compounds NiO and Ni-foil.
Figure 5. XAS spectra of the Ni-Fe mixed oxide catalyst recorded at the Fe K-edge.
XANES profiles are shown for the calcined, reduced-(O2)passivated and the used
samples together with the reference compounds α-Fe2O3, Fe3O4 and Fe-foil.
Figure 6. Least-squares fit (dashed line) of the XANES spectra (solid line)of the
reduced-(O2)passivated Ni-containing sample at the Ni K-edge. The fit is produced by
linear combination of the XANES spectra of the model components NiO and Ni-foil
using a Levenberg-Margquardt least squares algorithm in the WINXAS software.
Figure 7. XRD spectra of Co/SiO2 for calcined and reduced-(O2)passivated samples.
The symbols refer to different crystalline phases: ▼ = Co3O4, ● = CoO, = Co. The
profile of the reduced-(N2O)passivated sample is identical to the spectra of the O2-
passivated sample and is therefore not shown in the figure.
23
Figure 1
41
42
43
44
45
46
47
48
49
50
0 100 200 300 400 500 600 700 800 900time [min]
wei
ght l
oss
[ m
g ]
0
50
100
150
200
250
300
tem
pera
ture
[ C
]
Cu-350programme temperature
water,carbonates
N2O decomposition
Cureduction
41
42
43
44
45
46
47
48
49
50
0 100 200 300 400 500 600 700 800 900time [min]
wei
ght l
oss
[ m
g ]
0
50
100
150
200
250
300
tem
pera
ture
[ C
]
Cu-350programme temperature
water,carbonates
water,carbonates
N2O decompositionN2O decomposition
Cureduction
Cureduction
temperature program
24
Figure 2
0
1000
2000
3000
4000
5000
6000
25 30 40 50 60 70
Cu-350
Cu-400
Ce-bC-400
(b)
sign
al [a
.u.]
2 Θ [ ° ]
0
1000
2000
3000
4000
30 40 50 60 70 80 90
Cu-350
Cu-400
Ce-bC-400
(a)
sign
al [a
.u.]
2 Θ [ ° ]
25
Figure 3
0
0,02
0,04
0,06
0,08
0 2 4 6 8
t^(1/2) [ min^(1/2) ]
norm
aliz
ed w
eigh
t inc
reas
e [ -
]
Ce-bC-400-H2
Ce-bC-400-CO
26
Figure 4
0.5
1.0
1.5
8.328.34
8.368.38
2.25
4.5
Y
X
Z
photon energy [ keV ]
NiO Calcined
Red-pass Used
Ni foil
norm
. abs
orpt
ion
[ a.u
.]
27
Figure 5
0.5
1.0
1.5
7.1
7.15
7.2 norm
. abs
orpt
ion
[ a.u
.]
photon energy [ keV ] Fe foil
Used Red-pass
Calcined α-Fe2O3
Fe3O4
28
Figure 6
0.5
1.0
8.32 8.34 8.36 8.38
norm
. abs
orpt
ion
[a.u
.]
photon energy [keV]
29
Figure 7
sign
al [a
.u.]
2 Θ [ ° ]
calcined reduced-(O2)passivated
30
Table 1. Chemical composition and physical properties of the copper catalyst samples.
Cu crystallite size
(with O2 passivation)
Sample H2O/CO2
weight lossa
Total Cu mass fraction
in the samples
[ - ]
Cu
dispersiond
Cu passivated
with N2Oe
Cu passivated
with O2e Cu core
with XRDf
Core + pass.layer
(XRD + TGA)g
[ wt.% ] ICP-AESb TGA-TPRc [ % ] [ % ] [ % ] [ nm ] [ nm ]
Cu-350 5.7 0.29 0.28 6.5 10 40 19 23
Cu-400 1.7 0.27 0.26 5.4 7 45 20 25
Ce-bC-400 1.7 0.25 0.25 8.2 11 40 20 24
a Thermogravimetric measurements performed in Ar on the calcined samples up to the reduction temperature (260 C) resulted in a weight loss assigned to
adsorbed water and/or surface carbonates. b Normalised mass fractions, i.e. only CuO, ZnO, Al2O3 and CeO2 taken into account, with an estimated detection limit of 0.01 – 0.03 mg/g.
The elementary analysis was performed on as-prepared calcined samples, thus containing water/carbonates adsorbed on the surface. c For better comparison with the ICP-AES results the amount of copper determined by TGA-TPR is normalized with the sample mass before the drying procedure. d Copper dispersion measured by means of selective oxidation via N2O surface titration of the reduced catalyst samples taking into account copper bulk oxidation
[38,42], with percentage values being based on mole fraction. e Normalized with amount of Cu determined by TGA-TPR prior to passivation, with percentage values being based on mole fraction. f Performed on the reduced and passivated samples. The copper crystallite size is estimated using the Scherrer equation for the Cu (111) reflection. g The thickness of the passivation layer is taken into account assuming a cubic particle shape.
31
Table 2. Phase composition of the Ni-Fe catalyst sample (passivated with O2),
determined by a linear combination of XANES data of the reference materials at the Ni
and the Fe K-edge applying a least-squares fitting algorithm in the software package
WINXAS.
Ni K-edge:
Sample Ni metal [ % ] NiO [ % ]
Calcined 7 93
Reduced-passivated 70 30
Used 98 2
Fe K-edge:
Sample Fe metal [ % ] α-Fe2O3 [ % ] Fe3O4 [ % ]
Calcined 0 65 35
Reduced-passivated 57 39 4
Used 82 10 8
32
Table 3. Thermodynamics of the metal oxidation by O2 and N2O, based on the
standard enthalpies of formation for the bulk metals.
a From standard enthalpies of formation: N2O: 82 kJ/mole, Cu2O: -168 kJ/mole, NiO: -236 kJ/mole,
Fe2O3: -815 kJ/mole, Fe3O4: -1103 kJ/mole, CoO: -234 kJ/mole, Co3O4: -885 kJ/mole. Per definition,
the standard enthalpy of O2 is zero. The enthalpy of the metals is set to zero as well, although small
enthalpies of formation (absolute values < 7 kJ/mol) are tabulated for some metallic phases of Ni, Fe
and Co ([56] and references therein).
Standard reaction enthalpy [ kJ/mole ]a
N2O (per mole N2O) O2 (per mole O2)
Cu → Cu2O -250 -337
Ni → NiO -318 -472
Fe → Fe2O3 -354 -543
Fe → Fe3O4 -358 -408
Co → CoO -316 -468
Co → Co3O4 -303 -442
Paper III
Preparation and characterization of nanocrystalline, high-surface area
Cu-Ce-Zr mixed oxide catalysts from homogeneous
co-precipitation
Manuscript in preparation.
1
Preparation and characterization of nanocrystalline, high-surface area Cu-Ce-Zr
mixed oxide catalysts from homogeneous co-precipitation
Florian Huber1, Hilde Venvik1,*, Magnus Rønning1, John Walmsley2, Anders Holmen1
1Department of Chemical Engineering, Norwegian University of Science and
Technology (NTNU), N-7491 Trondheim, Norway 2SINTEF Materials and Chemistry, N-7465 Trondheim, Norway
* Corresponding author; E-mail: [email protected], Tel: +47-73592831,
Fax: +47-73595047
Cu0.23Ce0.54Zr0.23-mixed oxides were prepared by homogeneous co-precipitation with
urea. The resulting materials exhibit high surface area and nanocrystalline primary
particles. The material consists of a single fluorite-type phase according to XRD and
TEM. STEM-EDS analysis shows that Cu and Zr are inhomogeneously distributed
throughout the ceria matrix. EXAFS analysis indicates the existence of CuO-type
clusters inside the ceria-zirconia matrix. This type of mixed oxide materials should
therefore be described as heterogeneous single-phase materials rather than
homogeneous solid solutions. The pore structure and surface area of the mixed oxides
are affected by preparation parameters during both precipitation (stirring) and the
following heat treatment (drying and calcination). TPR measurements show that most of
the copper is reducible and not inaccessibly incorporated into the bulk structure.
Reduction-oxidation cycling shows that the reducibility improves from the first to the
second reduction cycle, probably due to a local phase segregation in the metastable
mixed oxide with gradual local copper enrichment during heat treatment.
KEY WORDS: homogeneous alkalinization; urea hydrolysis; solid solution; Cu-Ce-Zr
mixed metal oxide; reducibility; drying; surface-to-volume ratio.
2
1. Introduction
1.1. Cu-Ce-Zr mixed metal oxides (MMO)
Mixed metal oxides (MMO) containing copper, cerium and zirconium are applied in
several areas of heterogeneous catalysis. Ce-Zr mixed oxides are extensively used in
three-way catalysts [1,2]. Cu-Ce-Zr-based mixed oxides are applied within the field of
hydrogen production: Water-gas shift [3-6], steam reforming of methanol [7-10] and
selective oxidation of CO [6,11-17]. In addition, they are used as NO reduction catalysts
[18], for oxidation of methane [19,20], wet oxidation of phenol [21] and acetic acid [22],
methanol synthesis [23,24], direct oxidation of hydrocarbons in solid-oxide fuel cells
(SOFC) [25-27], and storage of reactive hydrogen for alkadiene hydrogenation [28].
Copper - in its reduced state - is typically regarded as the active catalyst component in
MMO materials (except for SOFCs, where copper the main function is electronic
conductivity [25]). Ceria acts as a reducible oxide support, enhancing the catalytic
activity via metal-support interaction and/or improved dispersion of the active metal
component [8,29]. An important property of ceria is the oxygen storage capacity (OSC),
i.e. the ability to adsorb and release oxygen under oxidizing and reducing conditions,
respectively, according to the reaction [30]:
22 2
2 2
. ( / ). ( / ) x
red H COCeO CeO
ox H O CO − (1)
In addition, ceria is found to stabilize the catalyst against deactivation [8,29] due to a
higher thermal stability and/or better dispersion of the active metal. ZrO2 is also known
to improve the activity and stability of MMO-based catalysts [9,31]. Zirconium added to
ceria to form Ce-Zr mixed oxides inhibits the thermal sintering of CeO2 [2,30,32,33].
Incorporation of Zr into the ceria lattice enhances the reducibility of ceria [34-36],
which may improve the catalytic activity of MMO catalysts relative to single oxides
[37,38]. The amount of Zr-dopant also affects surface area and crystallite size of the
MMO [37,39], in conjunction with the calcination temperature [40]. In addition to its
presumed function as an active catalytic species, Cu as a dopant in the fluorite lattice is
3
found to improve the reducibility of ceria [41,42], and affect particle size and surface
area [24,43].
The activity and stability of MMO catalysts therefore depends on the interaction
between the single components, for which a homogeneous distribution of the
components throughout the material without pronounced segregation is a pre-requisite.
The catalytic activity often scales proportionally with the surface area of the active
components [4,44] since the reaction takes place at the surface of the catalyst. A
homogeneous distribution of the components and a high surface area of the material
have to be assessed simultaneously during catalyst preparation.
Co-precipitation of metal (hydrous) oxides in aqueous solution at high pH has
successfully been applied for preparing different metal oxide catalyst formulations. It
has been stated that co-precipitation results in more active and stable catalysts than
impregnation methods [17,20,45-47], because of a more homogeneous distribution of
the elements. For impregnation-deposition methods, the interaction between the
different components depends on the surface area of the support material and the
amount of material impregnated or deposited. The existence of separate phases at higher
loadings is likely [4].
1.2. Homogeneous co-precipitation with urea (HCP)
Homogeneous alkalinization via urea hydrolysis is an efficient co-precipitation
procedure for preparation of MMO with high surface area and well-defined particle size
and shape [3,19,48-51]. Urea decomposes at elevated temperatures in a two-step
reaction releasing ammonium and carbonate ions into the metal salt solution
accompanied by a simultaneous increase in pH, which leads to the precipitation of basic
carbonates [48,50,52]:
22 2 2 4 3( ) 2 2CO NH H O NH CO+ −+ → + (2)
The decomposition rate strongly depends on the temperature [52], the rate constant
increasing by a factor of about 200 as the temperature increases from 60 to 100 °C
4
[50,52]. The kinetics of the metal ion hydrolysis, and hence the nucleation rate, can be
tuned through controlled release of hydroxide ions to obtain well-defined particle
shapes with a narrow particle size distribution [48-50]. Constant-pH co-precipitation
procedures with other hydroxide ion sources, such as NaOH, are claimed to facilitate
stronger agglomeration of primary particles into irregularly shaped clusters, resulting in
a broader particle size distribution [50,51].
In terms of synthesizing MMO solid solutions, co-precipitation in general has to be
regarded as a heterogeneous process. Because of different hydrolytic properties of the
metal ions in aqueous solution, simultaneous nucleation is rarely the case. A phase
containing only one cation usually nucleates to serve as site for the heterogeneous
nucleation of a second solid. Further growth proceeds incorporating both cations at
different rates [49]. As a result, the internal local composition of such composites
usually varies from the center to the periphery of each particle [48]. The detection of a
homogeneous solid solution inside the MMO material depends on the characterization
technique used. The detection of single-phase MMO by conventional XRD techniques
does not exclude the presence of nanodomains of single-component-rich phases [2],
since XRD averages properties over a macroscopic scale as well as being insensitive to
amorphous phases and ordered structures below about 2 nm [53,54].
1.3. Effect of synthesis parameters
Synthesis parameters that may affect the resulting catalyst material properties such as
surface area, particle size and morphology are [21,48,50]: Type of precursor salt,
organic additives (from simple organic solvents to surfactants), total metal
concentration, metal ratio, urea concentration, procedure of urea addition (at ambient or
high temperature), pH, procedure of mixing, synthesis temperature, aging time and post-
treatment (drying and calcination conditions).
Nitrate precursor salts can easily be decomposed (an exothermic reaction), are
inexpensive and have high solubility in aqueous media. Sulphates and chlorides are
usually excluded since their residue in the catalyst material might accelerate catalyst
5
deactivation [57,58]. The type of anion present in the solution can, however, have an
impact on the precipitated particles [48].
The urea concentration (and initial ratio of urea to total metal in the solution), synthesis
temperature and initial metal ratio in the solution can affect the pH evolution rate, and
hence the nucleation and growth. Small particles with a narrow particle size distribution
has be obtained for high urea-total metal ratio and high reaction temperature [48,50].
Total metal concentration and acidic/basic additives affect pH and hence actual metal
contents and phases present in the final product [48,49]. A high total metal
concentration in the solution may result in enhanced particle agglomeration that reduces
the surface area [48]. The aging time can affect crystallinity, particle size [50] and phase
changes [49,59] and influence the final metal content [48,49]. All the parameters
mentioned in this paragraph affect the nucleation and growth of mixed metal (hydrous)
oxides/carbonates at different stages in aqueous solution, and may be highly correlated.
The effect of calcination temperature is well established for ceria-based systems. As for
many systems, the particle size generally increases with a corresponding decrease in the
surface area with increasing calcination temperature. As a consequence, the oxygen
storage capacity of ceria and the catalytic activity of the material decrease
[4,21,33,40,43,60-63].
There appears to be an optimum range for the metal loading of ceria-based mixed
oxides, irrespective of preparation method [20]. The best redox properties and highest
OSC for Ce-Zr solid solutions have been obtained within the range 0.2 ≤ x ≤ 0.4, where
x is the atomic fraction of Ce atoms replaced by Zr relative to pure CeO2 [1]. A
maximum activity for water-gas shift [38] and for CO2 reforming of methane [64] on
Ce-Zr supported Pt catalysts was found with x = 0.5. A maximum methanol
decomposition activity was reached at x = 0.3 for a Pd-Ce-Zr system [37]. For the same
components, a maximum at x = 0.2 was found for CO and C3H8 oxidation [40].
A roughly linear relationship between the rate constant and Cu content up to 20 at% was
found for Cu-Ce mixed oxide catalysts for wet oxidation of phenol [21]. During
6
selective CO oxidation over Cu-Ce catalysts, the highest reaction rate was obtained for a
Cu-content of 14 at% as compared to 7 and 21 at% [15]. This catalyst was also the one
exhibiting the highest surface area. For the same reaction, catalyst system and
preparation method, albeit higher surface areas, Kim et al. [6] obtained the highest
activity at 20 at% Cu as compared to 10 and 50 at%, but state that the difference
between the three catalyst formulations is not large. Tang et al. [17] reported the CO
conversion at low temperatures to be higher for 24 at% Cu than for 7 and 13 at% Cu in
co-precipitated catalysts. In methanol steam reforming over Cu-Ce catalysts, the highest
methanol conversion was observed for about 50 at% Cu [46]. The maxium TOF was
obtained for about 10 at%. Shen et al. [24] found the initial maximum space-time yield
in the methanol synthesis for 47 at% Cu, but after about 25 h on stream the maximum
shifted to the Cu-Ce catalyst with 23 at% Cu. The difference between the 12, 23 and 47
at% catalysts decreased with time on stream. For oxidation of CO and CH4 over
Cu/ZrO2, the catalyst containing 20 at% Cu in ZrO2 was found to be most active [65].
Kundakovic et al. [19] achieved higher methane oxidation reaction rates for 15 at% than
for 5 at% Cu in a Cu-Ce-La catalyst containing 4.5 at% La. For the water-gas shift
reaction, the highest CO conversion was measured for 15 – 20 at% Cu in Cu-Ce-La
catalysts containing 10 at% La, but the difference is not large between catalysts with 5
at% to 40 at% under the reaction conditions used [3]. In a more recent study of the same
system [5], a 10 at% Cu catalyst was found to perform slightly better than 5 and 15 at%
catalysts, and all were better than a 40 at% catalyst. All these catalysts contained 8 at%
La in ceria, and the reaction conditions were similar in the two studies. Investigating the
impact of the third metal component, they observed increased activity in the order 30
at% La > 24 at% Zr > 8 at% La, and conclude that this effect is of chemical origin and
not scalable with the surface area [5]. An improvement in the long-term stability and
suppressed CO formation during methanol steam reforming over Cu-Ce-Zr catalysts by
increasing the amount of Cu from 4 to 12 at% has also been reported [66]. No
significant enhancement was observed above 12 at%, and this could be related to the
detection of a separate CuO phase for samples with higher loading. The molar
composition (Cu:Ce:Zr) of the 4 at% and 12 at% catalyst were 4.4:51.0:44.6 and
12.1:44.3:43.6, respectively. Usachev et al. [14] achieved the best selectivity for
selective oxidation of CO in excess hydrogen on Cu-Ce-based catalysts for Cu:Ce:Zr =
7
0.23:0.54:0.23. Interestingly, an optimum composition for Co-Ce-Zr catalysts for
hydrogen production by ethanol steam reforming was found at Co:Ce:Zr =
0.225:0.500:0.275 [67].
Qualitative conclusions that can be drawn from the brief literature review above with
regard the Cu-Ce-Zr composition are:
(1) A Ce:Zr ratio between 3:1 and 2:1 appears to be a reasonable choice.
(2) The optimum amount of copper incorporated into the ceria structure lies in the
range 10 - 20 at %, probably limited by the amount of Cu that can be dispersed
in ceria (or zirconia) without forming separate CuO phases.
(3) In 3-component-mixtures, good performance is achieved with 40 - 50 % of the
Ce-atoms replaced by Cu and Zr in equimolar amounts, i.e. an atomic ratio
Cu:Ce:Zr ≈ 1:2:1.
Some of the studies suggest that the catalytic properties of these MMO are insensitive to
variations in the Cu content within a range of 10 – 25 at%. In a study dealing with the
preparation method, an experimental optimization of the composition may thus not be
first priority, and reasonable values taken from literature should be adequate. An
optimization of the metal composition has to be established for the specific application
under relevant reaction conditions.
When using urea for precipitation of copper salts, the ammonia complexes (NH4+/NH3)
released during urea decomposition (equation (2)) interact with copper ions to form the
deep blue copper-ammonia complex cations, [Cu(NH3)4]2+ or [Cu(NH3)6]2+, depending
on the ammonia concentration. These complexes retain the copper ions in solution and
decrease the amount of Cu in the final mixed metal precipitate. In a closed system
(reflux conditions), an equilibrium will be established between precipitated copper
hydroxide-carbonate phases and complex copper ions in solution. In an open system,
this equilibrium can be shifted in favour of copper precipitation by releasing ammonia
into the gas phase at elevated temperatures.
8
1.4. Description of the surface area of MMO
When characterizing the surface area of a certain material, the interesting parameter is
the surface-to-volume ratio (stv, m2/m3), i.e. the fraction of the material exposed to the
surrounding gas. The BET-deduced surface area (m2/g) is useful for comparison of
solids of equal composition. For materials with varying composition, however, the
effect of the density has to be taken into account [68]. If the structure remains more or
less unaffected by varying composition, the molar mass can simply be used to eliminate
the mass effect, resulting in a mole-based BET value (m2/mole). The BET results
obtained by Kapoor et al. [37], for mesostructured Ce-Zr mixed oxides, and Hirano et al.
[39], for non-ordered microporous Ce-Zr mixed oxides, are useful for exemplifying the
effect of the metal composition on the surface area (Figure 1). The stv (m2/m3) curves
show the true impact of varying metal composition on the surface area of the materials.
The densities used to calculate the stv from the BET data are based on a linear
combination of tabulated values. The use of measured densities in such evaluations
would further increase the precision. Comparing the surface-composition dependence
for the specific BET (m2/g) and the mole-based BET (m2/mole) indicates the impact of
the change in molar mass (with varying metal composition) on the BET surface-
composition dependence. Comparing the surface-composition dependence for mole-
based BET and stv shows the impact of the change in lattice spacing (with varying
metal composition) on the mole-based BET surface-composition dependence. In this
simplified re-evaluation, the effect of a change in lattice spacing and hence the volume
of the crystal unit cell by varying the metal composition affects the BET surface area
only to a minor degree. A considerable contribution to the change in specific BET
surface area with variation of the metal composition arises from the different molar
mass of cerium and zirconium.
The effect of copper loading on the surface area of Cu-Ce mixed oxides can be
evalutated in a similar way, for example by using data from references [24] and [43].
Here, however, the surface area decreases with increasing Cu loading after an initial
strong increase at very low Cu level. These curves (not shown) contain local minima
and maxima, and the overall trend depends on the resolution of the experimental points.
The same might be true for the Ce-Zr system. The BET data obtained for a series of
9
coprecipitated Cu-Zr mixed oxides can be used in a similar way to evaluate these binary
MMO [69].
1.5. Aim of the study
The aim of the present study is to determine relevant material properties of Cu-Ce-Zr
mixed oxides prepared by homogeneous co-precipitation. This includes the actual
catalyst composition relative to the nominal composition, as well as particle
morphology, surface area, pore structure, metal distribution or degree of homogeneity,
local atomic structure, and finally redox behaviour.
The initial concentration of the metal salts in aqueous solution (= nominal MMO
composition) is kept constant. The nominal molar metal ratio used is: Cu:Ce:Zr =
0.23:0.54:0.23. This ratio is based on the literature, as described above. The parameters
studied, in terms of their effect on surface area/particle size and pore structure, are:
Synthesis setup with (a) two types of stirring/heat transfer configurations and (b) reflux
conditions vs. open system, drying conditions and heating rate during calcination.
An elementary analysis of the actual catalyst composition is carried out with Inductively
Coupled Plasma Atomic Emission Spectroscopy (ICP-AES). Morphology and particle
shape are examined with Transmission Electron Microscopy (TEM). The particle size of
the MMO is estimated with X-ray Diffraction (XRD) and TEM. The homogeneity of the
MMO is investigated using XRD (phase identification) and Scanning-TEM (STEM,
mapping of metal composition on a nm-scale). X-ray Absorption Spectroscopy (XAS)
has been applied for eludication of the oxidation state of bulk copper and its local
environment (first coordination shell). The redox behaviour is characterized by means
of Temperature Programmed Reduction and Oxidation (TPR-TPO) cycles. The pore
structure of the MMO is deduced from nitrogen adsorption-desorption isotherms. The
specific surface area is estimated by the Brunauer-Emmet-Teller (BET) method.
10
2. Experimental
2.1. Synthesis
Copper(II)-nitrate-trihydrate (> 99 %), cerium(III)-nitrate-hexahydrate (> 99.5 %) and
zirconyl(IV)-nitrate-hydrate (> 99.5 %) were purchased from Acros Organics. Urea (>
99.5 %) was purchased from Merck. Ethylene glycol (EG) (> 99.5 %) was purchased
from Fluka. Ethanol was purchased from Arcus. All chemicals were used as-received.
Deionized water was used for all catalyst preparations.
The two different setups used for catalyst preparation are shown in Figure 2. The first
setup consists of a hot plate equipped with a magnetic stirrer(Figure 2A). The precursor
solution is stirred in a 600-ml glass beaker with a magnetic stirring bar rotating at
approx. 1200 rpm. The temperature of the mixture is monitored with a thermometer
immersed in the solution. This setup is operated in an open mode, i.e. water evaporating
during operation is compensated by continuous refilling. The glass beaker is partly
covered to limit the evaporation. The second setup consists of a 500-ml 5-neck glass
flask immersed in a stirred oil bath (Figure 2B). The temperature is monitored with an
immersed thermometer, and the aqueous mixture is stirred with a blade agitator at
approx. 750 rpm. In the closed mode, a reflux condenser is connected to the flask, and
the unused flask necks are plugged. In the open mode, the three unused flask necks are
left open to allow the evaporation as well as the continuous refilling of water.
A total amount of 0.1047 mole of Cu-, Ce- and Zr-salt (nominal composition: Cu:Ce:Zr
= 0.23:0.54:0.23) were dissolved in 450 ml of water at ambient temperature. 0.9037
mole of urea were then added to the solution. After all solid components were dissolved
to give a pH of approx. 1.8 at 20 °C, the light blue, transparent solution was placed on
the hot plate/in the hot oil bath and heated up to 95 °C (unless otherwise noted) under
stirring. It took about 30 min. to reach the final temperature for both setups. The mixture
was kept on the hot plate/in the oil bath for a period of 8 hours, including precipitation
and aging. The precipitation started after approx. 30 min in both setups. The colour of
the suspension then changed from light-blue to green over the next 45 mins. and then
remained unchanged for the rest of the time. In the open mode, water at ambient
temperature was added continuously to the suspension to compensate the evaporation.
11
Ethanol was added instead of water in a single experiment, and in total 1,5 l of ethanol
was added during synthesis.
The suspension was removed from the hot plate/oil bath after 8 hours, and cooled to
room temperature using water. The solid precipitate was filtered off and washed twice
with 200 ml of deionized water (40 – 50 °C) under stirring (20 – 30 min.). Finally, the
precipitate was dried (about 11 h at 100 °C, unless otherwise noted) and calcined in a
muffle furnace for 2 h at 350 °C, either by placing the sample directly into the hot
furnace (normal method) or by increasing the calcination temperature at 2 °C/min up to
350 °C.
In two preparations, an ethylene glycol-water mixture (18 mole% or 40 V% EG) was
used instead of pure water, thereby increasing the boiling point of the mixture by about
5 °C. The synthesis temperatures were 95 °C and 100 °C. The resulting samples were
used for studying the effect of the drying conditions. The pore size distribution of the
sample synthezised at 100 °C was different from the typical distribution of the other
samples. For more information on this matter, we refer to Adachi-Pagano et al. [50].
The sample notation takes into account deviations from the typical synthesis conditions
given above and is classified as follows: HP = hot plate, B = oil bath, c = closed system,
o = open system, EG = ethylene glycol–water mixture, Eth = ethanol refilling, s98 =
synthesis temperature raised to 98 °C, d150 = drying temperature 150 °C, dry = dried,
not calcined, 2C = calcination rate 2 °C/min, N = drying in nitrogen, redox = after
reduction and reoxidation.
2.2. Characterization
Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) was used to
determine the actual amount of copper, cerium and zirconium after calcination of the
precipitated materials. The samples were dissolved in hydrochloric acid before analysis
without any visible residues.
12
Nitrogen adsorption-desorption isotherms were measured using a Micrometrics
TriStar 3000 instrument. The data were collected at -196 °C. The BET surface area
was calculated by the BET equation in the relative pressure interval ranging from 0.01
to 0.30. The pore volume was estimated by the Barrett-Joyner-Halenda (BJH) method
[70] as the adsorption cumulative volume of pores between 1.7 nm and 300.0 nm width.
This method is based on the assumption of cylindrical pores, and the capillary
condensation in the pores is taken into account by the classical Kelvin equation. The
pore size distributions were calculated by non-local density functional theory (NLDFT,
Original DFT model with N2 [71], DFT Plus software package [72]) assuming a slit-like
pore geometry. For comparison, pore size distributions were also determined with the
Harkins-Jura model (cylindrical geometry), a classical method based on the Kelvin
equation [73]. Although the models gave different absolute values for the total pore
volume, the trends were essentially the same for the samples investigated. For more
details on the comparison of DFT and classical models based on the Kelvin equation,
we refer to Rouquerol et al. [74] and Chytil et al. [75], and the references cited therein.
XRD data were recorded on a Siemens diffractometer D-5005 (dichromatic CuKα+β-
radiation). Average crystal size estimates and crystal size distributions for the MMO
powders were obtained from a software-based X-ray line broadening analysis (XLBA).
The analysis was performed in two steps. Selected experimental XRD fluorite peaks
((111), (200) and (220)) were simulated by means of the software package Profile [76]
using the Pearson VII model function. The contribution of CuKβ to the peak intensities
is removed in this step. The program Win-crysize [77], utilizing the Warren-Averbach
method [78], was then used to estimate the crystallite size taking into account
contributions from microstrain (scaled as mean square root of the average squared
relative strain). Contributions from instrumental line broadening were removed using
LaB6 as reference (selected peaks: (110) at 30.38°, (111) at 37.44° and (210) at 48.96°).
Thermogravimetric Analysis (TGA) was performed with a Thermogravimetric
Analyser (Perkin Elmer TGA 7). The device was used to check the metal oxide content
in the nitrate precursors, the weight loss of the calcined samples assigned to
water/carbonates adsorbed on the oxide powders, and small residues of precursor
13
species not removed after calcination at 350 °C (approx. 1 wt% at temperatures up to
500 °C). TGA was also used for temperature-programmed reduction and reoxidation
(redox). The redox procedure was conducted as follows using crushed oxide powders:
(1) pre-oxidation in air up to 500 °C (heating rate: 10 °C/min, dwell: 10 min) in order to
remove water/carbonates and precursor residues; (2) cooling down to ambient
temperature in air, flushing and stabilization of the baseline, (3) reduction in 7 vol-% H2
in nitrogen up to 300 °C (5 °C/min); (3) cooling down in H2/N2, flushing and
stabilization; (4) reoxidation in air up to 350 °C (5 °C/min, 1 h); (5) cooling down in air,
flushing and stabilization; (6) reduction in 7 vol-% H2 in nitrogen up to 300 °C (5
°C/min). The gas flow rate was 80 Nml/min in all steps.
Transmission XAS data were collected at the Swiss-Norwegian Beam Lines (SNBL) at
the European Synchrotron Radiation Facility (ESRF), France. Spectra were obtained at
the Cu K-edge (8.979 keV) using a channel-cut Si(111) monochromator. Higher order
harmonics were rejected by means of a chromium-coated mirror aligned with respect to
the beam to give a cut-off energy of approximately 15 keV. The maximum resolution
(∆E/E) of the Si(111) bandpass is 1.4 x 10-4 using a beam of size 0.6 x 7.2 mm. Ion
chamber detectors with their gases at ambient temperature and pressure were used for
measuring the intensities of the incident (I0) and transmitted (It) X-rays. The catalyst
powder sample was diluted with boronitride. A copper foil, and Cu2O and CuO powders
diluted with boron nitride were used as reference materials. The recorded XAS spectra
were energy calibrated, pre-edge background subtracted (linear fit) and normalised
using the WinXAS software package [79]. For Extended X-ray Absorption Fine
Structure (EXAFS) analysis the data were converted to k-space using WinXAS, and the
least-square curve fitting was performed with the EXCURVE 98 program [80] based on
small atom approximation and ab initio phase shifts calculated by the program.
TEM data were recorded on a JEOL 2010F transmission electron microscope. Small
amounts of the catalyst samples were put into sealed glass containers containing ethanol
and placed in an ultrasonic bath for a couple of minutes to disperse the individual
particles. The resulting suspension was dropped onto a holey carbon film, supported on
a titanium mesh grid, and dried. Conventional TEM images were recorded onto a CCD
14
camera. Samples were also examined in scanning transmission electron microscope
(STEM) mode, with a nominal probe size of approx. 0.7nm. Bright field and dark field
STEM images were acquired. Energy dispersive x-ray spectroscopy (EDS) analysis and
mapping were performed using an Oxford Instruments INCA system. Drift
compensation was employed to correct for movement of the sample during the time
taken for the acquisition of maps.
3. Results and discussion
3.1. General features of the synthesized Cu-Ce-Zr mixed oxides
The colour change from light-blue via dark blue/green to green observed during the first
45 min of precipitation is an indication of the heterogeneous nature of the co-
precipitation of different metal ions. The pH at which the precipitation of the single
metals begins depends on temperature, concentrations and the anions present (see e.g.
the titration experiment performed by Lamonier et al. [43] for Cu-Ce mixtures). With
urea as the source for hydroxide ions, the formation of carbonate precipitates, which
depends on the solubility of the corresponding metal carbonates in aqueous solution as
well as the interaction of Cu cations with NH3, has to be taken into account. This holds
especially for rare-earth metals whose carbonates are insoluble in water [81]. In addition,
the metal ions may affect each other in respect of their precipitation behaviour. A
detailed experimental study, similar to the one performed by Soler-Illia et al. [49] on
Cu-Zn basic carbonates, would be necessary to elucidate the reaction path of the
nucleation and growth process in the temperature-dependent phase diagram of the
Cu/Ce/Zr-nitrate-urea system.
The total (not optimized) yield of precipitation was approximately 80 wt% based on the
nominal oxide composition, including losses due to the post-synthesis treatment.
3.1.1. TEM
TEM images of a calcined Cu-Ce-Zr mixed oxide sample are shown in Figure 3. Co-
precipitation of copper, cerium and zirconium ions results in the formation of
nanocrystalline Cu-Ce-Zr mixed oxide particles. Primary particles formed through
nucleation and growth have agglomerated to larger clusters (Figure 3A), perhaps
15
already during the aging of the precipitate in solution. According to Matijevic [48], the
degree of particle agglomeration in solution increases with initial metal concentration.
By using lower concentrations of the nitrate precursors, the degree of agglomeration
could be reduced and the total surface area of the powder possibly improved. However,
particle agglomeration will also occur during drying and calcination. Zhang et al. [82]
suggested the addition of an anionic surfactant before drying in order to reduce the
degree of agglomeration during drying and calcination.
Figure 3B displays the crystalline nature of the nanoparticles. A rough particle size
estimate lies in the range 3-5 nm. The lattice spacing is visible, and nanocrystalline
particles are randomly oriented. The d-spacing mainly observed in the individual
particles is approximately 3.1 Å, corresponds to the (111) spacing of the fluorite
structure for the CeO2 and Ce-Zr mixed oxides. The lattice constant is thus estimated to
5.4 Å, in accordance with literature data for Ce-Zr mixed oxides [2,83]. The image of
Figure 3B is obtained after reduction and reoxidation of the calcined sample in Figure
3A. Comparable images (not shown here) were obtained for the calcined sample,
displaying similar features. The fluorite-type structure is therefore preserved upon redox
treatment.
3.1.2. XRD
Representative XRD spectra of the Cu-Ce-Zr mixed oxides after different steps in the
synthesis are included in Figure 4. Figure 4a is recorded after drying at 100 °C. The
characteristic peaks of the fluorite-type structure show that this structure is present
already after drying. This is in agreement with Hirano et al. [39], who observed the
fluorite structure in as-precipitated Ce-Zr (hydrous) oxide samples after drying at 60 °C.
Lamonier et al. [43] made a corresponding observation for Cu-Ce mixed oxides dried at
90 °C. No significant change in the fluorite structure can be observed upon calcination
(Figure 4c) and redox treatment (Figure 4d), in accordance with the TEM images.
Figure 4b will be discussed in section Particle size estimation from XRD data (Table 1,
Bo-samples) shows that the particles present after drying of the precipitate are preserved
in size after calcination and redox treatment, with a slight tendency towards sintering.
Figure 5 shows the crystal size distribution and the microstrain as a function of crystal
16
size for sample Bo as obtained by XLBA. The average crystal size for this sample is
approx. 3 nm (Figure 5a, Table 1). The primary crystal XLBA size estimate reaches up
to 9 nm, with 90 % of the particles being smaller than about 6 nm (Figure 5b). The
XLBA and TEM result are thus in good agreement. 10 % of the particles with size ≤ 1
nm may, however, be questionable. This is further discussed in chapter 3.1.4.
3.1.3. XAS
The XANES profile of a typical calcined Cu-Ce-Zr mixed oxide (sample Bo-d120) at
the Cu K-egde is compared with the profile of reference materials: Cu foil, CuO and
Cu2O powder in Figure 6. The XANES fingerprint of Cu in the MMO resembles that of
CuO, indicating a major fraction of the Cu atoms to be in the (2+)-oxidation state.
A standard EXAFS analysis (i.e. k ≥ 3 [54,84]) was performed on the XAS spectra of
the same sample. The structural model is limited to the first oxygen shell around the
copper atoms. The fitting was carried out in different ways to check whether the
structural parameters obtained are sensitive to the fitting procedure: The k-intervals
used for fitting were [3;8], [3;9] and [3;12]. For the range k = [3;9], the analysis
includes both fitting with k1- and k3-weighting in order to account for the high
correlation between the coordination number N and the Debye-Waller factor ∆σ2 [84].
In Table 2, the structural parameters obtained by fitting the structural model for the first
oxygen shell to the experimental data are shown (R = distance of the oxygen shell to the
copper central atom, E0 = energy shift to correct for deviations from the theoretical edge
value). The initial values for the fitting were taken from the crystallographic data for
CuO (E0 = 0, ∆σ2 = 0.01). The variance takes into account the uncertainty given by
EXCURVE as well as the variation of different fitting approaches. Figure 7A shows the
experimental and theoretical EXAFS spectra in k-space with k = [3;9] and k3-weighting.
The Fourier transform profile indicates that the first oxygen shell is the dominant shell
in the experimental EXAFS spectra (Figure 7B). This is consistent with spectra reported
by Shen et al. [24] for Cu-Ce mixed oxides. It is thus reasonable to limit the fitting to
the first shell. No significant improvement was obtained by inclusion of further shells.
17
Table 2 also includes the structural parameters obtained for the CuO reference powder.
Here, the coordination number was not included in the fitting procedure (initial value: 4),
but was simply refined at the end of the fitting. The structural parameters for the first
oxygen shell around the Cu atoms in the mixed oxide are similar to the ones in CuO, in
accordance with the XANES comparison. It can therefore be concluded, that the major
part of the Cu atoms in the MMO exhibit the typical chemical and structural
environment of Cu in copper-(2+) oxide. Ramaswamy et al. [85] and Lamonier et al.
[43] identified different Cu species in electron paramagnetic resonance (EPR) studies on
sol-gel prepared, nanocrystalline Cu-Zr oxides and co-precipitated Cu-Ce oxides,
respectively. They distinguished framework-substituted Cu ions and extra-lattice
species, such as interstitial copper ions or copper dimers, dispersed ions bound to the
surface and CuO-type clusters or small particles. The ratio between these species was
reported to depend on the amount of Cu in the samples and the temperature pre-
treatment. Crystalline CuO, which is not visible to EPR but is to XRD [43], was
reported to form at high Cu loadings. Consistent with our XAS results, Lamonier et al.
[43] conclude from their EPR studies on coprecipitated Cu-Ce mixed oxides calcined at
400 °C, that a major part of copper exists as CuO-like clusters well dispersed in the
solid matix.
The reduced coordination number (N1) of oxygen atoms compared to CuO (Table 2)
may relate to the small CuO cluster size, with a significant number of low-coordinated
Cu atoms at the particle surface [53-56]. Neglecting other possible effects [54], a rough
estimate for the size of spherical CuO clusters would be 8 Å (with error deviation: 5-15
Å) [55], i.e. in the range of the STEM resolution. However, defects in the oxide
structure, such as oxygen vacancies, can not be excluded. Standard EXAFS analysis
tends to underestimate coordination numbers, especially for small particles [54].
Applying the correction function developed by Clausen & Nørskov [54] in a modified
form (N-bulk = 4 instead of 12) to our system, the coordination number does not
increase significantly. This is only a rough indication, since the correction was
developed for metallic copper with fcc structure and not for copper oxide. A multi-data-
set EXAFS analysis [56] including asymmetric pair distribution functions for nano-
sized particle clusters [54], could further minimize correlation effects.
18
3.1.4. STEM-EDS
As a result of the nucleation and growth process, the synthesized material should exhibit
gradient in composition at the nanoscale. Figure 8 shows the STEM-EDS elemental
mapping of sample Bo-redox. Cu, Ce and Zr are distributed over the whole mapping
area (approx. 30 x 50 nm2), and the average metal composition is close to the ICP-AES
measured composition (Table 3). Within the mapping area, regions of a few nanometers
high in Cu or Zr concentration relative to the surroundings could be observed (Figure
8C, E). This is in agreement with the XAS results (chapter 3.1.3) that suggested the
existence of CuO-type clusters. As mentioned above, initially formed nuclei probably
serve as sites during co-precipitation for the heterogeneous nucleation of a second phase,
which grows incorporating different cations at different rates [49]. Thus, the particles
should incorporate the three metals in an inhomogeneous way, although governed by
their hydrolytic behaviour. Ce and Zr do not necessarily precipitate simultaneously or at
the same rate under the conditions used here (in contrast to the findings in reference
[39]) and/or the formation of single element clusters might be more favourable than a
homogeneous distribution of all components.
A similar elemental mapping was conducted for sample Bo, i.e. the same material,
calcined but before any redox treatment. Inhomogeneous metal distributions at the local
scale were found similar to those shown in Figure 8. Therefore, the redox treatment
itself does not significantly change the metal distribution. Qi et al. [5] found by XPS
analyses of Cu-Ce-La mixed oxides that the amount of surface Cu increased after water-
gas shift reaction as compared to the freshly calcined material. Lamonier et al. [43]
observed enrichment of Cu at the surface of Cu-Ce mixed oxide particles with
increasing temperature. For a discussion on the phase segregation in Ce-Zr mixed
oxides under high temperature treatment we refer to Di Monte & Kašpar [2] and
references cited therein.
It is known that the lattice parameters (a/b/c) of the crystalline fluorite phase change
upon doping of ceria with other metals [2,7,43,46,86]. The corresponding 2Θ-shift in
the XRD peaks depends on type and amount of dopant. A powder sample can be
considered as a collection of crystallites with different d-spacings with a broadening of
19
the XRD peaks as a result of the bandwidth of the d-value within the powder. An
inhomogeneous composition therefore introduces uncertainty into the determination of
particle size from XLBA analysis, since the peak broadening in locally inhomogeneous
phases no longer only stems from particle size, strain or the instrument. Based on the
TEM images in Figure 3, it may be assumed that the error introduced is not significant,
but the XLBA crystallite size distribution appears to produce somewhat smaller
diameters than observed in TEM, down to 1 nm. This phenomenon should not be
ignored in quantitative analyses of materials inhomogeneous composition by XRD, and
the use of complementary characterization techniques, such as TEM, is advisable.
3.1.5. TGA-TPR
The calcination temperature was chosen low enough to avoid sintering of the mixed
oxide particles, and high enough to ensure the removal of most of the synthesis residue
as gaseous decomposition products (N2/NOx, CO2 and H2O) from the basic hydrous
oxides. The freshly calcined samples were examined by TGA. Temperature-
programmed oxidation (TPO, data not shown) confirmed that most synthesis residues
were removed upon calcination. A weight loss observed at 100 – 150 °C can be
assigned to adsorbed water and carbonate species [29]. A second weight loss of about 1
wt% close to 500 °C may correspond to the amount of synthesis chemicals remaining
after calcination at 350 °C. Prior to further redox experiments in the TGA set-up, the
mixed oxide samples were therefore calcined at 500 °C to eliminate disturbances to the
reduction quantification.
Figure 9 shows the TPR profiles of the calcined Cu-Ce-Zr mixed oxide obtained during
redox cycling in the TGA. The maximum of the reduction peak lies at approx. 170 °C
with the peak of the second TPR slightly shifted to higher temperatures. This could be
caused by densification during the redox treatment also in agreement with changes in
BET and pore volume of the corresponding samples Bo and Bo-redox in Table 1.
Reoxidation of the reduced catalyst after the first TPR step results in a temperature
increase of 50 - 100 °C for 1 g of sample in flowing air at ambient temperature, due to
the pyrophoric properties of reduced copper. The temperature range for the reduction
peaks observed coincide with reduction temperatures obtained for Cu-Ce mixed oxides
20
prepared by conventional co-precipitation [6,12,46]. The main reduction peaks are
found at similar temperatures as for classical Cu-Zn-Al mixed oxide catalysts [29].
While Cu-Ce-Zr mixed oxides have the fluorite structure as the only XRD/TEM-visible
crystalline phase, the Cu-Zn-Al systems usually exhibit different oxide phases; Cu/CuO,
ZnO, according to XRD as well as (sometimes amorphous) Al2O3. Ce-Zr mixed oxides
are known to be reducible, but the degree of reduction is low for T < 200 °C according
to Overbury et al. [87]. A major contribution to the low temperature reduction stems
from the formation of OH-groups at the surface [35]. This can be neglected in TGA
because of the marginal weight of hydrogen atoms. The formation of oxygen vacancies
by H2O release would be more TGA sensitive, but plays a significant role above 500 °C
only [35]. Addition of reducible metals may improve the low temperature reducibility of
Ce-Zr mixed oxides, but this is also mainly related to the formation of OH-groups [35].
Cu has a moderate effect on the reducibility of ceria [41] that is dependent on the Cu
content [42]. In agreement with Overbury et al. [87] and Norman & Perrichon [35], who
studied the impact of noble metals, Wrobel et al. [42] found the Cu-promoted reduction
of Ce4+ to Ce3+ in Cu-Ce mixed oxides with Cu/Ce < 0.5 to occur mainly above 200 °C.
No quantitative discrimination between the formation of OH-groups and oxygen
vacancies was made in this study.
With reference to the literature summarized above, we assign the observed reduction of
the Cu-Ce-Zr mixed oxide at T < 200 °C mainly to the reduction of copper oxide. The
small, constant weight loss at T > 200 °C may originate from the reduction of Ce and
possibly to a fraction of less reducible Cu species. The existence of a broad TPR peak
with a pronounced shoulder may be related to a stepwise reduction of Cu(2+) via Cu(1+)
[42], as well as to the inhomogeneous distribution of copper in the mixed oxide sample,
as observed by STEM (Figure 8). The reducibility of single Cu atoms incorporated in
the MMO bulk could be lower than that of surface Cu or Cu particles due to
accessibility and/or stability of the oxide state. This interpretation of varying
reducibility of different copper species incorporated in the Ce-Zr mixed oxide is in
accordance with results obtained by Ramaswamy et al. [85] for nanocrystalline Cu-Zr
oxides. They found the extra-lattice species to be reduced more easily than the
framework-substituted ions. Consistently, Wrobel et al. [42] reported Cu clusters in Cu-
21
Ce oxides to be more reducible than isolated Cu(2+) ions. However, based on the
assumed heterogeneous mechanism of the co-precipitation, also sterical hindrances, i.e.
encapsulation of in principle reducible copper species by the Ce-Zr matrix, probably
causes variations in the reduction behaviour of copper.
The degree of reduction was quantified by integrating over the low temperature
reduction peaks and assuming a stoichiometry of Cu:O = 1, not taking into account
(possible) partial reduction to Cu(1+) [42]. The obtained copper content in Bo is given
in Table 3 (Bo-TPR-1 and BoTPR-2). Systematic errors may result from the choice of
baseline and the simplified reduction stoichiometry. The deviation between the fraction
of Cu obtained from TPR-1 (0.16) to that determined by ICP-AES (0.23(5)) indicates
that Cu is not completely reduced upon the first TPR. The reduction degree increases
upon reoxidation and a second TPR, but even then a small amount of Cu seems to
remain in the oxidised state. A similar increase in the extent of Cu reduction from the
first to the second TPR was observed for a sample prepared according to the same
procedure and, in addition, aged in ethanol under reflux conditions, with a somewhat
higher overall reduction degree (TPR-1: 0.18, TPR-2: 0.22). The improvement in
accessibility and reducibility of the Cu could be related to changes in the nanostructure
of the primary particles upon redox cycling. The segregation of copper atoms to the
surface of the mixed oxide particles would be in agreement with XPS results reported
by Qi et al. [5] for a used Cu-Ce-La catalyst sample and Lamonier et al. [43] for a high
temperature treated Cu-Ce sample. Cu segregation is not necessarily in conflict with our
STEM-EDS results, since a slow structural change is not as easily quantified as with
XPS.
The enhanced reducibility upon redox cycling was not observed in a Cu-Ce-Zr mixed
oxide prepared by a nitrate decomposition method that resulted in two XRD-visible
phases (CuO and fluorite, data not shown). Cu-Zn-Al mixed oxides with separate phases
(CuO, ZnO, and presumably XRD-amorphous Al2O3) also usually reach complete
reduction after a single TPR [88]. It may therefore be concluded that the main part of
copper in Cu-Ce-Zr mixed oxides prepared by co-precipitation has similar reduction
behaviour as in Zn-Al supported systems, whereas a certain part of the Cu is
22
incorporated in the fluorite structure (and/or interacting with Ce atoms [42]) and hence
more difficult to reduce. The key factor for such reduction properties appears to be the
existence of the fluorite structure, with the improved reducibility interpreted as a sign of
segregation within the metastable MMO phase.
3.1.6. N2 adsorption-desorption isotherm
Figure 10 shows a N2 adsorption-desorption isotherm representative of the Cu-Ce-Zr
mixed oxides prepared by co-precipitation. The isotherm exhibits a type-IV character
[89] with a hysteresis loop commonly interpreted as a consequence of capillary
condensation in the mesopores present in the material [90]. The inset in Figure 10
displays the pore size distribution of sample Bo obtained by NLDFT analysis of the
adsorption isotherm indicating which pore sizes that contribute to the overall pore
volume. The pore sizes in the MMO material mainly lie in the mesoporous range (2 –
50 nm).
3.2. Effect of ammonia
Table 3 shows the composition of some of the prepared Cu-Ce-Zr mixed metal oxides
determined by ICP-AES. Bo is representative of an open system (setup with oil bath)
preparation with evaporation and refilling of water. The measured metal composition of
Bo is corresponds well with the nominal composition based on the amount of precursor
salts used. Bo-d120 and Bo were prepared similarly and represent an indication of the
reproducibility of the preparation in terms of elemental composition. Bc was prepared
under reflux conditions in the same apparatus. Ammonia formed after urea
decomposition was not evaporated but remained in the solution under these conditions.
As a consequence, ammonia reacts with copper to form complexes that keep a
significant amount of copper in solution, while the ratio between Ce and Zr is not
significantly affected. An open system is therefore preferred for an optimal utilization of
the copper precursor, as well as a straightforward control of the catalyst composition via
the initial precursor concentrations.
23
3.3. Effect of synthesis temperature
Spectrum (b) in Figure 4 belongs to sample Bo-s98, similar to Bo except the
temperature during precipitation and aging being gradually increased to approximately
98 °C. The colour of the precipitate gradually changed from green to dark green/grey.
Copper hydroxide is known to decompose to black CuO when boiling in aqueous
solution, may also do so at lower temperature at a lower rate [91,92]. This explains the
darkening of the solution when approaching the boiling point of water. The additional
XRD peaks could not be ascribed to one, specific compound. Instead, the peaks can be
assigned to a mixture of basic copper hydroxide and hydroxynitrate phases, including
tenorite (CuO), malachite (CuCO3·Cu(OH)2), azurite (2CuCO3·Cu(OH)2) and
gerhardtite (Cu2(OH)3NO3) ([92]). These additional phases disappeared after calcination
of the dried sample (spectrum not shown), to retain the typical peaks of the fluorite-type
structure (such as spectrum (c)).
Sample Bo-s98 has suffered loss of surface area and pore volume compared to Bo
(Table 1). This indicates a larger extent of agglomeration of the primary crystallites,
since these are of similar size. One may suspect that the formation of intermediate Cu
hydroxide species results in a higher degree of Cu segregation in the final MMO, even if
a separate CuO phase could not be detected in XRD after calcination at 350 °C.
Lamonier et al. [43] investigated Cu-Ce mixed oxides with different Cu loading and
related the existence of intermediate copper hydroxynitrates at higher Cu loadings to the
appearance of a CuO phase in the calcined MMO.
Sample BoEth (Table 1) was precipitated in the standard way (as Bo), except the
gradual addition of ethanol instead of water about 1 h after the onset of the precipitation,.
The synthesis temperature was simultaneously reduced to the boiling point of the water-
ethanol mixture (approx. 83 °C at the end of the synthesis). The aging of the precipitate
thus proceeded in boiling solution, and the colour of the suspension changed from green
via grey to red-brown in the course of precipitation. This colour change was also
observed for precipitation in the water-ethylene glycol solution at 100 °C (BoEG-s100).
During washing with ethanol and filtration in contact with air, the colour of the
precipitate changed from red-brown to (light) green. The XRD spectra of the dried as
24
well as calcined samples contained only the typical fluorite peaks and no additional
phases related to Cu species. BoEG-s100 has lower surface area and smaller pore
volume than Bo, as well as a larger primary crystallite size.
The lower Cu content of BoEth (Table 3) is probably caused by the less efficient
removal of ammonia at reduced temperature, or a change in the decomposition rate of
urea with a change in temperature and solvent. McFadyen & Matijevic [93] also
observed a red-brown precipitate during precipitation of colloidal copper hydrous oxide
sols from copper nitrate solutions at 75 °C and high pH. The red-brown material, with
larger particle size than a blue-coloured mixture of copper hydroxide and hydroxynitrate,
was assigned to CuO. Alternatively, the colour change to red-brown could be related to
reduction of Cu2+ to Cu1+, accompanied by oxidation of ethanol, in analogy with
chemical reduction of transition metals by alcohols or polyols. The colour change
during washing may thus be interpreted as the reverse redox process.
Synthesis temperatures too close to the boiling point of the solvent apparently have an
undesirable effect on surface area and pore volume, and possibly also the homogeneity
of the Cu-Ce-Zr mixed oxides. Increasing the temperature during aging enhances the
agglomeration of the primary crystallites and to give lower pore volume and accessible
surface area. Too high temperature during the early stage additionally results in larger
crystallites. An optimal MMO preparation requires optimization of synthesis
temperature and aging time, parameters that are connected to the pH of the solution via
urea decomposition. An elucidation of the reaction path of the precipitation in the
temperature-dependent phase diagram of the Cu/Ce/Zr-nitrate-urea system may explain
the appearance of additional copper phases in the dried precipitate under some
conditions.
3.4. Effect of set-up
According to the ICP-AES data in Table 3, the use of different preparation set-ups had
no significant effect on the resulting catalyst composition. HPo was prepared in the
simple set-up with the hot plate, (Figure 2A), and exhibits within experimental
25
uncertainty the same metal composition as Bo that was prepared in the oil bath (Figure
2B).
In terms of structural characteristics (Table 1), Bo has a higher surface area and pore
volume than HPo, while the size of the primary crystallites is comparable. Both
catalysts have pores in the mesoporous range (2 – 50 nm, Figure 10, pore size distr. of
HPo not shown), but pores around 20 – 30 nm are more abundant in Bo than in HPo.
The higher mesoporosity of Bo is in agreement with the trend obtained by the BJH
method (Table 1). The synthesis set-up therefore affects the agglomeration of the
primary particles during aging rather than the formation of the primary particles. Since
the time scale for temperature increase, onset of precipitation and change of colour were
similar in both set-ups, heat management and hence urea decomposition should be
comparable. This may explain why the primary particle formation is unaffected by the
choice of set-up.
The agglomeration of the primary particles is presumably affected by the flow field of
the synthesis reactor. Different types of stirrers are used in the two set-ups (Figure 2),
that probably create different flow patterns. The flow pattern controls the fluid shear
rate acting on the primary crystallites in the solution and hence affects the rate of
collisions between the particles. An increase in the average fluid shear rate may
decrease the aggregation rate due to reduced contact time between two colliding
particles [94]. Hocevar et al. [21] observed an increase in the catalytic activity of Cu-Ce
mixed oxides with increasing stirring velocity during co-precipitation. The BET surface
area of these oxides was not reported, but beside the effect of enhanced metal
distribution, the surface area of the final oxide powder may also be affected. Kunz et al.
[95] showed that ultrasonic vibration (US) can be used to enhance mixing during
precipitation to give small particles and high surface area. Thus, an increase in the
disruptive hydrodynamic forces acting on the particles in the flow field of the synthesis
reactor, by increased stirring, application of ultrasound or use of a setup with more
efficient mixing (increased average fluid shear rate), should slow down the
agglomeration of primary particles and lead to a more open pore structure.
26
3.5. Effect of drying and calcination conditions
Several samples were calcined under identical conditions but dried under different
conditions. The comparison of sample Bo (drying in static air at 100 °C) to Bo-d150
(drying in static air at 150°C), as well as BoEG-s100 (drying in static air at 100°C) to
BoEG-s100-d250 (drying in flowing air at 250°C, heating rate: 5 °C/min) in Table 1
indicates that the drying temperature affects the surface area and pore structure of the
subsequently calcined MMO. Increasing the drying temperature to 150°C or 250 °C
decreases the BET surface area approximately 29 % relative to Bo and BoEG-s100,
respectively. Figure 11 shows the pore size distributions of BoEG-s100 and BoEG-
s100-d250. The higher drying temperature has resulted in a collapse of pores larger than
10 nm. Pei et al. [96] observed a substantial reduction in pore volume and surface area
when increasing the drying temperature from 90 °C to 250 °C for mesoporous silica
prepared by spray drying. A similar, ultrasonic-treated mixed oxide exhibited a decrease
of the surface area of about 13 % upon drying at 250 °C (flowing air, heating rate: 5
°C/min) relative to 100 °C (static air, 113 m2/g). The treatment in an ultrasonic bath for
2 h hence resulted in a more dense material with lowered pore volume and surface area
already before drying. For more details on ultrasonic treatment we refer to [95].
The pore structure formed by agglomeration during aging in solution is presumably
fragile with weak bonds between the primary particles (probably OH- and a few O-
bridges). Too high drying temperature may then cause collapse of the structure because
of thermal stress and onset of precursor decomposition. The extent of structural collapse
depends on the type of material and the preparation conditions prior to drying. Materials
with high initial surface area and large pore volume may be more brittle in this context
than samples with a dense structure. Drying the precipitates at lower temperature (e.g.
90 – 100 °C) may facilitate consolidation of the pore structure by strengthening of the
bonds between primary particles through a condensation-type reaction that forms
oxygen bridges [97], while thermal stress and precursor decomposition are kept at a
minimum. The densification during following heat treatment is less pronounced.
The gas atmosphere during drying may also surface area and pore volume of the
calcined mixed oxide. For the mixed oxide subjected to ultrasonic treatment before
27
drying, drying in flowing nitrogen at 250 °C resulted in 12 % higher surface area than
drying in flowing air under equivalent conditions. Samples prepared with ethylene
glycol and water at 95 °C were dried both in static air at 100°C (BoEG) and in flowing
nitrogen at 250 °C (BoEG-d250N; analog to BoEG-s100-d250, heating rate: 5 °C/min in
flowing air). The surface areas were similar despite the different drying temperatures, in
contrast to BoEG-s100 vs BoEG-s100-d250. The pore volume of BoEG-d250N is 13 %
lower than BoEG (0.3031 cm3/g), as compared to 34 % lower for sample BoEG-s100-
d250 in flowing air relative to in static air BoEG-s100 (Table 1). The same trend was
also found for a sol-gel prepared Cu-Ce-Zr mixed oxide catalyst dried at 150 °C. Drying
in flowing nitrogen resulted in 29 % higher surface area than flowing air. Xu et al. [98]
observed a surface area increase of about 23 % for nitrogen-dried ZrO2 powder pre-
calcined at 270 °C compared to the air-dried sample.
The effect of the gas atmosphere should be related to precursor decomposition. For the
mixed oxides prepared by co-precipitation, the main decomposition peak in air lies
around 200 °C. Using nitrogen instead of air or oxygen may retard the precursor
decomposition under elevated temperatures and allow consolidation of the pore
structures. At temperatures below the decomposition temperature of remaining
precursor compounds, the impact of the gas atmosphere is negligible. This was
confirmed by Cu-Ce-Zr precipitate samples dried at 100 °C in static air, flowing air and
flowing nitrogen. Approximately the same surface area was measured within
experimental uncertainty for all three drying conditions: 123 m2/g, 119 m2/g and 118
m2/g, respectively.
Bo and HPo were placed directly in a muffle furnace at 350 °C to calcine, whereas Bo-
2C and HPo-2C were heated at 2 °C/min up to 350 °C the muffle furnace. Table 1
shows that the heating rate during calcination has a minor effect on pore volume and
surface area. Drying at 100 °C for about 11 h seems to sufficiently consolidate the
structure to tolerate the different heating rates without significant impact. An increase in
the heating rate and hence in thermal stress and rate of precursor decomposition can
possibly induce a slight densification of the materials [97]. This would depend on the
28
initial pore structure, however, since large surface area materials with open pore
structure appear more susceptible to structural collapse.
3.6. Comparison with conventional co-precipitation
Conventional co-precipitation is normally carried out with NaOH or Na2CO3 as base.
Precursor solution and alkaline solution are prepared separately and then mixed in the
precipitation, either by successively adding the alkaline solution to the precursor
solution or the other way around. Some selected examples from the literature are:
(1) Hocevar et al. [21]: 10 at% Cu in ceria, base: Na2CO3, pH < 5.5, ambient temp.,
calcination at 500 °C for 1 h, surface area: 22.5 m2/g or 0.17 m2/cm3 (with the
densities given in Table 1).
(2) Liu et al. [7,46]: 10 at% Cu in ceria, base: NaOH, pH = 10, ambient temp. (then
1h at 90 °C), calcined at 450 °C for 3 h, surface area: 96 m2/g or 0.72 m2/cm3.
(3) Kim et al.[6]: 20 at% Cu in ceria, base: NaOH, pH = 10, 70 °C, calcined at 500
°C for 5 h, surface area: 91 m2/g or 0.67 m2/cm3.
(4) Shen et al. [24]: 12 at% Cu in ceria, base: Na2CO3, added to precursor, 70 °C,
calcined at 350 °C for 12 h, surface area: 191 m2/g or 1.43 m2/cm3.
(5) Lamonier et al. [43]: 17 at% Cu in ceria, base: NaOH, precursor solution into
NaOH solution, start pH = 14, 60 °C, calcined at 400 °C for 4 h, surface area: 80
m2/g or 0.59 m2/cm3.
The surface area strongly depends on the method used and not only the calcination
conditions. According to the study of Shishido et al. [51] on Cu-Zn oxide catalysts,
homogeneous co-precipitation yields catalyst materials superior to the ones obtained by
conventional co-precipitation.Conventional co-precipitation can under certain
conditions result in surface areas comparable to those obtained by homogeneous co-
precipitation (see example (4)). Homogeneous co-precipitation with urea as base
precursor allows premixing of all components in the initial solution, and no dosing
system is required. Salt and base precursors are mixed on a molecular level in the
solution, whereas efficient mixing of salt and alkaline solution during dosing is
necessary to minimize concentration gradients for conventional techniques [95]. The
29
use of urea (or ammonia for the conventional technique) as alkaline source facilitates
easy removal of the decomposition products, while metal cations from ordinary bases,
such as NaOH or Na2CO3, might be incorporated into the precipitate and hence affect
the properties of the final material.
4. Conclusions
Homogeneous co-precipitation with urea as base precursor results in a Cu-Ce-Zr mixed
oxide of composition close to the nominal (0.23:0.54:0.23), as given by the amount of
precursor salts in solution. The material exhibits high surface area (> 100 m2/g) and
nanocrystalline primary particles (3-5 nm) composed of a single fluorite-type phase
according to XRD and TEM. This type of single-phase material is often denoted as a
solid solution, but STEM-EDS elemental mapping shows that Cu and Zr are
inhomogeneously distributed throughout the ceria matrix as a result of the
heterogeneous nature of the co-precipitation process. The EXAFS analysis indicates the
existence of CuO-type clusters within the ceria-zirconia matrix. We therefore propose
that this type of mixed metal oxide materials is viewed and denoted as heterogeneous
single-phase materials rather than as homogeneous solid solutions. XRD-based
characterization methods, such as XLBA or conventional Rietveld profile fitting, are
found to be insufficient when dealing with this type of heterogeneous single-phase
materials, and complementary characterization techniques such as TEM are needed.
The pore structure and surface areas of the mixed oxide catalysts are affected by
preparation parameters related to the precipitation stage and the subsequent heat
treatment, i.e. drying and calcination. The surface area is governed by the degree of
agglomeration of the primary crystallites.
Temperature programmed reduction (TPR) experiments find most of the Cu atoms to be
reducible and not inaccessibly incorporated into the bulk, which is important since
reduced copper is the active catalyst component in most cases. The reducibility of the
mixed oxide improves from the first to the second reduction, between which the catalyst
was re-oxidized. We attribute this to improved accessibility of the reducible components
(mostly copper) upon heat treatment, as a result of gradual, local phase segregation in
30
the metastable mixed oxide, leading to copper enrichment at the surface. Both XRD and
TEM confirm that the crystal structure of the mixed oxide was preserved upon reduction
and reoxidation. No phase separation could be detected using these two techniques,
which have limitations with respect to local variations in compositions and structure.
The results and discussion given in this paper are not necessarily limited to Cu-Ce-Zr
mixed oxides, but can to some degree apply to other catalyst formulations prepared by
co-precipitation, e.g. other single-phase materials such as Co-Ce-Zr, or multi-phase
materials such as Cu-, Ni- or Fe-based mixed oxides.
Acknowledgements
This work was supported by the Research Council of Norway through Grant No.
140022/V30 (RENERGI) and 158516/S10 (NANOMAT). Statoil ASA through the Gas
Technology Center NTNU-SINTEF is also acknowledged for their support. We
gratefully acknowledge the project team at the Swiss-Norwegian Beam Lines (SNBL) at
the ESRF for their assistance with XAS. Elin Nilsen (Department of Materials
Technology, NTNU) is gratefully acknowledged for her assistance with the XRD device.
References
[1] Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Catalytic Science
Series Vol. 2 (Ed.: G.J. Hutchings), Imperial College Press, London, 2002.
[2] R. Di Monte, J. Kašpar, J. Mater. Chem. 15 (2005) 633-648.
[3] Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179-191.
[4] E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347-357.
[5] X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062.
[6] D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107-112.
[7] Y. Liu, T. Hayakawa, T. Tsunoda, K. Suzuki, S. Hamakawa, K. Murata, R.
Shiozaki, T. Ishii, M. Kumagai, Top. Catal. 22 (2003) 205-213.
[8] X. Zhang, P. Shi, J. Mol. Catal. A 194 (2003) 99-105.
[9] X.R. Zhang, P. Shi, J. Zhao, M. Zhao, C. Liu, Fuel Process. Tech. 1680 (2003) 1-
10.
[10] A. Szizybalski, Ph.D. Thesis, Technical University Berlin, Germany, 2005.
31
[11] P. Bera, S. Mitra, S. Sampath, M.S. Hegde, Chem. Comm. 10 (2001) 927-928.
[12] T.-J. Huang, Y.-C. Kung, Catal. Lett. 85 (2003) 49-55.
[13] A. Martinez-Arias, A.B. Hungria, M. Fernandez-Garcia, J.C. Conesa, G.
Munuera, Proc. of the 13th Int. Cong. on Catal., Paris, France, July 2004.
[14] N.Y. Usachev, I.A. Gorevaya, E.P. Belanova, A.V. Kazakov, O.K. Atal´yan, V.V.
Kharlamov, Russ. Chem. Bull., Int. Ed. 53 (2004) 538-546.
[15] G. Avgouropoulos, T. Ioannides, H.K. Matralis, J. Batista, S. Hocevar, Catal.
Lett. 73 (2001) 33-40.
[16] G. Avgouropoulos, T. Ioannides, H. Matralis, Appl. Catal. B 56 (2005) 87-93.
[17] X. Tang, B. Zhang, Y. Li, Y. Xu, Q. Xin, W. Shen, Catal. Today 93-95 (2004)
191-198.
[18] P. Bera, S.T. Aruna, K.C. Patil, M.S. Hegde, J. Catal. 186 (1999) 36-44.
[19] Lj. Kundakovic, M. Flytzani-Stephanopoulos, J. Catal. 179 (1998) 203-221.
[20] Y-j. Wang, X-f. Dong, W-m. Lin, Dianyuan Jishu 26 (2002) 134-136.
[21] S. Hocevar, J. Batista, J. Levec, J. Catal. 184 (1999) 39-48.
[22] C. de Leitenburg, D. Goi, A. Primavera, A. Trovarelli, G. Dolcetti, Appl. Catal. B
11 (1996) 29-35.
[23] E.E. Ortelli, J.M. Weigel, A. Wokaun, Catal. Lett. 54 (1998) 41-48.
[24] W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 83 (2002) 33-35.
[25] C. Lu, W.L. Worrell, J.M. Vohs, R.J. Gorte, J. Electrochem. Soc. 150 (2003)
1357-1359.
[26] S. McIntosh, J.M. Vohs, R.J. Gorte, J. Electrochem. Soc. 150 (2003) 1305-1312.
[27] Y-j. Wang, X-f. Jie, X-f. Dong, W-m. Lin, Dianyuan Jishu 26 (2002) 43-46.
[28] L. Jalowiecki-Duhamel, A. Ponchel, C. Lamonier, Int. J. Hydrogen Energy 24
(1999) 1083-1092.
[29] M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal.
Today 100 (2005) 249-254.
[30] J. Kašpar, P. Fornasiero, J. Solid State Chem. 171 (2003) 19-29.
[31] M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003)
11-13.
[32] M. Ozawa, J. Alloy Compd. 275-277 (1998) 886-890.
32
[33] S.-P. Wang, X.-C. Zheng, X.-Y. Wang, S.-R. Wang, S.-M. Zhang, L.-H. Yu, W.-
P. Huang, S.-H. Wu, Catal. Lett. 105 (2005) 163-168.
[34] S. Bedrane, C. Descorme, D. Duprez, Catal. Today 75 (2002) 401-405.
[35] A. Norman, V. Perrichon, Phys. Chem. Chem. Phys. 5 (2003) 3557-3564.
[36] A. Suzuki, T. Yamamoto, Y. Nagai, T. Tanabe, F. Dong, T. Sasaki, T. Taniike,
M. Nomura, Y. Iwasawa, Proc. of the 13th Int. Cong. on Catal., Paris, France,
July 2004.
[37] M.P. Kapoor, A. Raj, Y. Matsumura, Micropor. Mesopor. Mater. 44-45 (2001)
565-572.
[38] J.P. Breen, R. Burch, D. Tibiletti, in: S.D. Jackson, J.S.J. Hargreaves, D. Lennon
(Eds.), Catalysis in Application, Royal Society of Chemistry, Cambridge, 2003,
pp. 227-232.
[39] M. Hirano, T. Miwa, M. Inagaki, J. Solid State Chem. 158 (2001) 112-117.
[40] J.A. Wang, M.A. Valenzuela, S. Castillo, J. Salmones, M. Moran-Pineda, J. Sol-
Gel Sci. Tech. 26 (2003) 879-882.
[41] Y. Zhang, S. Andersson, M. Muhammed, Appl. Catal. B 6 (1995) 325-337.
[42] (a) G. Wrobel, C. Lamonier, A. Bennani, A. D`Huysser, A. Aboukais, J. Chem.
Soc., Faraday Trans. 92 (1996) 2001-2009; (b) C. Lamonier, A. Ponchel, A.
D`Huysser, L. Jalowiecki-Duhamel, Catal. Today 50 (1999) 247-259.
[43] C. Lamonier, A. Bennani, A. D`Huysser, A. Aboukais, G. Wrobel, J. Chem. Soc.,
Faraday Trans. 92 (1996) 131-136.
[44] M. Kurtz, H. Wilmer, T. Genger, O. Hinrichsen, M. Muhler, Catal. Lett. 86 (2003)
77-80.
[45] S. Golunski, R. Rajaram, N. Hodge, G.J. Hutchings, C.J. Kiely, Catal. Today 72
(2002) 107-113.
[46] Y. Liu, T. Hayakawa, K. Suzuki, S. Hamakawa, T. Tsunoda, T. Ishii, M.
Kumagai, Appl. Catal. A 223 (2002) 137-145.
[47] Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki, K. Eguchi, Appl. Catal. A 238 (2002)
11-18.
[48] E. Matijevic, Chem. Mater. 5 (1993) 412-426.
[49] G.J.A.A. Soler-Illia, R.J. Candal, A.E. Regazzoni, M.A. Blesa, Chem. Mater. 9
(1997) 184-191.
33
[50] M. Adachi-Pagano, C. Forano, J.-P. Besse, J. Mater. Chem. 13 (2003) 1988-1993.
[51] T. Shishido, Y. Yamamoto, H. Morioka, K. Takaki, K. Takehira, Appl. Catal. A
263 (2004) 249-253.
[52] W.H.R. Shaw, J.J. Bordeaux, J. Am. Chem. Soc. 77 (1955) 4729-4733.
[53] R.B. Greegor, F.W. Lytle, J. Catal. 63 (1980) 476-486.
[54] B.S. Clausen, J.K. Nørskov, Top. Catal. 10 (2000) 221-230.
[55] M. Borowski, J. Phys. IV 7 (1997) 259-260.
[56] A.I. Frenkel, C.W. Hills, R.G. Nuzzo, J. Phys. Chem. B 105 (2001) 12689-12703.
[57] M.V. Twigg, M.S. Spencer, Appl. Catal. A 212 (2001) 161-174.
[58] A. Wootsch, C. Descorme, D. Duprez, Proc. of the 13th Int. Cong. on Catal.,
Paris, France, July 2004.
[59] M. Yada, H. Kitamura, A. Ichinose, M. Machida, T. Kijima, Angew. Chem. Int.
Ed. 38 (1999) 3506-3510.
[60] T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 14 (1997) 105-115.
[61] T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 15 (1998) 107-114.
[62] D.-H. Tsai, T.-J. Huang, Appl. Catal. A 223 (2002) 1-9.
[63] X. Zheng, X. Zhang, X. Wang, S. Wang, S. Wu, Appl. Catal. A 295 (2005) 142-
149.
[64] F.B. Noronha, E.C. Fendley, R.R. Soares, W.E. Alvarez, D.E. Resasco, Chem.
Eng. J. 82 (2001) 21-31.
[65] M.K. Dongare, V. Ramaswamy, C.S. Gopinath, A.V. Ramaswamy, S. Scheurell,
M. Bruechner, E. Kemnitz, J. Catal. 199 (2001) 209.
[66] A. Mastalir, B. Frank, A. Szizybalski, H. Soerijanto, A. Deshpande, M.
Niederberger, R. Schomäcker, R. Schlögl, T. Ressler, J. Catal. 230 (2005) 464-
475.
[67] J.C. Vargas, E. Vanhaecke, A.C. Roger, A. Kiennemann, Stud. Surf. Sci. Catal.
147 (2004) 115-120.
[68] F. Schueth, Chem. Mater. 13 (2001) 3184-3195.
[69] J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157-161.
[70] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373-380.
[71] P.B. Balbuena, K.E. Gubbins, Fluid Phase Equilib. 76 (1992) 21-35.
34
[72] DFT Plus for Windows, version 3.00, User’s Manual, Micrometrics Instrument
Corporation.
[73] W.D. Harkins, G. Jura, J. Am. Chem. Soc. 66 (1944) 1366-1373.
[74] F. Rouquerol, J. Rouquerol, K. Sing, Adsorption by Powders and Porous Solids,
Academic Press, London, 1999.
[75] S. Chytil, W.R. Glomm, E. Vollebekk, H. Bergem, J. Walmsley, J. Sjöblom, E.A.
Blekkan, Micropor. Mesopor. Mater. 86 (2005) 198-206.
[76] Diffracplus Profile, Profile fitting program, User’s Manual, Siemens.
[77] Diffracplus Win-crysize, Crystallite size and microstrain, User’s Manual, Bruker
Analytical X-Ray Systems.
[78] B.E. Warren, in: B. Chalmes, R. King (Eds.), Progress in Metal Physics Vol. 8,
Pergamon Press, London, 1959, pp. 147-202.
[79] T. Ressler, J. Synch. Rad. 5 (1998) 118-122.
[80] N. Binsted, EXCURV98: CCLRC Daresbury Laboratory computer program,
1998.
[81] F.H. Firsching, J. Mohammadzadel, J. Chem. Eng. Data 31 (1986) 40-42.
[82] F. Zhang, S.-P. Yang, H.-M. Chen, X.-B. Yu, Ceram. Int. 30 (2004) 997-1002.
[83] Y. Nagai, T. Yamamoto, T. Tanaka, S. Yoshida, T. Nonaka, T. Okamoto, A.
Suda, M. Sugiura, Catal. Today 74 (2002) 225-234.
[84] D.C. Koningsberger, B.L. Mojet, G.E. van Dorssen, D.E. Ramaker, Top. Catal.
10 (2000) 143-155.
[85] V. Ramaswamy, M. Bhagwat, D. Srinivas, A.V. Ramaswamy, Catal. Today 97
(2004) 63-70.
[86] A.S. Deshpande, N. Pinna, P. Beato, M. Antonietti, M. Niederberger, Chem.
Mater. 16 (2004) 2599-2604.
[87] S.H. Overbury, D.R. Huntley, D.R. Mullins, G.N. Glavee, Catal. Lett. 51 (1998)
133-138.
[88] Florian Huber, Zhixin Yu, Sara Lögdberg, Magnus Rønning, Hilde Venvik,
Anders Holmen, submitted.
[89] S. Brunauer, L.S. Deming, W.S. Deming, E. Teller, J. Am. Chem. Soc. 62 (1940)
1723-1732.
35
[90] K.S.W. Sing, D.H. Everett, R.A.W. Haul, L. Moscou, R.A. Pierotti, J. Rouquerol,
T. Siemieniewska, Pure Appl. Chem. 57 (1985) 603-619.
[91] A.F. Hollemann, E.Wiberg, Lehrbuch der Anorganischen Chemie, 101st ed.
(Walter de Gruyter, Berlin, 1995).
[92] R.J. Candal, A.E. Regazzoni, M.A. Blesa, J. Mater. Chem. 2 (1992) 657-661.
[93] P. McFadyen, E. Matijevic, J. Inorg. Nucl. Chem. 35 (1973) 1883-1893.
[94] J.-P. Andreassen, M.J. Hounslow, AIChE J. 50 (2004) 2772-2782.
[95] (a) U. Kunz, C. Binder, U. Hoffmann, in: Preparation of catalysts VI, eds. G.
Poncelet, J. Martens, B. Delmon, P.A. Jacobs, P. Grange (Elsevier, Amsterdam,
1995) 869-878; (b) J. Krüger, U. Hoffmann, U. Kunz, ECCE-1 Proceedings Vol.
2 (1997) 1507-1510.
[96] L. Pei, K.-I. Kurumada, M. Tanigaki, M. Hiro, K. Susa, J. Mater. Sci. 39 (2004)
663-665.
[97] C.J. Brinker, G.W. Scherer, Sol-Gel Science: The Physics and Chemistry of sol-
gel processing (Academic Press, London, 1990).
[98] B.-Q. Xu, J.-M. Wei, Y.-T. Yu, J.-L. Li, Q.-M. Zhu, Top. Catal. 22 (2003) 77-85.
[99] http://www.matweb.com; primary literature for ZrO2: ASM Engineered Materials
Reference Book, second ed., Michael Bauccio (Ed.), ASM International,
Materials Park, OH, USA, 199
[100] 4; for CeO2/CuO: CRC Handbook of Chemistry and Physics, David R. Lide (Ed.),
79th ed., CRC Press, Boca Raton, FL, USA, 1998.
[101] S. Asbrink, L.J. Norrby, Acta Cryst. B 26 (1970) 8-15.
[102] M. Wolcyrz, L. Kepinski, J. Solid State Chem. 99 (1992) 409-413.
36
Figure captions
Figure 1. Effect of material density and composition on BET data on Ce-Zr mixed
oxides. Reported in the literature [37,39]. The surface area data of Kapoor et al. [37]
(left axis) and Hirano et al. [39] (right axis) are given as surface area ratio normalized
with the values of the pure CeO2 samples (Ce mole fraction x =1). M-BET denotes the
BET surface area normalized with the molar mass (m2/mole). The surface-to-volume
ratio (stv, m2/m3) is calculated using densities based on a linear combination of
tabulated values [99] for CeO2 (7650 kg/m3) and ZrO2 (monoclinic for pure ZrO2: 5680
kg/m3, tetragonal for Ce-Zr mixed oxides: 6100 kg/m3, according to the Ce-Zr phase
diagram [1,2,30]).
Figure 2. The two different set-ups used for catalyst preparation. A: Hot plate equipped
with a magnetic stirrer, B: 500-ml 5-neck glass flask stirred with a blade agitator and
immersed in a stirred oil bath.
Figure 3. TEM images of a representative Cu-Ce-Zr mixed metal oxide powder (sample
Bo). A) As calcined sample (scale bar 50 nm). B) After pre-oxidation (heating rate: 10
°C/min, dwell: 10 min at 500 °C), reduction (up to 300 °C, 5 °C/min) and reoxidation (5
°C/min, 1 h at 350 °C). Scale bar: 5 nm. Insertion: Lattice fourier transform.
Figure 4. XRD spectra of Cu-Ce-Zr mixed metal oxide powders at different stages in
the preparation: A) Bo-dry; dried at 100 °C for about 11 h, not calcined. B) Bo-s98;
aging temperature approx. 98 °C. C) Bo; sample of (A) after calcination at 350 °C for 2
h. D) Bo-redox; sample of (C) after reduction (up to 300 °C, 5 °C/min) and reoxidation
(5 °C/min, 1 h at 350 °C).
Figure 5. Crystal size distribution and microstrain (RMS = root mean square) for the
Cu-Ce-Zr mixed metal oxide (sample Bo) as estimated from XRD data by means of a
software-based X-ray line broadening analysis (XLBA). A) Relative frequency. B)
Cumulative frequency. C) RMS strain.
37
Figure 6. Normalized Cu K-edge XANES spectra of a typical calcined Cu-Ce-Zr mixed
metal oxide powder (sample Bo-d120) and reference materials (Cu foil, CuO and Cu2O
powder). The spectra of Cu2O and Cu foil are shifted upwards along the ordinate axis
for visualisation.
Figure 7. A) Cu K-edge EXAFS spectra and B) corresponding Fourier transforms of a
Cu-Ce-Zr mixed oxide catalyst (sample Bo-d120, k3 weighting, fitting interval: ∆k = 3-
9 Å-1), containing both the experimental and the fitted theoretical curve based on small
atom approximation. The structural model takes into account the first oxygen shell
around Cu.
Figure 8. STEM-EDS elemental mapping of a calcined Cu-Ce-Zr mixed oxide powder
(sample Bo-redox) after reduction (up to 300 °C, 5 °C/min) and reoxidation (5 °C/min,
1 h at 350 °C). A) Mapping area, marked with white frame. B) EDS metal ratio of the
mapping area. C) Cu Kα1. D) Ce Lα1. E) Zr Kα1. F) O Kα1. G) EDS spectrum of the
mapping area. Scale bar: 30 nm in all images.
Figure 9. TPR profiles of a Cu-Ce-Zr mixed metal oxide (sample Bo) obtained during
the first (TPR1) and second (TPR2) in the TGA. Between TPR1 and TPR2 the catalyst
was reoxidised in air (5 °C/min, 1 h at 350 °C).
Figure 10. N2 adsorption-desorption isotherm of a Cu-Ce-Zr mixed metal oxide (sample
Bo) and the corresponding pore size distribution estimated from the adsorption isotherm
by NLDFT.
Figure 11. Effect of the drying temperature on the pore size distribution estimated from
the adsorption isotherm by NLDFT. BoEG-s100 was dried at 100 °C whereas BoEG-
s100-d250 was dried at 250 °C in air.
38
Figure 1
1
1,2
1,4
1,6
1,8
2
0 0,2 0,4 0,6 0,8 1 1,2Ce mole fraction x [ - ]
surf
ace
area
ratio
A/A
(x =
1)
[ - ]
1
3
5
7
9BET [Kap]stv [Kap]m-BET [Kap]BET [Hir]stv [Hir]m-BET [Hir]
39
Figure 2
T
T
M
A) B)
40
Figure 3
A)
B)
41
Figure 4
5 20 30 40 50 60 70 80
2 theta
sign
al [a
. u.]
(a)
(b)
(c)
(d)
42
Figure 5
A)
0
5
10
15
20
25
0 2 4 6 8 10
crystal size estimate [ nm ]
rela
tive
freq
uenc
y [ %
]
B)
0
20
40
60
80
100
0 2 4 6 8 10crystal size estimate [ nm ]
cum
ulat
ive
freq
uenc
y [ %
]
C)
0
0,4
0,8
1,2
1,6
2
0 2 4 6 8 10crystal size estimate [ nm ]
RM
S st
rain
[ 1e
-2, -
]
43
Figure 6
0,00
0,50
1,00
1,50
2,00
8,96 8,98 9,00 9,02 9,04 9,06
photon energy [ keV ]
norm
. abs
orpt
ion
[ - ]
Bo-d120CuOCu2OCu-foil
44
Figure 7
A)
B)
FT k
3 ·χ
(k)
k3 ·χ
(k)
FT k
3 ·χ
(k)
k3 ·χ
(k)
45
Figure 8
A)
B)
Atomic
[ % ]
Molar frac.
[ - ]
Cu K 10.14 0.27
Zr L 9.91 0.26
Ce L 17.88 0.47
C)
D)
E)
F)
G)
46
Figure 9
-0,4
-0,3
-0,2
-0,1
0
60 100 140 180 220 260
temperature [ °C ]
deriv
ate
wei
ght [
mg/
min
]
TPR1
TPR2
47
Figure 10
20
60
100
140
180
220
0 0,2 0,4 0,6 0,8 1
relative pressure [ - ]
nitr
ogen
ads
orbe
d [ c
m³/g
STP
]
0,000
0,002
0,004
0,006
0,008
0,010
1 10 100pore width [ nm ]
pore
vol
ume
[ cm
³/g S
TP ]
20
60
100
140
180
220
0 0,2 0,4 0,6 0,8 1
relative pressure [ - ]
nitr
ogen
ads
orbe
d [ c
m³/g
STP
]
0,000
0,002
0,004
0,006
0,008
0,010
1 10 100pore width [ nm ]
pore
vol
ume
[ cm
³/g S
TP ]
48
Figure 11
0
0,002
0,004
0,006
0,008
0,01
0,012
1 10 100
pore width [ nm ]
pore
vol
ume
[ cm
³/g S
TP ]
BoEG-s100
BoEG-s100-d250
49
Table 1. Structural parameters of Cu-Ce-Zr mixed metal oxides. For simplification, the
densities used for the calculation of the surface-to-volume ratio (stv) and the pore
volume in cm3/cm3 (based on the sample volume) are estimated by linear combination
of the tabulated densities of CeO2 (7650 kg/m3), ZrO2 (5680 kg/m3, tetragonal phase)
and CuO (6310 kg/m3).
BETa stv pore volume XRD crystal size
[ m2/g ] [ m2/mm3 ] [ cm3/g ] [ cm3/cm3 ] [ nm ]
Bc 121.13 0.86 0.1347 0.96
HPo 89.24 0.62 0.1288 0.90 3.1
HPo-2C 89.26 0.62 0.1352 0.95
Bo-d120b 145.20 1.01 0.2720 1.90 3.8
Bo-2C 163.79 1.15 0.2632 1.84 3.4
Bo 154.42 1.08 0.2678 1.87 3.3
Bo-dry 3.1
Bo-redox 141.69 0.99 0.2464 1.72 3.6
Bo-d150 109.34 0.2344
BoEth 110.49 0.78 0.1943 1.37 5.1
BoEth-dry 4.2
BoEG-s100 170.65 1.19 0.3345 2.34 3.1
BoEG-s100-dry 2.8
BoEG-s100-d250 120.76 0.85 0.2201 1.54 3.5
Bo-s98 115.29 0.1893 3.6
a Estimated error of BET smaller than 5 % of the total value. b Sample Bo-d120 suffered a temporary temperature hot spot in the initial period of the precipitation, and
the nominal drying temperature was higher than for Bo (approximately 120 °C in a different type of
drying furnace); similar XRD profile as Bo, with fluorite structure only, as well as similar N2 isotherm
and pore size distribution.
50
Table 2. Structural parameters of the first oxygen coordination shell around the Cu
central atom in nanocrystalline Cu-Ce-Zr mixed metal oxide (sample Bo-d120, fitting
interval: ∆k = 3-9 Å-1) and the reference compound CuO (k3 weighting, fitting interval:
∆k = 3-12 Å-1) obtained from a standard EXAFS analysis at the Cu K-edge. The
EXCURVE parameter AFAC reflects the amplitude reduction factor. The AFAC value
used for fitting the oxide sample was obtained by fitting the reference material CuO
with k3 weighting of the EXAFS function, but deviated only slightly from the
theoretical value 1. The parameters given for Bo-d120 result from fitting with both k1
and k3 weighting.
AFAC 0.94
E0 R1 N1 ∆σ2
[ eV ] [ Å ] [ - ] [ Å2 ]
Bo-d120 -10.5 ± 1.0 1.92 ± 0.01 2.5 ± 0.5 0.003 ± 0.002
CuO -15.2 ± 0.8 1.96 ± 0.01 3.9 ± 0.2 0.008 ± 0.001
CuOa 1.96 4.0
CeO2a 2.34 8.0
a Crystallographic data obtained from literature for CuO [101] and CeO2 [102].
51
Table 3. Composition of Cu-Ce-Zr mixed metal oxides with ICP-AES. The deviation
between the parallel runs was within 5 %. The estimated detection limit is 2000 mg/kg
for Ce, 300 mg/kg for Cu and 200 mg/kg for Zr (based on sample weight). In addition,
the copper amount determined by TPR-TGA (integration over the TPR peaks) is also
included.
Catalyst composition - molar fraction of metals
[ - ]
Cu Ce Zr
Nominal comp. 0.23 0.54 0.23
Bc 0.10 0.63 0.27
HPo 0.22 0.55 0.23
Bo-d120 0.23 0.54 0.23
Bo 0.23(5) 0.54 0.22(6)
Bo-TPR 1 0.16a
Bo-TPR 2 0.20
BoEth 0.19 0.58 0.23
BoEG-s100 0.22 0.55 0.23
a The first TPR was reproduced three times for the freshly calcined catalyst. The deviation between the
reduction degrees determined for these three runs was less than 1 % (with 0.16 corresponding to 100 %).
Paper IV
Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for
water-gas shift
Manuscript submitted.
1
Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift
Florian Huber, Hilde Meland, Magnus Rønning, Hilde Venvik*, Anders Holmen
Department of Chemical Engineering, Norwegian University of Science and Technology
(NTNU), N-7491 Trondheim, Norway
* Corresponding author; E-mail: [email protected], Tel: +47-73592831,
Fax: +47-73595047
Cu-Zn-Al and Cu-Ce-Zr mixed oxide catalysts were prepared by two different methods,
co-precipitation and flame spray pyrolysis. The performance of the catalysts was
evaluated using the water-gas shift reaction with and without CO2 and H2 added to the
feed. Cu-Ce-Zr catalysts are found not to be superior to Cu-Zn-Al catalysts in terms of
initial activity and short-term stability. Their apparent activation energy appears to be
less affected by increased concentrations of CO2 and H2.
KEY WORDS: Cu-Zn-Al; Cu-Ce-Zr; mixed metal oxides; water-gas shift.
1. Introduction
The water-gas shift (WGS) reaction is one of the oldest catalytic processes employed in
the chemical industry. There is renewed interest in this reaction because of its relevance
for producing hydrogen for use in fuel-cell systems as well as its key role in automotive
exhaust processes, since the hydrogen produced is an efficient reductant for NOx
removal [1]. Improved WGS catalysts with high activity at relatively low temperatures
and better stability than commercial Cu-Zn-Al catalysts are needed. Ceria is at present
one of the most investigated metal oxides, especially in connection with its oxygen
storage capacity (OSC) applied in three-way catalyst systems [2]. CeO2 is reported to
improve the stability of classical Cu-Zn-Al formulations [3], as is ZrO2 [4]. In addition,
the application of Ce-Zr mixed oxides has become attractive [5], since CeO2 doped with
Zr shows improved OSC and better stability [2]. In the present study, Cu-Zn-Al and Cu-
Ce-Zr mixed metal oxide (MMO) catalysts were prepared both by co-precipitation and
2
flame spray pyrolysis, and investigated under WGS conditions. The aim of the study
was to compare classical Cu-Zn-Al formulations to novel Cu-Ce-Zr MMO catalysts
from different preparation procedures at various reaction conditions.
2. Experimental
Four MMO catalysts were prepared. CuZn-CP was prepared by co-precipitation from
nitrate salts with (NH4)2CO3 in aqueous solution at ambient temperature according to a
patent of Norsk Hydro, dried at 90 °C and calcined at 350 °C (3 °C/min, 1h) [6]. CuCe-
UCP was prepared by homogeneous co-precipitation from nitrate precursors with urea
in a ethylene glycol-water mixture at 95 °C, dried at 100 °C and calcined at 250 °C (2
°C/min, 30 min) [7]. CuZn-FSP and CuCe-FSP were prepared by flame spray pyrolysis
(FSP) of organo-metallic salts dissolved in toluene [8]. The MMO catalysts were
characterised by Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-
AES), X-ray diffraction (XRD) and N2 adsorption-desorption (BET). For more details
on preparation and characterization of the catalysts we refer to [7] and [8] for the co-
precipitated and flame spray pyrolysed catalysts, respectively. The Cu dispersion of Cu-
Zn-Al catalysts has previously been determined by N2O titration (e.g. 8 % for CuZn-CP)
[3,9]. However, N2O titration was not applied for Cu-Ce-based catalysts, since
contributions from ceria may obscure the results [10].
The WGS activity was tested at atmospheric pressure in an externally heated tubular
fixed-bed reactor setup with on-line GC analysis, with the catalysts pre-reduced in
H2/N2. More details on the reduction and WGS testing of CuZn-CP/CuCe-UCP and
CuZn-FSP/CuCe-FSP are given in [7] and [8], respectively. Three different feed
compositions were applied; CuZn-CP and CuCe-UCP were investigated under a simple
WGS reactant mixture (25/125/350 Nml/min CO/H2O/N2, catalyst amount: 0.10 g) and
a simulated reformate product mixture (25/125/60/175/115 Nml/min
CO/H2O/CO2/H2/N2, 0.20 g). CuZn-FSP and CuCe-FSP were studied under a simple
WGS reactant mixture (50/100/50 Nml/min CO/H2O/N2, 0.10 g). The selectivity was
100 % in all measurements. Activation energies (Ea) were determined by the integral
method (irreversible reaction, first order in CO and zero order in H2O, plug-flow) for
3
the simple WGS reactant mixture applied to CuZn-CP and CuCe-UCP, and by the
differential approach with an Arrhenius plot for the two other WGS conditions.
3. Results and discussion
The chemical composition and structural characteristics of the catalysts are given in
Table 1. Figure 1 shows the WGS activity of the flame-sprayed (A) and co-precipitated
(B) Cu-Zn-Al and Cu-Ce-Zr MMO catalysts as a function of reaction temperature at
different WGS feed conditions. Based on the catalyst mass, the Cu-Ce-Zr catalysts show
either lower, similar or higher WGS activity compared to classical Cu-Zn-Al
formulations. Cu-Ce-Zr catalysts appear not to be generally superior to classical Cu-Zn-
Al formulations. Instead, the difference in activity seems to be related to structural
characteristics of the catalysts as well as the reaction conditions. Koryabkina et al. [10]
suggest that neither ceria nor ZnO promotes the WGS activity of copper. In an earlier
study, a promoting effect of ceria on the WGS turnover frequency of copper could not
be confirmed [3]. The activity curves for Cu-Ce-Zr in Figure 1A and 1B in the low CO
conversion range (< 20 %) can be normalized by the ratio of the Cu content in the
corresponding Cu-Zn-Al and Cu-Ce-Zr samples. The Cu-Ce-Zr catalysts then show a
similar or higher relative activity as compared to Cu-Zn-Al, indicating a good utilization
of the active material. Beside a somewhat lower Cu content in the Cu-Ce-Zr catalysts
studied here, the system has a general disadvantage in the higher molar masses of Ce
and Zr relative to Zn and Al using mass-based criteria.
The apparent activation energies (Ea) for the MMO catalysts under the chosen WGS
conditions are shown in Table 1. Figure 1C shows the Arrhenius-type plot used to
estimate Ea for CuZn-FSP/CuCe-FSP under simple WGS reactant conditions and CuZn-
CP/CuCe-UCP under simulated reformate feed conditions. The estimated activation
energies vary between 35 and 48 kJ/mole for the simple WGS reactant mixtures. Ea
increases from 44 to 83 kJ/mole with the addition of CO2 and H2 into the WGS feed for
CuZn-CP. This can be explained by a change in the rate-determining step [11]. Ea
remains unaffected by the addition of CO2 and H2 for CuCe-UCP, similar to Cu-Ce-Zr
MMO/carbon nanofiber composite catalysts [7]. Mann et al. [12] obtained activation
energies of 48 and 86 kJ/mole for forward and reverse WGS, respectively, over a Cu-
4
Zn-Al catalyst. A corresponding trend for Cu-Zn-Al MMO catalysts is found from other
reports [10,13-15]. Thus, Cu-Zn-Al MMO catalysts appear to exhibit a higher apparent
activation energy under reaction conditions where reverse WGS has to be considered,
e.g. simulated reformate feed. Consistent with our results, activation energies of 56 and
32 kJ/mole have been obtained for Cu-Ce and Cu-Ce-Al MMO catalysts, respectively,
under reformate feed conditions [10]. Ea increased from 54 to 71 kJ/mole for a Cu-Ce-
La MMO catalyst with the addition of CO2 and H2, while for a different metal ratio and
lower calcination temperature, 60 kJ/mole was obtained under a different simulated
reformate feed composition [16], as well as 19 and 30 kJ/mole under simple WGS
conditions [17]. It therefore appears that Cu-Ce-based catalysts do not display the same
increase in apparent activation energy as Cu-Zn-based catalysts upon increased
concentration of hydrogen and CO2 in the feed. Further studies that ensure the
comparability and consider the impact of additional factors such as temperature range,
conversion level, initial deactivation and Ea estimation method are necessary.
Figure 2 shows the normalized short-term deactivation behaviour of the Cu-Zn-Al and
Cu-Ce-Zr catalysts under the different WGS feed conditions and constant reaction
temperatures (300, 310 or 350 °C). All catalysts exhibit a similar deactivation behaviour
between 300 and 350 °C, independent of the feed conditions. The Cu-Zn-Al and Cu-Ce-
Zr catalysts do not differ much with respect to short-term deactivation. CuZn-FSP
shows a somewhat higher deactivation than CuCe-FSP, but has a higher surface-to-
volume ratio (Table 1). Hence, stronger deactivation of the flame-sprayed catalyst may
be related to a higher and more unstable surface, prone to sintering. CuZn-CP exhibits a
slightly lower deactivation than CuCe-UCP at 300 °C under simple WGS feed
conditions. Beside the different chemical composition of the two catalyst formulations,
a higher surface-to-volume ratio could be connected to the higher deactivation.
In summary, the performance of the catalysts depends on reaction conditions and
catalyst properties. Cu-Ce-Zr catalysts appear not to be generally superior to Cu-Zn-Al
catalysts in terms of activity or short-term stability. Their apparent activation energies
appear, however, to be less affected by the increased concentrations of CO2 and H2.
5
Acknowledgements
This work was supported by the Research Council of Norway and Statoil ASA through
the Gas Technology Center NTNU-SINTEF. Cathrine Bræin Nilsen (Depart. of Chem.
Eng., NTNU) and Bjørnar Arstad (Sintef Materials and Chemistry) are gratefully
acknowledged for preparing one of the copper catalysts and performing the WGS
measurements on the FSP samples, respectively. Tue Johannessen et al. are gratefully
acknowledged for the flame-spray synthesis.
References
[1] D. Andreeva, V. Idakiev, T. Tabakova, L. Ilieva, P. Falaras, A. Bourlinos, A.
Travlos, Catal. Today 72 (2002) 51.
[2] A. Trovarelli, Catalysis by Ceria and Related Materials, Catalytic Science Series,
Vol. 2 (Imperial College Press, London, 2002).
[3] M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal.
Today 100 (2005) 249.
[4] M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003)
11.
[5] R. Di Monte, J. Kašpar, J. Mater. Chem. 15 (2005) 633.
[6] L.A. Kristiansen, US Patent US4308176, 1981.
[7] F. Huber, Z. Yu, J. Walmsley, D. Chen, H. Venvik, A. Holmen, Appl. Catal. B,
in press.
[8] H. Meland, T. Johannessen, B. Arstad, H.J. Venvik, M. Rønning, A. Holmen,
Stud. Surf. Sci. Catal., accepted.
[9] F. Huber, Z. Yu, S. Lögdberg, M. Rønning, D. Chen, H. Venvik, A. Holmen,
Catal. Lett., in press.
[10] N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H. Ribeiro, J.
Catal. 217 (2003) 233-239.
[11] M.S. Spencer, Catal. Lett. 32 (1995) 9-13.
[12] R.F. Mann, J.C. Amphlett, B. Peppley, C.P. Thurgood, Int. J. Chem. React. Eng.
2 (2004) A5.
[13] R.L. Keiski, O. Desponds, Y.F. Chang, G.A. Somorjai, Appl. Catal. A 101 (1993)
317-338.
6
[14] Y. Choi, H.G. Stenger, J. Power Sources 124 (2003) 432-439.
[15] C.V. Ovesen, B.S. Clausen, B.S. Hammershøi, G. Steffensen, T. Askgaard, I.
Chorkendorff, J.K. Nørskov, P.B. Rasmussen, P. Stoltze, P. Taylor, J. Catal. 158
(1996) 170-180.
[16] X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062.
[17] Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179-191.
7
Figure Captions
Figure 1. A) WGS activity of CuZn-FSP and CuCe-FSP as function of reaction
temperature (50/100/50 Nml/min CO/H2O/N2, catalyst amount: 0.10 g). CuCe-FSP-
norm corresponds to CuCe-FSP normalized with the ratio of the Cu content in CuZn-
FSP (28 wt%) and CuCe-FSP (10 wt%). B) WGS activity of CuZn-CP and CuCe-UCP
as function of reaction temperature for two different feeds: SWR (25/125/350 Nml/min
CO/H2O/N2, 0.10 g) and SRP (25/125/60/175/115 Nml/min CO/H2O/CO2/H2/N2, 0.20
g). CuCe-UCP-SWR/SRP-norm corresponds to CuCe-UCP-SWR/SRP normalized with
the ratio of the Cu content in CuZn-CP (24 wt%) and CuCe-UCP (10 wt%). C)
Arrhenius-type plot for CuZn-CP/CuCe-UCP (25/125/60/175/115 Nml/min
CO/H2O/CO2/H2/N2, 225 – 275 °C) and CuZn-FSP/CuZn-FSP (50/100/50 Nml/min
CO/H2O/N2, 180 – 310 °C) with all CO conversions lower than 13 %.
Figure 2. Normalized short-term deactivation behaviour, based on the initial CO
conversion. A) CuZn-FSP/CuCe-FSP at approx. 310 °C (50/100/50 Nml/min
CO/H2O/N2). The initial CO conversion is obtained from the y-interception of the linear
fit at TOS = 0 h. B) CuZn-CP/CuCe-UCP at 300 °C (SWR: 25/125/350 Nml/min
CO/H2O/N2) and 350 °C (SRP: 25/125/60/175/115 Nml/min CO/H2O/CO2/H2/N2). The
initial CO conversion is calculated as the average of the first three measurements
obtained at intervals of 3 minutes.
8
Figure 1
A)
0
3
6
9
12
15
180 200 220 240 260 280 300 320
temperature [ C ]
CO
con
vers
ion
[ % ]
CuZn-FSP
CuCe-FSP
CuCe-FSP_norm
B)
0
20
40
60
80
100
160 210 260 310 360temperature [ C ]
CO
con
vers
ion
[ % ]
CuCe-UCP_CO+H2O CuZn-CP_CO+H2OCuCe-UCP_reformate CuZn-CP_reformateCuCe-UCP_CO+H2O_norm CuCe-UCP_reformate_norm
9
C)
-1
0
1
2
3
0,0016 0,0018 0,002 0,0022 0,0024
1/T [ 1/K ]
ln(C
O c
onve
rsio
n) [
- ]
CuCe-UCP
CuZn-CP
CuZn-FSP
CuCe-FSP
10
Figure 2
A)
0
0,2
0,4
0,6
0,8
1
1,2
0 2 4 6 8 10 12
TOS [ h ]
norm
aliz
ed C
O c
onve
rsio
n [ -
]
CuZn-FSP
CuCe-FSP
B)
0
0,2
0,4
0,6
0,8
1
1,2
0 4 8 12 16 20
TOS [ h ]
norm
aliz
ed C
O c
onve
rsio
n [ -
]
CuCe-UCP_reformate
CuZn-CP_reformate
CuCe-UCP_CO+H2O
CuZn-CP_CO+H2O
Table 1. Chemical composition, structural characteristics and apparent WGS activation energies (Ea) of the catalysts.
catalyst composition
molar metal fraction a BET b csurface
volume XRD d Ea
Cu Ce or Zn Zr or Al crystal size CO+H2O reformate
[-] [-] [-] [m2/g] [m2/mm3] [nm] [kJ/mole]
CuZn-FSP 0.32(7) 0.43(6) 0.23(7) 160 0.86 n.d. 37
CuCe-FSP 0.23 0.54 0.23 93 0.65 n.d. 48
CuZn-CP 0.27 0.40 0.33 80 0.41 14 44 f 83
CuCe-UCP 0.23 0.54 0.23 164 1.15 2 32 f 33
a Nominal composition for CuZnAl-FS and CuCeZr-FS, composition of CuZnAl-CP determined by ICP-AES, nominal composition of CuCeZr-UCP verified by
ICP-AES. b Performed on the calcined samples, after drying at 150 °C under vaccum for 1.5 h. c Surface-to-volume ratio estimated by linear combination of tabulated densities for CeO2 (7650 kg/m3), ZrO2 (5680 kg/m3, tetragonal phase), CuO (6310 kg/m3),
ZnO (5660 kg/m3) and γ-Al2O3 (3650 kg/m3). d Crystal size of CuZn-CP estimated with Scherrer equation for the Cu (111) reflection taking into account the thickness of the passivation layer [3,9]; for CuCe-
UCP estimated by software-based X-ray line broadening analysis [7]. f Ea for the temperature range 165 – 210 °C.
Paper V
Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for water-gas shift:
Carbon nanofibers as dispersing agent for the mixed oxide particles
Applied Catalysis B: Environmental, accepted.
1
Nanocrystalline Cu-Ce-Zr mixed oxide catalysts for water-gas shift:
Carbon nanofibers as dispersing agent for the mixed oxide particles
Florian Hubera, Zhixin Yua,1, John C. Walmsleyb, De Chena, Hilde J. Venvika,*, Anders
Holmena
aDepartment of Chemical Engineering, Norwegian University of Science and
Technology (NTNU), N-7491 Trondheim, Norway bSINTEF Materials and Chemistry, N-7465 Trondheim, Norway
* Corresponding author; E-mail: [email protected], Tel: +47-73592831,
Fax: +47-73595047 1 Present address: SINTEF Materials and Chemistry, N-7465 Trondheim, Norway
Nanocomposite catalysts containing carbon nanofiber (CNF) and Cu-Ce-Zr mixed metal
oxide (MMO) have been prepared by homogeneous co-precipitation with urea. The
water-gas shift reaction (WGS) has been used as test reaction. The CNF-containing
nanocomposite catalysts exhibit similar overall catalytic activity and stability as the
corresponding CNF-free catalyst. 13 wt% of the MMO could be replaced by CNF
without decreasing the overall activity and stability of the catalyst. The specific activity
of the nanocomposites based on the total metal oxide content is similar or higher than
the activity of the CNF-free material, depending on the CNF content. Similar activation
energies are, however, obtained for the CNF-free and CNF-containing materials. We
can not exclude that the CNF material acts as reaction promoter under certain conditions,
but suggest that the impact of CNF addition on the precipitation of the mixed oxide
particles, and hence the catalytic activity relative to the CNF-free MMO, should also be
considered. CNF may be regarded as inert dispersing agent material improving the
precipitation of the MMO under conditions where the co-precipitation of the MMO
precursors does not result in materials with high surface area.
KEY WORDS: Carbon nanofibers; dispersing agent; Cu; Ce; Zr; mixed metal oxides;
nanocomposites; homogeneous co-precipitation; water-gas shift.
2
1. Introduction
Mixed metal oxide (MMO) catalysts containing Cu, Ce and Zr are widely used in
environmental heterogeneous catalysis. Examples include reactions relevant in
hydrogen production for fuel cell applications (methanol steam reforming [1-4], water-
gas shift [5-7], selective CO oxidation[8-13]), methanol synthesis [14,15] and selective
catalytic NO reduction for automotive emission control [16]. Copper in its reduced state
is typically held as the active catalyst component in these materials. Ceria and zirconia
as support materials enhance catalytic activity and stability via metal-support
interactions and/or improved dispersion of the active metal component [2,3,17-19]. A
large and stable catalyst surface area, as well as a homogeneous metal distribution, are
prerequisites for high catalytic activity.
Carbon nanofibers (CNF) have been attracting increasing interest as catalyst support
material over the recent years [20]. The material exhibits high mechanical strength, high
electrical conductivity and medium to high specific surface area. Moreover, the metal
dispersed on CNF can be readily recovered from a spent catalyst by burning off the
nanofibers. In MMO formulations, CNF can be incorporated as dispersing agent for the
metal oxide particles. By decreasing the degree of agglomeration of the metal oxide
particles, the exposed metal surface area and thereby the number of active surface sites
can be increased. CNF is in some cases proposed to serve as adsorbent and activator for
the reactant molecules [21,22].
Homogeneous alkalinization via urea hydrolysis is a useful technique for preparation of
MMO with high surface area, homogeneous metal distribution and well-defined particle
size and shape [5,7,23-26]. In principle, urea decomposes at elevated temperatures in a
two-step reaction releasing ammonium and carbonate ions into the metal salt solution
accompanied by a simultaneous increase in the pH, which leads to the precipitation of
metal basic carbonates [23,25,27]:
22 2 2 4 3( ) 2 2CO NH H O NH CO+ −+ → + (1)
The decomposition rate strongly depends on the reaction temperature [25,27].
3
In terms of synthesizing MMO ‘solid solutions’, one has to realize that co-precipitation
in general has to be regarded as a heterogeneous process. A phase containing only one
cation first nucleates to serve as locus for the heterogeneous nucleation of a second solid.
Further growth proceeds incorporating the different cations at different rates [24]. As a
result, the internal chemical composition of such composites usually varies from the
center to the periphery [23].
In the present study, a series of CNF-containing Cu-Ce-Zr mixed metal oxide catalysts
were prepared by homogeneous co-precipitation. The water-gas shift (WGS) reaction
was used as test reaction. The catalytic performance of the CNF-containing samples was
compared with that of the corresponding CNF-free Cu-Ce-Zr mixed oxide sample, thus
examining the effect of the carbon nanofibers in the CNF-metal oxide nanocomposite
catalyst.
2. Experimental
2.1. Chemicals
Copper(II)-nitrate-trihydrate (> 99 %), cerium(III)-nitrate-hexahydrate (> 99.5 %) and
zirconyl(IV)-nitrate-hydrate (> 99.5 %) were purchased from Acros Organics. Urea (>
99.5 %) was purchased from Merck. Ethylene glycol (EG) (> 99.5 %) was purchased
from Fluka. All chemicals were used as-received. All catalyst preparations were
conducted with deionized water.
2.2. Catalyst preparation
CNF with platelet structure were synthesized by catalytic chemical vapor deposition as
previously described [28], from CO/H2 (40/10 ml/min) decomposition on Fe3O4
nanoparticles at 600 °C. Such a procedure results in CNF with the graphite sheets
oriented perpendicular to the fiber axis, as shown in Figure 1, and average diameter of
approx. 116 nm. The XRD diffraction pattern exhibits sharp, graphitic (002) reflections,
indicating a high degree of structural order. The value of d-spacing examined from the
(002) line is 3.45 Å, close to that of graphite. The as-grown CNF were boiled in
concentrated nitric acid for 3 hours, followed by washing with water and drying
overnight in a muffle furnace at 120 °C. After such oxidation treatment, the remaining
4
amount of catalyst (Fe2O3) is decreased from 3.1% to 0.6%, and surface oxygen groups
have been introduced at the CNF surface to increase the hydrophilicity.
In the MMO/CNF synthesis, a total amount of 0.105 mole of Cu-, Ce- and Zr-salt
(nominal composition: Cu:Ce:Zr = 0.23:0.54:0.23) was dissolved in 300 ml of a mixture
of 40 vol% ethylene glycol (EG) in water at ca. 30 °C under stirring. The CNF (11.1,
22.2 and 44.2 g CNF/mole total metal) was suspended in 150 ml 40 vol% EG-water and
treated in an ultrasonic bath before adding to the salt solution. Finally, 0.904 mole of
urea was added to the mixture, resulting in a pH of approx. 2 of the mixture. The
mixture was poured in a 500-ml 5-neck glass flask, placed in a hot oil bath and rapidly
heated up to 95 °C under vigorous stirring with a blade agitator at approx. 750 rpm. The
temperature of the solution was monitored with an immerged thermometer.
The setup was operated in an open mode in order to avoid loss of Cu through
complexation with NH3 in the solution [29]. Water was allowed to evaporate through
the unused flask necks, and the loss was compensated by continuous refilling with water
(at ambient temperature, not pre-heated). The addition of water was carried out in the
same way for all catalyst preparations, but variations may influence the formation of the
precipitate and thereby the final material properties. The investigation of this parameter
was not part of this study.
The mixture was kept in the oil bath for a total period of 8 hours, including heating (ca.
15 – 20 min), precipitation and aging. The suspension was then removed and rapidly
cooled down to room temperature using cold water. The solid precipitate was filtered
off and washed three times with 200 ml of deionized water (40 – 50 °C). The precipitate
was dried for about 12 h at 100 °C and calcined for 30 min at 250 °C (heating up: 2
°C/min; cooling down: ca. 2 °C/min) in a muffle furnace.
The CNF-free sample was prepared in the same way as the three CNF-containing
samples, except that the nitrate precursors were dissolved in 450 ml 40 vol% EG-water.
Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) has been used
to determine the actual amount of copper, cerium and zirconium on CNF-free MMO
5
after calcination. The actual composition of the MMO catalysts was in good agreement
with the nominal composition: Cu:Ce:Zr ≈ 0.23:0.54:0.23 [29].
The catalysts are denoted CNF00, CNF07, CNF13 and CNF24 with the digits
specifying the CNF content (in wt%) in the CNF-metal oxide nanocomposites. 11.1,
22.2 and 44.2 g CNF/mole total metal corresponds to 7.2, 13.4 and 23.6 wt% CNF in
the nanocomposite, respectively. Pure platelet CNF after oxidation treatment is denoted
as P-CNF.
2.3. Characterization techniques
Nitrogen adsorption-desorption isotherms were measured using a Micromeritics TriStar
3000 instrument. The data were collected at -196 °C. The BET surface area was
calculated in the relative pressure interval ranging from 0.01 to 0.30. The pore volume
was estimated by the Barrett-Joyner-Halenda (BJH) method [30] as the adsorption
cumulative volume of pores between 1.7 and 300.0 nm width. This method is based on
the assumption of cylindrical pores, and the capillary condensation in the pores is taken
into account by the classical Kelvin equation.
XRD data were recorded on a Siemens diffractometer D-5005 (dichromatic CuKα+β-
radiation). Average crystal size estimates and crystal size distributions for the mixed
oxide powders were obtained from a software-based X-ray line broadening analysis
(XLBA). Selected experimental XRD peaks ((111), (200) and (220)) were simulated by
means of the software package Profile [31], using the Pearson VII model function,
removing the contribution of CuKβ to the peak intensities. The program Win-crysize
[32], utilizing the Warren-Averbach method [33], was then used to estimate the
crystallite size taking into account contributions from microstrain (scaled as mean
square root of the average squared relative strain). Contributions from instrumental line
broadening were removed using LaB6 as a reference (selected peaks: (110) at 30.38°,
(111) at 37.44° and (210) at 48.96°).
A Thermogravimetric Analyser (TA Instruments TGA Q500) in a temperature-
programmed oxidation mode (TPO) was applied to check the metal oxide content in the
6
nitrate precursors and the weight loss of the dried and calcined samples. A heating rate
of 2 °C/min and a gas flow composed of 90 ml/min air and 10 ml/min nitrogen (for
flushing the weight chamber) was applied over approx. 15 mg of sample for each TPO
uptake.
TEM data were recorded on a JEOL 2010F transmission electron microscope. Small
samples were put into sealed glass containers containing ethanol, and placed in an
ultrasonic bath to disperse the catalyst. A drop of the resulting suspension was placed
onto a holey carbon film, supported on a titanium mesh grid and dried. Conventional
TEM images were recorded onto a CCD camera. Samples were also examined in
scanning transmission electron microscope (STEM) mode, with a nominal probe size of
approx. 0.7 nm, acquiring bright field and dark field STEM images. Energy dispersive
x-ray spectroscopy (EDS) analysis and mapping were performed using an Oxford
Instruments INCA system, employing drift compensation to correct for sample
movement during the acquisition of maps.
2.4. Catalyst testing
The water-gas shift (WGS) activity was tested at atmospheric pressure in an externally
heated tubular fixed-bed reactor made of quartz with inner diameter of 10 mm. The
catalyst particles (50 – 125 µm) were loaded onto quartz wool. The quartz reactor was
heated inside an electric furnace. A carbonyl trap was connected to the outlet of the CO
gas cylinder to remove possible iron carbonyls. The gas flows were controlled by mass
flow controllers. Water was dosed with a liquid flow controller from a He-pressurized
cylinder. The water was injected into a vaporizing unit, and the steam was then mixed
into the gas stream. A microchannel heat exchanger (Forschungszentrum Karlsruhe)
operated with air was used for condensation of liquid products prior to analysis of dry
product gas with an on-line Agilent G2891A Micro GC equipped with thermal
conductivity detection (TCD). The gas composition in the dry gas was quantified by
using a range of certified calibration gases as external standards.
The catalyst samples were pre-reduced in 10 % H2 in nitrogen at 250 °C for 1 h (total
flow: 300 Nml/min; heating rate: 5 °C/min), and cooled in the H2/N2 atmosphere to the
7
starting temperature of the activity measurements. The water-gas shift reaction was
studied under two different feed conditions. A simple WGS reactant mixture (25
Nml/min CO, 125 Nml/min H2O and 350 Nml/min N2) was fed over 0.1 g (approx. 0.1
ml) of catalyst. A simulated reformate product mixture (25 Nml/min CO, 125 Nml/min
H2O, 60 Nml/min CO2, 175 Nml/min H2 and 115 Nml/min N2) was applied in
experiments with 0.2 g catalyst (approx. 0.2 ml). The initial stabilization of the feed was
carried out in the by-pass line. The temperature was increased in steps of 15 – 25 °C
during the activity measurements, and stabilized for 15 min. (approx. 30 min for the
starting temperature) before three GC analyses were taken at three minute intervals. The
short-term deactivation behaviour of the catalysts was recorded at the final temperature
for about 15 h.
The CO conversion (XCO) for the feed containing only CO and H2O (and balance N2)
can be calculated from the CO and CO2 concentrations in the dry exit gas measured at
each temperature:
2
2
100%( )CO
COXCO CO
= ⋅+
(2)
For the simulated reformate feed, the calculated CO conversion must take into account
the initial composition of the feed gas:
0
2 2,0
( ) 100%( )CO
CO COXCO CO CO
−= ⋅+ −
(3)
The selectivity was 1 for all activity measurements. Trace CH4 was detected during the
temperature scan as well as the following short-term deactivation experiment.
Apparent activation energies (Ea) were estimated from the temperature scans. For the
CO and H2O containing feed, the activation energy was determined by the integral
method assuming an irreversible reaction of first order in CO and zero order in H2O as
well as plug-flow conditions. According to Keiski et al. [34], the water-gas shift
8
reaction is not a simple first-order reaction in the CO activity, however, the first-order
rate equation describes the phenomenon quite well. The model was fitted to the
experimental data in the temperature range 165 – 210 °C applying the least-square
method. The CO conversion was below 30 % and hence the approach to WGS
equilibrium composition less than 1 % legitimating the use of an irreversible reaction
model. The approach to equilibrium is calculated with the exit gas concentrations and
the equilibrium constant:
2 2
2
100%eq
CO HEquilibrium approachK CO H O
⋅= ⋅⋅ ⋅
(4)
For the simulated reformer feed, the activation energies were estimated from an
Arrhenius plot in the temperature range 225 – 275 °C by the differential approach, with
the CO conversion always lower than 15 %. Possible heat and transport limitations were
considered by applying well-known empirical evaluation criteria [35].
The standard deviation of the CO conversion was estimated to be 4 %. The carbon
balance for CO and CO2 was within approx. 1 %, except for the three highest
temperature levels in the experiments with the simulated reformate feed. Here, the
negative deviation from the initial carbon content gradually increased to 3 %, as a result
of a systematic error connected to the GC calibration of CO2.
3. Results
3.1. Catalyst characterization
Figure 2 displays TPO profiles of the dried catalyst materials CNF00 and CNF13 as
well as calcined CNF13. The CNF-free MMO material shows a major decomposition
peak around 200 °C corresponding to about 10 % weight loss. This peak disappeared for
the corresponding calcined sample (data not shown) verifying that most of the precursor
remainding in the dried sample was removed by the calcination treatment. In addition, a
small peak corresponding to approx. 1 % weight loss is observed around 400 °C. The
CNF13 sample exhibits two major peaks after drying, at approx. 200 °C and 400 °C.
The peak at 200 °C, corresponding to approx. 9 % weight loss, can be assigned to
9
remaining precursor material in the dried sample. The peak at 400 °C corresponds to
approx. 13.5 % weight loss, and can be assigned to the oxidation of the carbon
nanofibers to CO and CO2. The pure CNF material oxidizes at around 500 °C (data not
shown). The metal oxides thus catalyze the oxidation of the nanofibers, decreasing the
oxidation temperature thereby. Only the peak at 400 °C remains after calcination of
CNF13, corresponding to approx. 13 % weight loss, i.e. the calcination treatment
removes remaining precursor material but does not oxidize the carbon nanofibers.
Temperature-programmed reduction (TPR) measurements performed on a CNF-free
Cu-Ce-Zr mixed oxide sample showed that Cu is reduced at around 170 °C, and that the
degree of Cu reduction increases from the first to the second reduction [29].
Figure 3 shows XRD spectra of the nanocomposites and the CNF-free material after
drying, as well as P-CNF for reference. The powder samples exhibit the fluorite-type
structure with the typical (111), (200), (220) and (311) reflections after drying at 100 °C.
With increasing CNF content, the graphitic (002) reflection becomes more significant in
the nanocomposites. The fluorite crystal size estimated from the XRD spectra (Table 1)
is similar for all samples and hence not significantly affected by the CNF content. The
XRD spectra and hence the estimated crystal size does not change significantly upon
calcination.
Figure 4 shows TEM images of CNF13 after calcination at 250 °C. The low-resolution
image (Figure 4A) displays agglomerates built up of the primary CuCeZr MMO
crystallites. The agglomerates are dispersed to a certain degree by the CNF, depending
on the CNF content. The high-resolution image (Figure 4B) displays the crystalline
structure of the CuCeZr MMO primary particles. The lattice spacing demonstrates the
random orientation of the nanocrystalline particles. These TEM data confirm the
existence of primary crystallites in the range of a few nanometers, for which the
software-based XLBA appears to slightly underestimate the size (Table 1). A reasonable
estimate for the primary crystallite size would be 3 – 5 nm.
10
3.2. Catalytic activity for water-gas shift
Figure 5 shows the conversion of CO over the catalysts for the simple WGS reactant
mixture as a function of temperature. CNF13 exhibits the highest WGS activity among
the three nanocomposite catalysts, and its performance is similar to the activity of
CNF00. In Table 2, the CO conversion of the three nanocomposites is normalized with
the weight content of the MMO and divided by the CO conversion of the CNF-free
catalyst. CNF07 and CNF24 show a relative WGS activity close to 1, i.e. the metal
oxide components in these two nanocomposites exhibit a catalytic performance similar
to the metal oxide in CNF00. CNF13 shows a relative WGS activity of 1.2, as reflected
in an overall WGS activity similar to CNF00 despite a lower MMO content. The
estimated activation energies of the four catalyst materials all lie in the range 31 – 34
kJ/mole (Table 2). The addition of the platelet CNF does not significantly affect the
WGS activation energy under the given temperatures and reaction conditions. Figure 6
shows the normalized short-term deactivation behaviour of all four catalysts at 300 °C
for the simple WGS reactant mixture. The catalysts show similar deactivation up to 15 h,
with CNF24 exhibiting a slightly higher deactivation rate than the other catalysts. The
addition of the platelet CNF is not found to improve the short-term stability of the metal
oxide catalyst under the given reaction conditions.
Figure 7 shows the CO conversion over CNF00, CNF13 and CNF24 for the reformate
reactant mixture as a function of temperature. CNF13 exhibits a WGS activity higher
than CNF24 and similar to CNF00, as for the CO/H2O/N2 mixture. CNF24 shows a
lower CO conversion than CNF00 and CNF13 at the highest temperatures (> 280 °C).
The MMO weight-normalized CO conversion of CNF13 and CNF24 divided by the CO
conversion of the CNF-free catalyst is given in Table 3. Under the reformate feed
conditions, the metal oxide components in both CNF13 and CNF24 exhibit a relative
WGS activity higher than 1. The activation energies of the three catalysts as determined
from the Arrhenius plot in Figure 8 are in the range 33 – 39 kJ/mole (Table 3). The
addition of the platelet CNF does not significantly affect the WGS activation energy for
the given temperatures and reaction conditions, as for to the CO/H2O/N2 mixture.
Furthermore, the activation energies are rather similar for the pure WGS and the
simulated reformate reactant mixture, i.e. the addition of CO2 and H2 does not affect the
11
activation energies considerably, bearing in mind that the values for pure WGS and the
reformate reactant mixture are determined in different ways.
Figure 9 shows the normalized short-term deactivation behaviour of CNF00, CNF13
and CNF24 at 350 °C for the reformate product mixture. The catalysts display similar
deactivation up to 16 h. In general, the addition of the platelet CNF does not improve
the short-term stability of the metal oxide catalyst under the reaction conditions applied.
The short-term deactivation behaviour is similar under pure WGS and the simulated
reformate product mixture. The deactivation test temperatures are, however, somewhat
different, and the approach to equilibrium conditions is 1 % for the pure WGS mixture
as compared to 50 % for the reformate reactant mixture.
4. Discussion
Dong et al. [21] found the optimum amount of Herringbone-type CNF as dispersing
agent in Cu-Zn-Al mixed oxide catalysts for methanol synthesis to lie around 11 g
CNF/mole total metal (approx. 13 wt% CNF). The same group also found the optimum
amount for Co-Cu catalysts for synthesis of higher alcohols to lie around 10 g
CNF/mole total metal (approx. 11 wt%) [22]. To some extent, the optimum amount of
CNF dispersing agent appears not to depend strongly on the MMO catalyst and the type
of reaction chosen. In our study, the catalyst prepared with 11 g platelet CNF/mole total
metal (approx. 7 wt%, CNF07) did not improve the catalyst properties significantly.
Instead the optimum amount appears to lie around the double amount, 22 g CNF/mole
(13 wt%, CNF13), with reservations concerning the limited resolution of the data points
for the CNF loading. Despite the decreasing amount of MMO in CNF07 relative to
CNF13, the catalyst performance remains at the same level. A further decrease in MMO
content and increase in CNF loading to 44 g/mole (24 wt%, CNF24) decreases the
catalyst activity and indicates that there is a limit to how much of the active catalyst
component can be replaced by inert material. The Cu-Ce-Zr mixed oxide materials
prepared and studied here are capable of utilizing the double amount of CNF per mole
total metal compared to the two studies mentioned above, whereas based on the weight
content of the CNF in the MMO catalyst our results are similar. The optimum amount of
CNF dispersing agent for MMO catalysts may thus depend on the type of CNF used, the
12
conditions applied for preparing the nanocomposites, as well as the basis for the
comparison.
All CNF-containing catalysts exhibit a superior overall activity compared to the CNF-
free samples in the previously mentioned studies [21,22]. Dong et al. [21] as well as
Zhang et al. [22] suggest the CNF to be more than a mere dispersing agent material. As
adsorbent and activator of reactant molecules (e.g. H2-adsorption/spillover and electron
transfer), they may facilitate higher stationary-state concentrations of the active species
on the catalyst surface, thereby improving activity and affecting the selectivity. This
effect would be less distinctive in our study, but can not be generally excluded (see for
example CNF-supported noble metal catalysts for hydrogenation of cinnamaldehyde
[36,37]). In our study, the CNF-containing samples exhibit a higher activity than the
CNF-free catalyst (Table 2 and Table 3) only when based on the metal content.
The influence of the preparation conditions when comparing CNF-containing and CNF-
free catalysts is, however, also important. The CNF may influence the precipitation,
even if appearing inert under reaction conditions. For CNF-free MMO catalysts
prepared by co-precipitation, the resulting surface area depends on the mixing efficiency
during the co-precipitation process [29,38,39]. Efficient stirring results in small particles
with high surface area and hence high catalytic activity. Small particles can be
introduced as seeds into the precipitation reactor to offer nucleation centers in the
growth process [40]. For carbon nanofibers with carboxylic acid surface groups, metal
ion adsorption may occur prior to nucleation of the metal hydroxides. Small adsorbed
metal clusters may then serve as nucleation sites for the formation of mixed metal
hydroxide crystallites. Further nucleation and crystal growth may then occur on these
primary nuclei [41,42]. Shape, size distribution and agglomeration degree of the final
mixed oxide crystals as well as the degree of attachment to the carbon support may
depend on size and shape of the CNF [40].
The deposition-precipitation on CNF may lead to smaller particles, dispersed by the
seed material, and hence higher surface area of the active materials. The addition of
CNF may thus compensate inefficient mixing during precipitation, whereas the impact
13
of CNF on nucleation and crystal growth may diminish under efficient mixing
conditions. The efficiency of the mixing in [21] and [22] is not clear. However, the
lower reduction temperatures of the CNF-containing materials as compared to the CNF-
free MMO may indicate the active particles to be of different size or exhibit a different
degree of agglomeration. The preparation of MMO by homogeneous co-precipitation
with urea as base precursor has been shown to result in high surface area materials
under the preparation conditions chosen [29], and CNF addition may thus not greatly
improve the particle dispersion during precipitation. The fluorite crystal size in our
catalysts is not significantly affected by the CNF addition, within experimental
uncertainty (Table 1). The similar overall activity and the slightly increased specific
activity of the nanocomposites compared to the CNF-free MMO in the present study
can be interpreted accordingly. The small difference in apparent activation energy
(Table 2 and Table 3) also indicates that the active particles are not fundamentally
different, and the addition of CNF does not appear to have an impact on the rate-
determining step. Frøseth et al. [43] observed no significant CO chemisorption to occur
on CNF identical to the ones used in our study, using steady-state isotopic transient
kinetic analysis (SSITKA). The study was performed at 100 °C and 1.85 bar with
CO/inert (1.5/33.5 Nml/min) and CO/H2/inert (1.5/15/33.5 Nml/min) gas mixtures.
Chemisorption experiments (Micrometrics ASAP 2010) at 40 °C also showed no
significant H2 chemisorption on the pure CNF material.
The addition of CNF did not improve the short-term stability of the nanocomposite
catalysts compared to the CNF-free MMO in our study. Consequently, the deactivation
of the active phase is not affected by the CNF-MMO interface. According to the
combined XRD and TEM data, the nanocomposites consist of agglomerates of mixed
oxide crystallites that are dispersed by the CNF. One reason for deactivation could be a
decrease in the catalytically active surface area of the MMO by sintering of the
secondary mixed oxide agglomerates. Another cause could be phase separation within
the mixed oxide structure, as described earlier [29]. Both effects are not directly
affected by the CNF, hence CNF should not have an influence on catalyst deactivation
caused by these two processes.
14
The impact of CNF on the catalyst activity and stability could be related to size-
matching [44]. The mixed oxide crystallites and the CNF exhibit a rather large size
difference in our study (Table 1). The addition of CNF might improve the catalyst
activity and stability in cases where CNF with smaller diameters are combined with
precipitation procedures resulting in larger mixed oxide crystals than under the given
conditions, possibly at the expense of a lower initial, catalytically active surface area. In
summary, effects of CNF size, structure and surface properties as well as all parameters
concerning preparation of the active phase have to be considered before concluding on
the optimum amount of CNF in nanocomposites, and especially if conclusions on the
activity of the CNF itself are to be drawn.
5. Conclusions
Under the preparation and reaction conditions chosen in this study, the CNF-containing
nanocomposite catalysts exhibit overall activity and stability similar to the CNF-free
catalyst. The main advantage of the nanocomposites relates to conserving raw materials
for MMO catalysts, since about 13 wt% of the MMO could be replaced by CNF without
decreasing the overall activity of the catalyst. CNF are also a recycling-friendly additive
in MMO materials, since they can be easily removed by oxidation at 400 °C.
The specific activity of the nanocomposites, based on the total metal oxide content, is
similar to or higher than the activity of the CNF-free material, with 13 wt% CNF found
as optimum in our study. We can not exclude the possibility that the CNF material acts
as a reaction promoter under certain conditions, but suggest that the impact of CNF
addition on the precipitation of the mixed oxide particles, and hence the catalytic
activity relative to the CNF-free MMO, is important. This is supported by the similar
activation energies obtained for the CNF-free and CNF-containing materials in our
study and SSITKA results on CO chemisorption on CNF. CNF may therefore be
considered as an inert dispersing agent material improving the precipitation of the
MMO under conditions where the co-precipitation of the MMO precursors does not
result in materials with high surface area. The size distribution and surface properties of
the CNF will also influence the precipitation.
15
Acknowledgements
This work was supported by the Research Council of Norway through Grant No.
140022/V30 (RENERGI) and 158516/S10 (NANOMAT). Statoil ASA through the Gas
Technology Center NTNU-SINTEF is also acknowledged for their support. Elin Nilsen
(Department of Materials Technology, NTNU) and Jon Arvid Lie (Department of
Chemical Engineering, NTNU) are gratefully acknowledged for their assistance with
the XRD and the TGA devices, respectively.
References
[1] Y. Liu, T. Hayakawa, T. Tsunoda, K. Suzuki, S. Hamakawa, K. Murata, R.
Shiozaki, T. Ishii, M. Kumagai, Top. Catal. 22 (2003) 205-213.
[2] X. Zhang, P. Shi, J. Mol. Catal. A 194 (2003) 99-105.
[3] X.R. Zhang, P. Shi, J. Zhao, M. Zhao, C. Liu, Fuel Process. Tech. 1680 (2003) 1-
10.
[4] A. Szizybalski, Ph.D. Thesis, Technical University Berlin, Germany, 2005.
[5] Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179-191.
[6] E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347-357.
[7] X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062.
[8] P. Bera, S. Mitra, S. Sampath, M.S. Hegde, Chem. Comm. 10 (2001) 927-928.
[9] T.-J. Huang, Y.-C. Kung, Catal. Lett. 85 (2003) 49-55.
[10] D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107-112.
[11] A. Martinez-Arias, A.B. Hungria, M. Fernandez-Garcia, J.C. Conesa, G.
Munuera, Proc. of the 13th Int. Cong. on Catal., Paris, France, July 2004.
[12] N.Y. Usachev, I.A. Gorevaya, E.P. Belanova, A.V. Kazakov, O.K. Atal´yan, V.V.
Kharlamov, Russ. Chem. Bull., Int. Ed. 53 (2004) 538-546.
[13] G. Avgouropoulos, T. Ioannides, H. Matralis, Appl. Catal. B 56 (2005) 87-93.
[14] E.E. Ortelli, J.M. Weigel, A. Wokaun, Catal. Lett. 54 (1998) 41-48.
[15] W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 83 (2002) 33-35.
[16] P. Bera, S.T. Aruna, K.C. Patil, M.S. Hegde, J. Catal. 186 (1999) 36-44.
[17] M. Rønning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal.
Today 100 (2005) 249-254.
16
[18] M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003)
11-13.
[19] R. Di Monte, J. Kašpar, J. Mater. Chem. 15 (2005) 633-648.
[20] K. P. De Jong and J.W. Geus, Catal. Rev. Sci. Eng. 42 (2000) 481–510.
[21] X. Dong, H.-B. Zhang, G.-D. Lin, Y.-Z. Yuan, K.R. Tsai, Catal. Lett. 85 (2003)
237-246.
[22] H.-B. Zhang, X. Dong, G.-D. Lin, X.-L. Liang, H.-Y. Li, Chem. Comm. 40 (2005)
5094-5096.
[23] E. Matijevic, Chem. Mater. 5 (1993) 412-426.
[24] G.J.A.A. Soler-Illia, R.J. Candal, A.E. Regazzoni, M.A. Blesa, Chem. Mater. 9
(1997) 184-191.
[25] M. Adachi-Pagano, C. Forano, J.-P. Besse, J. Mater. Chem. 13 (2003) 1988-1993.
[26] T. Shishido, Y. Yamamoto, H. Morioka, K. Takaki, K. Takehira, Appl. Catal. A
263 (2004) 249-253.
[27] W.H.R. Shaw, J.J. Bordeaux, J. Am. Chem. Soc. 77 (1955) 4729-4733.
[28] Z. Yu, D. Chen, B. Tøtdal, A. Holmen, J. Phys. Chem. B 109 (2005) 6096-6102.
[29] F. Huber, H. Venvik, M. Rønning, J. Walmsley, A. Holmen, Manuscript in
preparation.
[30] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373-380.
[31] Diffracplus Profile, Profile fitting program, User’s Manual, Siemens.
[32] Diffracplus Win-crysize, Crystallite size and microstrain, User’s Manual, Bruker
Analytical X-Ray Systems.
[33] B.E. Warren, in: B. Chalmes, R. King (Eds.), Progress in Metal Physics Vol. 8,
Pergamon Press, London, 1959, pp. 147-202.
[34] R.L. Keiski, O. Desponds, Y.F. Chang, G.A. Somorjai, Appl. Catal. A 101 (1993)
317-338.
[35] F. Kapteijn, J.A. Moulijn, Laboratory Reactors, in: G. Ertl, H. Knözinger, J.
Weitkamp (Eds.), Handbook of Heterogeneous Catalysis Vol. 3, VCH
Verlagsgesellschaft mbH, Weinheim, Germany, 1997, chapter 9, p. 1359.
[36] M.L. Toebes, F.F. Prinsloo, J.H. Bitter, A.J. van Dillen, K.P. de Jong, J. Catal.
214 (2003) 78-87.
17
[37] M.L. Toebes, Y. Zhang, J. Hájek, T.A. Nijhuis, J.H. Bitter, A.J. van Dillen, D.Y.
Murzin, D.C. Koningsberger, K.P. de Jong, J. Catal. 226 (2004) 215-225.
[38] S. Hocevar, J. Batista, J. Levec, J. Catal. 184 (1999) 39-48.
[39] (a) U. Kunz, C. Binder, U. Hoffmann, in: G. Poncelet, J. Martens, B. Delmon,
P.A. Jacobs, P. Grange (Eds.), Preparation of catalysts VI, Elsevier, Amsterdam,
1995, pp. 869-878; (b) J. Krüger, U. Hoffmann, U. Kunz, ECCE-1 Proceedings
Vol. 2 (1997) 1507-1510.
[40] A. Cacciuto, S. Auer, D. Frenkel, Nature 428 (2004) 404-406.
[41] M.K. van der Lee, J. van Dillen, J.H. Bitter, K.P. de Jong, J. Am. Chem. Soc. 127
(2005) 13573-13582.
[42] G.L. Bezemer, P.B. Radstake, V. Koot, A.J. van Dillen, J.W. Geus, K.P. de Jong,
J. Catal. 237 (2006) 291-302.
[43] V. Frøseth, Z. Yu, D. Chen, A. Holmen, Manuscript in preparation.
[44] B.-Q. Xu, J.-M. Wei, Y.-T. Yu, J.-L. Li, Q.-M. Zhu, Top. Catal. 22 (2003) 77-85.
18
Figure Captions
Figure 1. TEM images of as-grown carbon nanofibers (CNF) produced at 600 °C from
decomposition of CO/H2 (40/10 ml/min) over Fe3O4 nanoparticles. A) Scale-bar 10 nm.
B) Scale-bar 500 nm. The high-resolution image (A) confirms the platelet structure of
the CNF.
Figure 2. Temperature-programmed oxidation profiles of the catalyst samples CNF00
(dried sample) and CNF13 (both dried and calcined samples).
Figure 3. XRD patterns of CNF00, CNF07, CNF13 and CNF24 after drying and of P-
CNF after oxidation treatment.
Figure 4. TEM images of CNF13 after calcination at 250 °C. A) Scale-bar 50 nm. B)
Scale-bar 5 nm.
Figure 5. WGS activity of the CNF-containing and CNF-free Cu-Ce-Zr mixed oxide
catalysts as function of reaction temperature under the CO/H2O/N2 (25/125/350
Nml/min) reactant mixture.
Figure 6. Short-term deactivation of the CNF-containing and CNF-free Cu-Ce-Zr
mixed oxide catalysts at 300 °C under the CO/H2O/N2 (25/125/350 Nml/min) reactant
mixture. All curves are normalized with the initial CO conversion, calculated as the
average of the first three analyses obtained at of 3 min. intervals (CNF00: 63.77 %,
CNF07: 56.92 %, CNF13: 60.87 %, CNF24: 51.75 %).
Figure 7. WGS activity of the CNF-containing and CNF-free Cu-Ce-Zr mixed oxide
catalysts as function of reaction temperature under the CO/H2O/CO2/H2/N2
(25/125/60/175/115 Nml/min.) reactant mixture.
19
Figure 8. Arrhenius plots for the CO/H2O/CO2/H2/N2 (25/125/60/175/115 Nml/min.)
reactant mixture in the temperature range 225 – 275 °C with all CO conversions lower
than 15 %.
Figure 9. Short-term deactivation of the CNF-containing and CNF-free Cu-Ce-Zr
mixed oxide catalysts at 350 °C under the CO/H2O/CO2/H2/N2 (25/125/60/175/115
Nml/min.) reactant mixture. All curves are normalized with the initial CO conversion,
calculated as the average of the first three analyses obtained at of 3 min. intervals
(CNF00: 49.20 %, CNF13: 47.85 %, CNF24: 41.81 %).
20
Figure 1
A)
B)
21
Figure 2
22
Figure 3
20 30 40 50 60
CNF00
CNF07
CNF13
CNF24
P-CNF
(111) (200) (220) (311)
(002)
23
Figure 4
A)
B)
24
Figure 5
0
10
20
30
40
160 180 200 220 240 260
Temperature [ °C ]
CO
con
vers
ion
[ % ]
CNF00
CNF07
CNF13
CNF24
25
Figure 6
0
0,2
0,4
0,6
0,8
1
1,2
0 3 6 9 12 15
TOS [ h ]
Nor
mal
ized
CO
con
vers
ion
[ - ]
CNF00
CNF07
CNF13
CNF24
26
Figure 7
0
10
20
30
40
50
160 200 240 280 320 360
Temperature [ °C ]
CO
con
vers
ion
[ % ]
CNF00
CNF13
CNF24
27
Figure 8
1
1,5
2
2,5
3
0,0018 0,00185 0,0019 0,00195 0,002 0,00205
1/T [ 1/K ]
ln(C
O c
onv.
) [ -
]
CNF00
CNF13
CNF24
28
Figure 9
0
0,2
0,4
0,6
0,8
1
1,2
0 2 4 6 8 10 12 14 16
TOS [ h ]
Nor
mal
ized
CO
con
vers
ion
[ - ]
CNF00
CNF13
CNF24
29
Table 1. Structural characteristics of the dried and calcined CNF and Cu-Ce-Zr MMO
nanocomposites.
BET Pore volume XRD crystal sizea
[ m2/g ] [ cm3/g ] [ nm ]
P-CNF 172 0.326
CNF00 164 0.317 2.4 (3.4)
CNF07 n.d. n.d. 2.8
CNF13 164 0.307 2.8
CNF24b 139 0.262 3.1 (3.1)
a Determined on the dried samples. The values in parenthesis correspond to calcined
samples: CNF00 at 350 °C, placed directly into the hot furnace; CNF24 at 2 °C/min
up to 250 °C. b This sample was prepared with a different batch of platelet CNF, but with similar
structural properties after oxidation treatment: BET 155 m2/g, pore volume 0.297
cm3/g.
30
Table 2. Comparison of WGS activity between the nanocomposites and the CNF-free
catalyst for the CO/H2O/N2 (25/125/350 Nml/min) reactant mixture. The CO conversion
of the nanocomposites is normalized with the MMO weight content and divided by the
CO conversion of the CNF-free catalyst. The activation energies (Ea) determined by the
integral method in the temperature range 165 – 210 °C are also given.
CNF00 Rel. CO conv. normalized with metal oxide content
Temperature CO conv. CNF07/CNF00 CNF13/CNF00 CNF24/CNF00
[ °C ] [ % ] [ - ] [ - ] [ - ]
165 10.46 0.95 1.22 1.01
180 14.25 0.95 1.22 1.07
Ea [ kJ/mole ] 32 31 31 34
(165 – 210 °C)
31
Table 3. Comparison of WGS activity between the nanocomposites and the CNF-free
catalyst for the CO/H2O/CO2/H2/N2 (25/125/60/175/115 Nml/min.) reactant mixture.
The CO conversion of the nanocomposites is normalized with the MMO weight content
and divided by the CO conversion of the CNF-free catalyst. The activation energies (Ea)
determined by the differential method from an Arrhenius plot in the temperature range
225 – 275 °C are also given.
CNF00 Rel. CO conv. normalized with metal
oxide content
Temperature CO conv. CNF13/CNF00 CNF24/CNF00
[ °C ] [ % ] [ - ] [ - ]
225 6.23 1.19 1.10
250 8.97 1.30 1.27
275 9.66 1.31 1.25
Ea [ kJ/mole ] 33 38 39
(225 – 275 °C)
Paper VI
The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide
catalysts for water-gas shift
Manuscript in preparation.
1
The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-
gas shift
Florian Huber1, John Walmsley2, Hilde Venvik1,*, Anders Holmen1
1Department of Chemical Engineering, Norwegian University of Science and
Technology (NTNU), N-7491 Trondheim, Norway 2SINTEF Materials and Chemistry, N-7465 Trondheim, Norway
* Corresponding author; E-mail: [email protected], Tel: +47-73592831,
Fax: +47-73595047
Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts were prepared by homogeneous co-
precipitation with urea. Pt was wet-impregnated on mixed oxide catalysts. The WGS
activity of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts can be related to the Cu
reducibility in these catalysts. Low-temperature reducibility correlates with low-
temperature activity. Pre-reduction is not absolutely necessary when performing the
WGS reaction above the Cu reduction temperature of the catalyst. An adequate
reduction procedure may, however, be applied in order to optimize the CO conversion.
Pt had no significant effect on the Cu-Ce-Zr mixed oxide catalyst, but altered the
properties of the Cu-Zn-Al mixed oxide catalyst. Pt shifted the Cu reduction to lower
temperatures, indicating the existence of an interaction between Pt and Cu in the
bimetallic catalyst. The effect of Pt on the WGS activity and stability was dependent on
the pre-treatment procedure as well as the reaction conditions.
KEY WORDS: homogeneous alkalinization; urea; Cu-Ce-Zr; Cu-Zn-Al; mixed oxide;
bimetallic; platinum; pre-reduction; Cu reducibility; water-gas shift.
1. Introduction
The water-gas shift (WGS) reaction (CO + H2O ↔ CO2 + H2) is one of the oldest
catalytic processes employed in the chemical industry [1]. It is an important step in the
industrial production of hydrogen or synthesis gas. The role of WGS is to enhance the
2
production of hydrogen and remove CO before ammonia synthesis, refinery
hydroprocesses and bulk storage and redistribution of hydrogen, or to adjust the H2/CO
ratio in the production of methanol and the Fischer-Tropsch synthesis. The WGS
reaction has received renewed interest as a key step in fuel processing to optimise
hydrogen production and reduce the CO level for proton exchange membrane fuel cell
(PEMFC) applications [2-4].
Copper-based catalysts are considered as the most active for WGS [1,6]. Cu-Zn-Al
mixed oxides are classical catalyst formulations for WGS [2,7,8], while Cu-Ce or Cu-
Ce-Zr mixed oxides represent more recently studied systems [9-11]. The Cu-Ce-based
catalysts are claimed to be stable at high temperatures and applicable without activation
in a reducing gas such as hydrogen [9,10].
Supported bimetallic catalysts containing Cu and a noble metal, such as Pd or Pt, have
also been studied for the WGS reaction [12-14], in addition to CO oxidation [15-18],
methanol steam reforming [19], NOx reduction [15,20] and nitrile hydrogenation [21].
Bimetallic formulations can be used to tailor the reactivity of catalysts, governed by
electronic, ensemble and/or geometric effects [22,23].
In the present study, Cu-Zn-Al and Cu-Ce-Zr mixed metal oxide catalysts were
prepared by co-precipitation with urea as base precursor. A part of the precipitated
material was impregnated with Pt. The Pt-impregnated Cu-Zn-Al and Cu-Ce-Zr mixed
oxides were studied under WGS conditions and compared to the unimpregnated Cu-Zn-
Al and Cu-Ce-Zr systems, to investigate the effect of Pt on the catalyst performance.
2. Experimental
2.1. Synthesis
Cu(II)-nitrate-trihydrate (≥ 99 %), Ce(III)-nitrate-hexahydrate (≥ 99.5 %) and Zr(IV)-
nitrate-hydrate (≥ 99.5 %) purchased from Acros Organics, Zn(II)-nitrate-hexahydrate
(≥ 99 %) purchased from Fluka and Al(III)-nitrate-nonahydrate (≥ 98.5 %) purchased
from Riedel-de Haën were used as metal precursors. Tetraamineplatinum(II)-nitrate
(99.995 %) was purchased from Aldrich Chem. Co. Urea (≥ 99.5 %) was purchased
3
from Merck. Ethanol was purchased from Arcus. All chemicals were used as-received.
Deionized water was used for all catalyst preparations.
In a typical synthesis of Cu-Ce-Zr MMO, a total amount of 0.21 mole of Cu-, Ce- and
Zr-salt (nominal composition: Cu:Ce:Zr = 0.23:0.54:0.23) was dissolved in 1000 ml of
water at ca. 30 °C under stirring. 0.904 mole of urea was added to the mixture, resulting
in a pH of approx. 1.7 of the mixture. The mixture was poured into a 1000-ml 5-neck
glass flask, placed in a hot oil bath and rapidly heated up to 95 °C under vigorous
stirring with a blade agitator at approx. 750 rpm, while monitoring the temperature with
a thermometer immersed into the solution.
The setup was operated in an open mode, i.e. water was allowed to evaporate through
the unused flask necks, and the loss was compensated by continuous refilling with water
(at ambient temperature, not pre-heated). The addition of water was conducted in the
same way in all preparations, but variations in this manual procedure may influence the
formation of the precipitate and thereby the final material properties. The investigation
of this parameter was not part of the study.
The mixture was kept in the oil bath for a total period of 8 hours, including heating (ca.
20 min), precipitation and aging. The resulting suspension was then removed and
rapidly cooled down to room temperature using cold water. The solid precipitate was
filtered off and washed with deionized water and ethanol. The precipitate was then
divided into two equal parts. One part was dried for about 12 h at 100 °C and calcined
for 1 h at 350 °C (heating at 2 °C/min; cooling at approx. 2 °C/min) in a muffle furnace.
The other part was redispersed in 50 ml water at ambient temperature. 0.8 mmole of Pt-
salt was dissolved in 20 ml water and added dropwise to the redispersed precipitate
under stirring. The suspension was stirred for 1 h at ambient temperature. The
temperature was then increased to 70 °C, evaporating most of the water within approx.
1.5 h. The precipitate was then dried for about 14 h at 100 °C and calcined for 1 h at 350
°C in a muffle furnace (heating at 2 °C/min; cooling at approx. 2 °C/min).
4
The Cu-Zn-Al MMO catalysts were prepared in the same way as the Cu-Ce-Zr MMO,
with a total amount of 0.21 mole of Cu-, Zn- and Al- salt (nominal composition:
Cu:Zn:Al = 0.23:0.54:0.23). The initial pH of the salt-urea mixture was approx. 3.4. 0.8
mmole of Pt-salt was added to one half of the precipitate.
The four copper-containing catalysts are denoted as CeZr, CeZr-Pt, ZnAl and ZnAl-Pt.
2.2. Characterization techniques
Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) was used to
determine the actual metal content in the samples after calcination of the precipitated
materials. The samples were dissolved in acid before analysis without any visible
residues.
Nitrogen adsorption-desorption isotherms were obtained using a Micrometrics
TriStar 3000 instrument. The data were collected at -196 °C. The BET surface area
was calculated by the BET equation in the relative pressure interval ranging from 0.01
to 0.30. The pore volume was estimated by the Barrett-Joyner-Halenda (BJH) method
[24] as the adsorption cumulative volume of pores between 1.7 nm and 300.0 nm width.
This method is based on the assumption of cylindrical pores, and the capillary
condensation in the pores is taken into account by the classical Kelvin equation. The
pore size distributions were calculated by non-local density functional theory (NLDFT,
original DFT model with N2 [25], DFT Plus software package [26]), assuming a slit-like
pore geometry.
X-ray diffraction spectra (XRD) were recorded on a Siemens diffractometer D-5005
using dichromatic CuKα+β-radiation.
Transmission electron microscopy (TEM) images were recorded on a JEOL 2010F
transmission electron microscope. Small amounts of catalyst sample were placed in
sealed glass containers containing ethanol and immersed in an ultrasonic bath for a
couple of minutes to disperse the individual particles. The resulting suspension was
dropped onto a holey carbon film, supported on a titanium mesh grid, and dried.
5
Conventional TEM images were recorded onto a CCD camera. Samples were also
examined in scanning transmission electron microscope (STEM) mode, with a nominal
probe size of approx. 0.7nm. Bright and dark field STEM images were acquired. Energy
dispersive x-ray spectroscopy (EDS) analysis and mapping were performed using an
Oxford Instruments INCA system. Drift compensation was employed to correct for
movement of the sample during the time taken for the acquisition of maps.
Temperature-programmed reduction (TPR) experiments were performed with
CHEMBET 3000 from Quantachrome Instruments, based on the volumetric principle
with the gas flow being analyzed by a thermal conductivity detector (TCD). The catalyst
samples, 0.065-0.066 g and 0.133-0.137 g for Cu-Zn-Al- and Cu-Ce-Zr-based catalysts,
respectively, were loaded into a quartz reactor and held in place by quartz wool. The
samples were heated in 5 vol-% O2/Ar at 5 °C/min to 260 °C, and kept there for 10 min.
After cooling down to ambient temperature and flushing with Ar, the samples were then
heated in 7 vol-% H2/Ar at 2 °C/min.
2.3. Reaction experiments
The water-gas shift activity and stability were measured at atmospheric pressure in an
externally heated tubular fixed-bed reactor made of quartz. Quartz tubes with inner
diameters of 10 mm and 6 mm were used for the Cu-Zn-Al- and Cu-Ce-Zr-based
catalysts, respectively. The particle size fraction used for the experiments was 125 – 200
µm and 50 – 125 µm for the Cu-Zn-Al- and Cu-Ce-Zr-based catalysts, respectively.
Quartz wool was applied below and above the catalyst layer. A carbonyl trap was
connected to the outlet of the CO gas cylinder to remove possible iron carbonyls. The
gas flows were controlled by mass flow controllers. Water was dosed from a He-
pressurized cylinder using a liquid flow controller. The water was injected into a
vaporizing unit, and the steam was then mixed into the gas stream. A microchannel heat
exchanger (Forschungszentrum Karlsruhe) operated with air was used for condensation
of liquid products prior to analysis of the dry product gas with an on-line Agilent
G2891A Micro GC equipped with thermal conductivity detection (TCD). The gas
composition in the dry gas was quantified by using a range of certified calibration gases
as external standards.
6
The catalysts were studied under reaction conditions with and without pre-reduction.
Pre-reduction was conducted in 10 vol-% H2 in nitrogen at 250 °C for 1 h (total flow:
300 Nml/min; heating rate: 5 °C/min). The catalyst sample was then cooled in the H2/N2
atmosphere to the starting temperature of the activity measurements. When no pre-
reduction was carried out, the catalyst sample was heated to 250 °C (heating rate: 5
°C/min) in 300 Nml/min air and kept there for 45 min. The catalyst sample was then
flushed with 300 Nml/min N2 5.0 for 15 min and cooled down to the starting
temperature of the activity measurements.
The water-gas shift reaction was studied under two different reactant mixtures. All four
catalysts were subjected to a simple WGS reactant mixture containing 25 Nml/min CO,
125 Nml/min H2O and 350 Nml/min N2. The amount of catalyst used for these
experiments was 0.051 g. ZnAl and ZnAl-Pt were also studied under a simulated
reformate product mixture containing 25 Nml/min CO, 125 Nml/min H2O, 60 Nml/min
CO2, 175 Nml/min H2 and 115 Nml/min N2, using 0.110 g of catalyst. An initial
stabilization of the feed was carried out in a by-pass line. The temperature was
increased in steps of 15 – 25 °C during the activity measurements, and stabilized for 15
– 30 min. before four GC analyses were taken at three minute intervals. The short-term
deactivation behaviour of the catalysts was recorded at the final temperature.
The conversion of CO (XCO) for the feed containing CO, H2O and balance N2 only can
be calculated from the CO and CO2 concentrations in the dry exit gas:
2
2
100%( )CO
COXCO CO
= ⋅+
(1)
For the simulated reformate feed that also contains H2 and CO2, the calculated CO
conversion must take into account the initial composition of the feed gas:
0
2 2,0
( ) 100%( )CO
CO COXCO CO CO
−= ⋅+ −
(2)
7
The conversion of CO as a basis for comparison of the catalysts with and without Pt
impregnated can be justified, since equal amounts of catalyst were used and the mass
difference introduced by Pt is relatively small. Furthermore, the turnover frequency is
not easily defined or determined in such systems.
The selectivity was 100% in all experiments. Trace CH4 was detected in some of the
temperature scans as well as the following short-term deactivation experiment. The
standard deviation of the CO conversion was estimated to 4 %. The carbon balance for
CO and CO2 was always within ±1 %.
3. Results and discussion
3.1. Catalyst characterization
Table 1 shows chemical composition and structural parameters of the four catalysts. The
actual composition of the Cu-Zn-Al-based catalysts (Cu:Zn:Al = 0.25:0.53:0.22 for
ZnAl) is close to the nominal composition (Cu:Zn:Al = 0.23:0.54:0.23), while for the
Cu-Ce-Zr-based catalysts the actual composition (Cu:Ce:Zr = 0.19:0.62:0.19 for CeZr)
deviates somewhat from the nominal one. The Pt-content is approx. 1 wt% in CeZr-Pt
and 2 wt% in ZnAl-Pt, and hence the total molar metal fraction approx. 0.8 % and 1.0 %,
respectively.
The Cu-Ce-Zr- and Cu-Zn-Al-based catalysts have been prepared under identical
conditions, but the Cu-Ce-Zr-based systems exhibit significantly higher surface areas
and pore volumes than the Cu-Zn-Al-based systems (Table 1). This indicates that the
two systems exhibit different precipitation behaviour and/or behaviour under heat
treatment (drying/calcination). Adjusted preparation conditions have to be applied if
optimization of the surface area of the Cu-Zn-Al mixed oxide system is desired.
The N2 adsorption-desorption isotherms and the corresponding pore size distributions
for the four catalysts are shown in Figure 1. The majority of the pores in all four
catalysts lies in the mesoporous range (2-50 nm). The Cu-Ce-Zr-based systems show a
broad unimodal pore size distribution with a maximum at around 25 nm, while the Cu-
8
Zn-Al-based systems show a bimodal distribution with a pronounced maximum at
around 8 nm and a local maximum at around 28 nm.
The catalyst bed density was estimated to approximately 0.3 g/cm3 and 0.9 g/cm3 for
ZnAl and CeZr, respectively. CeZr thus had a bed density three times higher than ZnAl.
The impregnation step for deposition of Pt on the precipitates before drying resulted in a
decrease of the pore volume, and in case of CeZr-Pt also the surface area (Table 1). The
change in pore volume is more pronounced in the Cu-Ce-Zr- than in the Cu-Zn-Al-
based system, probably due to the larger surface area and pore volume, and hence more
fragile pore structure of the former. As can be seen in Figure 1B, the impregnation of
the Cu-Ce-Zr precipitate with Pt resulted in a break-down of pores larger than 25 nm,
reflected in a significant pore volume decrease.
Figure 2 shows the XRD spectra of the four catalysts after calcination. The Cu-Ce-Zr-
based systems exhibit the fluorite-type structure of ceria and zirconia. No additional
CuO phase is visible. The spectrum of CeZr-Pt does not show any sign of crystalline Pt
or Pt-oxide. The Cu-Zn-Al-based systems show reflections assignable to CuO, ZnO and
Al2O3, and possibly mixed oxide structures of these elements. The spectrum of ZnAl-Pt
shows weak reflections that may be assigned to crystalline Pt and/or PtO2, as compared
to the spectrum of ZnAl.
Cu-Ce-Zr mixed oxide catalysts have been investigated with TEM and STEM-EDS in
an earlier study [27], and will therefore not be discussed in detail in this work. In short,
the Cu-Ce-Zr mixed oxides exhibit a good distribution of all three metals, but without
forming a true solid solution. The distribution of Pt over the Cu-Ce-Zr mixed oxide in
CeZr-Pt was investigated with STEM-EDS, finding Pt distributed over the regions
studied with local variations in concentration (not shown). Figure 3 shows TEM images
of ZnAl-Pt. The bright, dense particles in Figure 3A and B are rich in Zn, with a
Cu:Zn:Al ratio of 1:8-9:1-2. The grey, porous areas are closer to the overall composition
as determined by ICP-AES, with a typical Cu:Zn:Al ratio of 1: 2-4:2-3. The overall
Cu:Zn:Al ratio in Figure 3A is 1:3.9:1.7. The particles in Figure 3C mainly contain Cu
9
and Zn, with Al present only in small quantities. Cu and Zn are not homogeneously
distributed in this region, since both Cu- and Zn-rich areas are found. The overall
Cu:Zn:Al ratio in Figure 3C is 1:0.7:0.02. Different preparation conditions and metal
ratios have to be applied if an optimization of the metal distribution of the Cu-Zn-Al
mixed oxide catalyst is desired..
The bimodal pore size distribution of the Cu-Zn-Al-based catalysts, shown in Figure 1B,
can be explained with the TEM images in Figure 3. The pores with a size around 8 nm
are probably located in the grey, porous areas, while the pores with a size around 28 nm
are probably built up by the bright, dense particles. The distribution of Pt over the Cu-
Zn-Al mixed oxide in ZnAl-Pt was investigated with STEM-EDS. The analysis showed
that Pt was distributed over the regions studied with local variations in concentration.
The TPR profile of the four catalysts is shown in Figure 4. The Cu-Ce-Zr-based
catalysts are reduced at lower temperatures than the Cu-Zn-Al-based catalysts, with the
maximum signal around 160 °C. In addition, CeZr and CeZr-Pt exhibit a small signal
already below 100 °C, a feature not present in the Cu-Zn-Al-based catalysts. The
addition of 1 wt-% Pt does not have a large effect on the reducibility of the Cu-Ce-Zr
mixed oxide catalyst. It slightly enhances the reduction at lower temperatures. ZnAl
exhibits a broad reduction peak between 180 and 280 °C. ZnAl-Pt exhibits a broad
reduction peak of different shape than ZnAl between 135 and 235 °C. The addition of 2
wt-% Pt to the Cu-Zn-Al mixed oxide clearly shifts the Cu reduction to lower
temperatures, possibly due to hydrogen spillover [28,29] from Pt to CuO, thus
indicating a significant interaction between Pt and Cu in the catalyst. Huang & Sachtler
[21] reported that the addition of Ru, Rh, Pd or Pt to NaY-supported Cu catalysts
enhanced the reduction of Cu2+.
3.2. Catalyst activity and stability
3.2.1. Cu-Ce-Zr-based mixed oxide catalysts
Table 2 summarizes the conversion of CO obtained for CeZr and CeZr-Pt under the
simple WGS reactant mixture between 165 and 195 °C. Both catalysts demonstrated
WGS activity without pre-reduction, and the activity did not increase significantly upon
10
pre-reduction. Hence, Cu-Ce-Zr mixed oxide catalysts do not require pre-reduction, but
the CO conversion can be optimized through an adequate pre-reduction procedure.
CeZr-Pt exhibited a somewhat lower CO conversion than CeZr. This may be related to
the lower surface area (Table 1). A slight negative impact of Pt on the reactivity can,
however, not be excluded, but Pt did not significantly alter the WGS activity of the Cu-
Ce-Zr mixed oxide catalyst under the given conditions. Hungría et al. [15] reported that
the catalytic performance of a Cu-Ce-Zr mixed oxide supported on alumina and
impregnated with 1 wt% Pd was governed by the Cu-(Ce,Zr)Ox character of its active
sites, when used for CO oxidation and NO reduction. In a similar way, the catalytic
activity of CeZr-Pt may be dominated by active Cu sites that are not significantly
altered by Pt under the given conditions, as also reflected in the reducibility (Figure 4).
This slight shift in reducibility to lower temperatures may, however, be correlated with
a lower impact of the pre-reduction of CeZr-Pt as compared to CeZr. Pt appears to make
the WGS activity of Cu-Ce-Zr mixed oxide catalyst more insensitive to the pre-
treatment.
Both catalysts, CeZr and CeZr-Pt, were more stable without than with pre-reduction, as
can be seen in Figure 5. The short-term stability was studied at 250 °C under the simple
WGS reactant mixture. Under these conditions, the not pre-reduced catalysts display a
somewhat lower CO conversion, 80 % and 95 % of the pre-reduced catalysts for CeZr
and CeZr-Pt, respectively.
3.2.2. Cu-Zn-Al-based mixed oxide catalysts
Figure 6 shows the CO conversion of ZnAl and ZnAl-Pt as function of reaction
temperature under the CO/H2O/N2 reactant mixture. In contrast to the Cu-Ce-Zr mixed
oxide catalysts, the Cu-Zn-Al mixed oxide catalyst ZnAl exhibited a different behaviour
with and without pre-reduction. Without pre-reduction, ZnAl did not show WGS
activity at 165 °C and then ignited between 180 and 195 °C, reaching more or less the
activity of the pre-reduced sample. The higher CO conversion of the not pre-reduced
catalyst in the range 195 – 220 °C as compared to the pre-reduced can be ascribed to a
contribution from Cu reduction by CO. This conclusion was reached through
comparison of the CO and H2 concentrations detected in the dry exit gas.
11
The reactivity of the not pre-reduced catalyst can be further explained with the TPR
profile of ZnAl in Figure 4, neglecting the impact of the type of reduction gas used. The
reduction of Cu in ZnAl started at about 180 °C and correlates thus with the onset of the
CO conversion, indicating the importance of reduced Cu for the WGS activity of the
catalyst. The TPR profile of CeZr exhibited a maximum reduction peak at around 160
°C and correspondingly showed a significant WGS activity already at 165 °C (Table 2).
A low Cu reduction temperature is therefore a prerequisite for WGS activity at low
temperature. Ko et al. [30] reported lower Cu reduction temperatures and higher
conversions in the WGS reaction for Cu-Zr relative to Cu-Zn-Al mixed oxide catalysts
at low temperatures. Tang et al. [31] stated that the redox ability of Cu-Ce mixed oxide
catalysts at low temperatures plays an essential role for the catalytic activity in CO
oxidation, and that the redox properties are determined by the dispersion of the copper
species and the degree of interaction between these species and ceria. Shen & Song [32]
reported that Cu-Zn-Al mixed oxide catalysts with lower Cu reduction temperature
showed higher activity for production of hydrogen from methanol steam reforming.
Above 210 °C, the CO conversion of not pre-reduced ZnAl was slightly lower than of
the pre-reduced ZnAl. Above the onset temperature for Cu reduction, pre-reduction is
not absolutely necessary, but as for Cu-Ce-Zr mixed oxides, an appropriate reduction
procedure may enhance catalyst activity and stability [33-42] and ensure avoidance of
hot spots (especially in fixed bed reactors) [43].
The addition of Pt to the Cu-Zn-Al mixed oxide in ZnAl-Pt resulted in an increased
reactivity of the not pre-reduced sample below 195 °C as compared to not pre-reduced
ZnAl (Figure 6). This can also be explained by the TPR profiles of both catalysts
(Figure 4) and neglecting the impact of the type of reduction gas used. Pt shifted the Cu
reduction to lower temperatures with a certain amount of Cu being reduced at 165 °C.
Above 195 °C, the not pre-reduced ZnAl-Pt showed a similar CO conversion as
compared to not pre-reduced ZnAl.
ZnAl-Pt without pre-reduction exhibits a similar initial WGS activity as the Cu-Zn-Al
mixed oxide catalyst ZnAl (Figure 6). The pre-reduced ZnAl-Pt sample, however,
showed a significantly lower activity than both ZnAl and not pre-reduced ZnAl-Pt.
12
Utaka et al. [14] also observed a decreased activity of pre-reduced, Pt-impregnated Cu-
Zn-Al mixed oxide catalysts for oxygen-assisted WGS as compared to pre-reduced Cu-
Zn-Al. Huang & Sachtler [21] reported that the addition of Ru, Rh, Pd or Pt to NaY-
supported Cu catalysts lowered the activity for acetonitrile hydrogenation.
Figure 7 shows the short-term stability of ZnAl and ZnAl-Pt at 250 °C. ZnAl and ZnAl-
Pt without pre-reduction exhibited a similar normalized deactivation behaviour (Figure
7B). The pre-reduced ZnAl-Pt sample, however, showed a completely different trend.
After a slight increase in the first 10 h on stream, the CO conversion remained constant
for the following 15 h on stream. The initial increase in CO conversion may be related
to structural changes and the formation of a slightly more active and stable phase under
the given reaction conditions.
The type of pre-treatment procedure appears to have a significant effect on initial
activity and stability of supported, bimetallic Pt-Cu catalysts. Epron et al. [44] also
report that the pre-treatment significantly affects the interaction between Pt and Cu in an
Al2O3-supported, bimetallic Pt-Cu system, and the catalytic activity for nitrate reduction
in water. The origin for the decreased activity and enhanced stability of the pre-reduced
ZnAl-Pt, as compared to ZnAl may be related to surface alloy effects [22,45-50]. Zhou
et al. [20] reported a higher activity and stability of Pt-on-Cu shell-core nanoparticles in
NO reduction compared to Cu-on-Pt shell-core particles and Pt-Cu alloy particles.
Persson et al. [51] reported a lower initial activity and increased stability of an Al2O3-
supported, bimetallic Pd-Pt catalyst for catalytic combustion of CH4, as compared to
Pd/Al2O3. The decreased initial activity and enhanced stability of this catalyst was
related to a partial formation of a Pd-Pt alloy. Besenbacher et al. [52] reported on a
supported Au-Ni surface alloy catalyst for steam reforming being less active and more
stable than a supported Ni catalyst, and ascribed it to a change in the surface reactivity
of Ni by Au.
The WGS activity and stability of pre-reduced ZnAl and ZnAl-Pt were also studied
under the simulated reformate product mixture, as shown in Figure 8. Pre-reduced
ZnAl-Pt exhibits lower CO conversion than pre-reduced ZnAl also under these reaction
13
conditions (Figure 8A). The stability as function of time on stream of both catalysts at
300 °C is indicated in Figure 8B. ZnAl shows the typical deactivation behaviour of Cu-
Zn-Al mixed oxide catalysts. ZnAl-Pt shows an increase in the CO conversion during
the first 6 h on stream, as observed under the simple WGS reactant mixture at 250 °C.
The CO conversion then decreases and approaches the conversion of ZnAl. A possible
explanation would be that the structure responsible for the high stability at 250 °C under
the simple WGS reactant conditions also existed initially at 300 °C under the simulated
reformate feed conditions but is not stable over a longer period under these reaction
conditions. Hungría et al. [15] reported that the Pd-Cu alloy formed in an Al2O3-
supported Pd-Cu/(Ce,Zr)Ox catalyst for NO reduction was gradually degraded under
reaction conditions at temperatures above 230 °C. Epron et al. [44] reported on
segregation of Pt and Cu and hence diminishing interaction at 400 °C, under reducing as
well as oxidizing conditions in an Al2O3-supported, bimetallic Pt-Cu catalyst.
3.3. Important aspects on bimetallic Pt-Cu
The change in reactivity of bimetallic surfaces is generally attributed to electronic,
ensemble and/or geometric effects [22,23,53].
Pt has a lattice constant that is approx. 9 % larger than the one of Cu. A pseudomorphic
Pt overlayer on a Cu substrate is therefore compressed, since the individual Pt atoms
have less space and thus interact/overlap more with surrounding neighbour states. This
results in a broader d-band and a downshift of the d-band center, away from the Fermi
level. Consequently, the binding energy/strength of adsorbates such as CO on the
surface decreases [45,53]. A decrease in the CO adsorption energy relative to pure Cu
may then result in a decreased WGS turnover rate of the bimetallic catalyst [54]. The
reactivity of the bimetallic surface may depend on the Pt coverage. Regarding the
supported Au-Ni surface alloy catalyst for steam reforming, it was shown that the
dissociation probability of CH4 decreases with increasing Au coverage on Ni [52,53]. In
analogy with Pd on Cu [22], real pseudomorphic Pt overlayers may not exist on Cu
because of the strong compression of the Pt layer by the Cu substrate lattice. Pt probably
forms a surface alloy with Cu [23,49]. Consequently, the above mentioned trends
14
derived from pseudomorphic overlayer systems are more of a qualitative rather than a
quantitative nature [22,53].
The thermodynamic behaviour of bimetallic systems is an important issue when
discussing surface reactivity. Two important quantities are the segregation energy and
the mixing energy [53]. Pt on Cu has a small negative surface segregation energy [50,53]
and a positive mixing energy [49,53], thus forming alloys in the surface layers of the Cu
lattice without significant Pt migration into the Cu bulk when considering energetics
without taking into account the temperature-dependent impact of entropy [53]. The
temperature may, however, also play an important role. With increasing temperature,
the entropy may eventually become dominant resulting in Pt migration into the bulk [53]
or destruction of the interactions between Cu and Pt [44]. The effect of the surrounding
gas may also have an important effect on the surface composition of the bimetallic
system [53]. The dependence of the catalyst activity on the pre-treatment conditions,
observed in this study and by Epron et al. [44], may be an indication for that.
While the decreased activity of the pre-reduced ZnAl-Pt, as compared to ZnAl, may be
explained by a d-band shift, the reason for the improved stability is not clear. Further
studies on the deactivation behaviour of Cu-Zn-Al mixed oxide catalysts as well as the
impact of Pt are necessary. The improved stability of a supported, bimetallic Au-Ni
steam reforming catalyst, as compared to a supported Ni catalyst, was explained with a
decrease in the coke formation rate [52]. For the bimetallic Pt-Cu catalyst, the improved
stability may be related to the suppression of gradual deterioration of active sites by
secondary chemical processes [35], in analogy with Au-Ni. It is known that the structure
of Cu particles/surface depends on the reaction environment [37,41]. A possible
question is therefore, whether reduced, active copper sites undergo gradual degradation
processes, possibly including sintering, Cu-Zn alloy formation and formation of surface
spinel species [33,38], partial reoxidation or degradation by H2O or CO2 [43,55-58], and
whether Pt could suppress these processes. The mechanism behind regeneration of
deactivated Cu-based catalysts by oxidation and subsequent reduction [35] is also of
significance in this respect.
15
4. Conclusions
Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts were prepared by homogeneous co-
precipitation with urea. While the Cu-Ce-Zr mixed oxide catalyst exhibited a high
surface area and a good distribution of all three metals, the Cu-Zn-Al mixed oxide
catalyst exhibited a relatively low surface area and an inhomogeneous distribution of the
three metals under the preparation conditions used. For the Cu-Ce-Zr mixed oxide
catalyst, the reduction of Cu could be achieved at lower temperatures than for the Cu-
Zn-Al mixed oxide. Improvement of the surface area, metal distribution and Cu
reducibility in the Cu-Zn-Al mixed oxide catalyst requires therefore different
preparation conditions.
Cu-Ce-Zr mixed oxide catalysts showed WGS activity without pre-reduction under the
reaction conditions used. The not pre-reduced sample was somewhat less active but
more stable than the pre-reduced sample. Pre-reduction is therefore not absolutely
necessary, but an adequate pre-reduction procedure may be applied to optimize the CO
conversion. Pt impregnated on the Cu-Ce-Zr mixed oxide had no significant effect on
Cu reducibility as well as WGS activity and stability.
The WGS activity of the Cu-Zn-Al mixed oxide catalyst correlated with the Cu
reducibility. Below the temperatures required for Cu reduction, the not pre-reduced
sample did not show significant WGS activity. At temperatures where Cu is at least
partly reduced, the catalyst showed significant WGS activity. Above the temperatures
required for partial Cu reduction, the not pre-reduced sample exhibited a somewhat
lower CO conversion than the pre-reduced sample, under similar short-term stability.
Consequently, pre-reduction is not absolutely necessary at these reaction temperatures,
but an adequate pre-reduction procedure may improve the CO conversion.
Pt impregnated on the Cu-Zn-Al mixed oxide catalyst shifted the Cu reduction and
hence also the WGS activity of the unreduced sample to lower temperatures. The not
pre-reduced sample showed similar WGS activity and stability, as compared to the Cu-
Zn-Al mixed oxide catalyst. The pre-reduced sample, however, exhibited a lower CO
conversion and a better stability at 250 °C than the Cu-Zn-Al mixed oxide catalyst
16
under CO/H2O/N2-feed conditions, indicating the existence of an interaction between
Cu and Pt in the bimetallic catalyst. Under CO/H2O/CO2/H2/N2-feed conditions at 300
°C, the impact of Pt diminished and the CO conversion of the Pt-impregnated catalyst
approached the one of the Cu-Zn-Al mixed oxide catalyst with time on stream,
indicating a destruction of the interaction between Cu and Pt.
Improvement of the performance of the bimetallic Pt-Cu catalyst requires optimization
of the Pt-loading as well as further studies on appropriate temperatures and ambient gas
atmospheres in order to facilitate the interaction between Pt and Cu.
Acknowledgements
This work was supported by the Research Council of Norway through Grant No.
140022/V30 (RENERGI) and 158516/S10 (NANOMAT). Statoil ASA through the Gas
Technology Center NTNU-SINTEF is also acknowledged for their support.
References
[1] D. Andreeva, Gold Bull. 35 (2002) 82-88.
[2] J.R. Ladebeck, J.P. Wagner, in: Handbook of Fuel Cells – Fundamentals,
Technology and Applications, Vol. 3: Fuel Cell Technology and Applications,
eds. W. Vielstich, H.A. Gasteiger, A. Lamm, Wiley and Sons, New York, 2003,
pp.190-201.
[3] R. Farrauto, S. Hwang, L. Shore, W. Ruettinger, J. Lampert, T. Giroux, Y. Liu, O.
Ilinich, Annu. Rev. Mater. Res. 33 (2003) 1-27.
[4] Y. Choi, H.G. Stenger, J. Power Sources 124 (2003) 432-439.
[5] N. Schumacher, A. Boisen, S. Dahl, A.A. Gokhale, S. Kandoi, L.C. Grabow, J.A.
Dumesic, M. Mavrikakis, I. Chorkendorff, J. Catal. 229 (2005) 265-275.
[6] D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90-102.
[7] C. Rhodes, G.J. Hutchings and A.M. Ward, Catal. Today 23 (1995) 43-58.
[8] D.S. Newsome, Catal. Rev. Sci. Eng. 21 (1980) 275-318.
[9] Y. Li, Q. Fu and M. Flytzani-Stephanopoulos, Appl. Catal., B 27 (2000) 179-191.
[10] X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062.
17
[11] N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H. Ribeiro, J.
Catal. 217 (2003) 233-239.
[12] E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347-357.
[13] E.S. Bickford, S. Velu, C. Song, Preprints of American Chemical Society
Symposia, Division of Fuel Chemistry 49 (2004) 649-651.
[14] T. Utaka, K. Sekizawa, K. Eguchi, Appl. Catal. A 194-195 (2000) 21-26.
[15] A.B. Hungría, A. Iglesias-Juez, A. Martínez-Arias, M. Fernández-García, J.A.
Anderson, J.C. Conesa, J. Soria, J. Catal. 206 (2002) 281-294.
[16] R.E.R. Colen, M. Kolodziejczyk, B. Delmon, J.H. Block, Surf. Sci. 412/413
(1998) 447-457.
[17] N.N. Hoover, B.J. Auten, B.D. Chandler, J. Phys. Chem B 110 (2006) 8606-8612.
[18] P.C. Liao, J.J. Carberry, T.H. Fleisch, E.E. Wolf, J. Catal. 74 (1982) 307-316.
[19] E.S. Ranganathan, S. Bej, L.T. Thompson, Preprints of American Chemical
Society Symposia, Division of Fuel Chemistry 48 (2003) 381-382.
[20] S. Zhou, B. Varughese, B. Eichhorn, G. Jackson, K. McIlwrath, Angew. Chem.
Int. Ed. 44 (2005) 4539-4543.
[21] Y. Huang, W.M.H. Sachtler, J. Catal. 188 (1999) 215-225.
[22] A. Groβ, Top. Catal. 37 (2006) 29-39.
[23] V. Ponec, Appl. Catal. A 222 (2001) 31-45.
[24] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373-380.
[25] P.B. Balbuena, K.E. Gubbins, Fluid Phase Equilib. 76 (1992) 21-35.
[26] DFT Plus for Windows, version 3.00, User’s Manual, Micrometrics Instrument
Corporation.
[27] F. Huber, H. Venvik, M. Rønning, J. Walmsley, A. Holmen, Appl. Catal. A,
submitted.
[28] W.C. Conner, J.L. Falconer, Chem. Rev. 95 (1995) 759-788.
[29] M.H. Kim, J.R. Ebner, R.M. Friedman, M.A. Vannice, J. Catal. 208 (2002) 381-
392.
[30] J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157-161.
[31] X. Tang, B. Zhang, Y. Li, Y. Xu, Q. Xin, W. Shen, Catal. Today 93-95 (2004)
191-198.
[32] J.-P. Shen, C. Song, Catal. Today 77 (2002) 89-98.
18
[33] C.E. Quincoces, N.E. Amadeo, M.G. Gonzalez, Stud. Surf. Sci. Catal. 111 (1997)
535-541.
[34] W.F. Ruettinger, O.M. Ilinich, R.J. Farrauto, US Patent US2004082669, 2004.
[35] O.M. Ilinich, W.F. Ruettinger, R.T. Mentz, R.J. Farrauto, US Patent
US2004082471, 2004.
[36] K.-D. Jung, O.-S. Joo, S.H. Han, Catal. Lett. 68 (2000) 49-54.
[37] B.H. Sakakini, J. Tabatabaei, M.J. Watson, K.C. Waugh, J. Mol. Catal. A 162
(2000) 297-306.
[38] H. Wilmer, O. Hinrichsen, Catal. Lett. 82 (2002) 117-122.
[39] S. Fujita, S. Moribe, Y. Kanamori, M. Kakudate, N. Takezawa, Appl. Catal. A
207 (2001) 121-128.
[40] W.P.A. Jansen, J. Beckers, J.C. Van den Heuvel, A.W. Denier van den Gon, A.
Bliek, H.H. Brongersma, J. Catal. 210 (2002) 229-236.
[41] P.L. Hansen, J.B. Wagner, S. Helveg, J.R. Rostrup-Nielsen, B.S. Clausen, H.
Topsøe, Science 295 (2002) 2053-2055.
[42] A.A.G. Lima, M. Nele, E.L. Moreno, H.M.C Andrade, Appl. Catal. A 171 (1998)
31-43.
[43] C. Rhodes, G.J. Hutchings and A.M. Ward, Catal. Today 23 (1995) 43-58.
[44] F. Epron, F. Gauthard, J. Barbier, J. Barbier, Appl. Catal. A 237 (2002) 253-261.
[45] A. Ruban, B. Hammer, P. Stoltze, H.L. Skriver, J.K. Nørskov, J. Mol. Catal. A
115 (1997) 421-429.
[46] J. Greeley, M. Mavrikakis, Nat. Mater. 3 (2004) 810-815.
[47] D. Woodruff (Ed.), Surface alloys and alloy surfaces, The Chemical Physics of
Solid Surfaces Series Vol. 10, Elsevier, Amsterdam, 2002.
[48] L. Yang, Curr. Top. Catal. 2 (1999) 59-71.
[49] A. Christensen, A.V. Ruban, P. Stoltze, K.W. Jacobsen, H.L. Skriver, J.K.
Nørskov, F. Besenbacher, Phys. Rev. B 56 (1997) 5822-5834.
[50] A.V. Ruban, H.L. Skriver, J.K. Nørskov, Phys. Rev. B 59 (1999) 15990-16000.
[51] K. Persson, A. Ersson, S. Colussi, A. Trovarelli, K. Jansson, S.G. Järås, Proc. of
the 12th Nordic Symp. on Catal., Trondheim, Norway, May 2006, 80-81.
[52] F. Besenbacher, I. Chorkendorff, B.S. Clausen, B. Hammer, A.M. Molenbroek,
J.K. Nørskov, I. Stensgaard, Science 279 (1998) 1913-1915.
19
[53] J.H. Larsen, I. Chorkendorff, Surf. Sci. Rep. 35 (1999) 163-222.
[54] D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90-102.
[55] Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki, K. Eguchi, Appl. Catal. A 238 (2003)
11-18.
[56] O. Hinrichsen, T. Genger, M. Muhler, Stud. Surf. Sci. Catal. 130 (2000) 3825-
3830.
[57] (a) K.C. Waugh, Solid State Ionics 168 (2004) 327-342; (b) R.A. Hadden, H.D.
Vandervell, K.C. Waugh, G. Webb, Catal. Lett. 1 (1988) 27-33.
[58] D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107-112.
20
Figure captions
Figure 1. N2 adsorption-desorption isotherms (A) and corresponding pore size
distributions (B) of the four catalysts after calcination.
Figure 2. XRD spectra of the four catalysts after calcination.
Figure 3. TEM images of ZnAl-Pt after calcination.
Figure 4. TPR profiles of the four catalysts. The samples, 0.065-0.066 g and 0.133-
0.137 g for ZnO- and CeO2-based catalysts, respectively, were heated in 5 vol-% O2/Ar
to 260 °C at a heating rate of 5 °C/min, kept there for 10 min, cooled down to ambient
temperature and flushed with Ar, and then heated in 7 vol-% H2/Ar at a heating rate of 2
°C/min.
Figure 5. Normalized short-term deactivation of CeZr and CeZr-Pt at 250 °C under the
CO/H2O/N2 (25/125/350 Nml/min) reactant mixture.
Figure 6. Initial CO conversion of ZnAl and ZnAl-Pt as function of reaction
temperature under the CO/H2O/N2 (25/125/350 Nml/min) reactant mixture. The
catalysts (0.051 g) were studied with and without pre-reduction.
Figure 7. A) Short-term deactivation of ZnAl and ZnAl-Pt at 250 °C under the
CO/H2O/N2 (25/125/350 Nml/min) reactant mixture, with and without pre-reduction. B)
Normalized short-term deactivation of ZnAl and ZnAl-Pt. All curves are normalized
with the initial CO conversion, calculated as the average of the first three analyses
obtained at 3 min. intervals.
Figure 8. A) CO conversion of ZnAl and ZnAl-Pt as function of reaction temperature
under the CO/H2O/CO2/H2/N2 (25/125/60/175/115 Nml/min.) reactant mixture. The
catalysts (0.110 g) were pre-reduced at 250 °C for 1 h in 10 vol-% H2/N2. B) Short-term
deactivation of pre-reduced ZnAl and ZnAl-Pt at 300 °C under the CO/H2O/CO2/H2/N2
(25/125/60/175/115 Nml/min.) reactant mixture.
21
Figure 1
A)
0
100
200
300
0 0,2 0,4 0,6 0,8 1Relative pressure [-]
Nitr
ogen
ads
orbe
d [c
m3 /g
STP
] CeZrCeZr-PtZnAlZnAl-Pt
B)
0
0,004
0,008
0,012
0,016
0,02
1 10 100 1000
Pore width [nm]
Pore
vol
ume
[cm
3 /g S
TP]
CeZrCeZr-PtZnAlZnAl-Pt
22
Figure 2
0
10000
15 20 30 40 50 60 70 80
2 [ ° ]
Inte
nsity
[a. u
.]
CeZr
CeZr-Pt
ZnAl
ZnAl-Pt
Pt
PtO2
0
10000
15 20 30 40 50 60 70 80
2 [ ° ]
Inte
nsity
[a. u
.]
CeZr
CeZr-Pt
ZnAl
ZnAl-Pt
Pt
PtO2
23
Figure 3
A)
B)
100 nm100 nm100 nm
200 nm200 nm200 nm
24
C)
50 nm50 nm50 nm
25
Figure 4
-20
20
60
100
140
180
30 70 110 150 190 230 270 310
Temperature [ °C ]
Sign
al [
a. u
. ]
CeZrCeZr-PtZnAlZnAl-Pt
26
Figure 5
0,5
0,7
0,9
1,1
0 4 8 12 16
TOS [ h ]
Nor
mal
ized
CO
con
vers
ion
[ - ]
CeZr redCeZr not redCeZr-Pt redCeZr-Pt not red
27
Figure 6
0
4
8
12
16
20
24
150 170 190 210 230 250 270
temperature [ C ]
CO
con
vers
ion
[ %
]
ZnAl red
ZnAl not red
ZnAl-Pt red
ZnAl-Pt not red
28
Figure 7
A)
0
4
8
12
16
20
0 5 10 15 20 25
TOS [ h ]
CO
con
vers
ion
[ %
]
ZnAl red
ZnAl not red
ZnAl-Pt red
ZnAl-Pt not red
B)
0
0,4
0,8
1,2
0 5 10 15 20 25
TOS [ h ]
Nor
mal
ized
CO
con
vers
ion
[ - ]
ZnAl redZnAl not redZnAl-Pt redZnAl-Pt not red
29
Figure 8
A)
0
5
10
15
20
25
170 190 210 230 250 270 290 310
temperature [ C ]
CO
con
vers
ion
[ %
]
ZnAl red
ZnAl-Pt red
B)
0
6
12
18
24
0 8 16 24 32 40
TOS [ h ]
CO
con
vers
ion
[ %
]
ZnAl red
ZnAl-Pt red
30
Table 1. Composition and structural parameters of the mixed oxide catalysts.
Catalyst composition – metal content
Cu Ce/Zn Zr/Al Pt BET
asurfacevolume
pore volume a
[% (g/g-cat.)] [m2/g] [m2/mm3] [cm3/g] [cm3/cm3]
CeZr 9.4 69.4 14.0 161 1.12 0.40 2.79
CeZr-Pt 9.2 67.2 13.5 0.7 147 0.30
ZnAl 19.4 42.0 7.3 74 0.40 0.20 1.08
ZnAl-Pt 18.8 40.3 7.1 2.2 79 0.18
a Surface-to-volume ratio and pore volume in [cm3/cm3] estimated by linear
combination of tabulated densities for CeO2 (7650 kg/m3), ZrO2 (5680 kg/m3,
tetragonal phase), CuO (6310 kg/m3), ZnO (5660 kg/m3) and γ-Al2O3 (3650 kg/m3).
31
Table 2. Initial CO conversion of CeZr and CeZr-Pt under the CO/H2O/N2 (25/125/350
Nml/min) reactant mixture. The catalysts (0.051 g) were studied with and without pre-
reduction.
CO conversion [ % ]
Temperature CeZr CeZr-Pt
[ °C ] pre-reduced not pre-reduced pre-reduced not pre-reduced
165 6.47 6.39 5.14 5.76
180 9.27 8.39 7.94 8.36
195 12.53 10.55 11.10 11.10