Anda di halaman 1dari 246

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

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 Rnning 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 Haans, 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 Torbjrn Vrlstad are gratefully acknowledged for assistance with the BET device. Hilde Meland and Cathrine Brin 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, Asbjrn Lindvg, Rune Myrstad, Edd A. Blekkan, Ingrid Aartun, Bozena Silberova and Rune Ldeng 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 NTNUSINTEF is gratefully acknowledged.

Acknowledgements

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

ii

Abstract

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

iii

Abstract

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, coprecipitation 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 CuCe-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.

iv

Abstract

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-ZnAl mixed oxide catalysts can be related to the Cu reducibility in these catalysts. Lowtemperature 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

Table of contents
Acknowledgement Abstract Table of contents List of publications and presentations List of symbols and abbreviations i iii vi viii xi

1. Introduction
1.1. Hydrogen society 1.2. Fuel processing for fuel cell applications 1.3. Scope of the present work

1
1 1 3

2. Theory and literature


2.1. Water-gas shift reaction (WGS) 2.2. WGS catalysts for fuel processing 2.2.1. Catalyst requirements 2.2.2. New catalyst materials 2.3. Mixed metal oxide catalysts

5
5 8 8 9 11

3. Experimental
3.1. Catalyst preparation 3.2. Characterization of catalyst materials 3.2.1. Inductively coupled plasma atomic emission spectroscopy (ICP-AES) 3.2.2. X-ray diffraction (XRD) 3.2.3. Nitrogen adsorption-desorption isotherms 3.2.4. Thermogravimetric analysis (TGA) 3.2.5. Volumetric TPR measurements 3.2.6. Transmission electron microscopy (TEM) 3.2.7. Transmission X-ray absorption spectroscopy (XAS) 3.3. Setup for water gas shift testing

13
13 13 13 14 14 15 15 15 16 18

vi

Table of contents

4. Results and discussion


4.1. Pre-studies 4.1.1. Total pore volume and pore size distribution 4.1.2. XRD particle size estimates with Win-crysize 4.1.3. Reproducibility of the WGS testing By-pass effects 4.2. Summary of results and discussion 4.2.1. Relating catalyst structure and composition to the water-gas shift activity of Cu-Zn-based mixed-oxide catalysts (Paper I) 4.2.2. Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts (Paper II) 4.2.3. Preparation and characterization of nanocrystalline, high-surface area Cu-Ce-Zr mixed oxide catalysts from homogeneous co-precipitation (Paper III) 4.2.4. Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift (Paper IV) 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) 4.2.6. The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift (Paper VI)

23
23 23 25 26 30 30 34

38 42

43 44

5. Concluding remarks 6. Suggestions for further work


References List of appendices

47 50
51 62

vii

List of publications and presentations

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. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Relating catalyst structure and composition to the water-gas shift activity of Cu-Znbased mixed-oxide catalysts, Catalysis Today 100 (2005), 249-254. II F. Huber, Z. Yu, S. Lgdberg, M. Rnning, 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. Rnning, 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. Rnning, H. Venvik, A. Holmen, Comparison of CuCe-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 CuCe-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift, Manuscript in preparation.

viii

List of publications and presentations

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. Tangers, 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. Rnning, 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, Humlebk, Denmark, August 9 -15, 2003 and EuropaCat VI, Innsbruck, Austria, August 31 September 04, 2003. M. Rnning, 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. Rnning, 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. Rnning, 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. Rnning, 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.

ix

List of publications and presentations

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 Topse A/S, Lyngby, Denmark, Mai 10, 2006. F. Huber, H. Meland, Z. Yu, M. Rnning, 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. Rnning, 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

List of symbols and abbreviations


Latin symbols B Be BET dC dP E Ea E0 I0 It k K Keq N R stv T xp Line width at half maximum (FWHM) [ or radian] Instrumental line broadening [ or radian] BET surface area [m2/g] Size of the core of passivated metal particles [nm] Particle size estimate [nm] Energy [eV] Activation energy [kJ/mole] Energy shift to correct for deviation from theoretical edge value [eV] Incident X-ray intensity [-] Transmitted X-ray intensity [-] Wave number of the photoelectron [-1] Scherrer constant [-] Equilibrium constant [-] Coordination number [-] Distance of the central atom to the neighbouring atom (shell) [] Surface-to-volume ratio [m2/m3] Temperature [K or C] Fraction of passivated metal atoms [-]

Greek symbols 2 Wavelength [] Debye-Waller factor (disorder in the neighbour distance) [2] Bragg angle [ or radian]

Abbreviations AFAC BET BJH Amplitude reduction factor Brunauer Emmet Teller Barrett Joyner Halenda

xi

List of symbols and abbreviations

CCD CNF DFT EDS EG EPR ESRF EXAFS FT FWHM HCP HTS ICP-AES LFC LTS MCFC MFC MMO NLDFT NPD OSC PAFC PCA PEMFC PSD RMS SOFC SNBL SSITKA STEM TCD

Charge-coupled device Carbon nanofibers Density functional theory Energy dispersive X-ray spectroscopy Ethylene glycol Electron paramagnetic resonance spectroscopy European synchrotron radiation facility Extended X-ray absorption fine structure Fourier transform Full width half maximum Homogeneous co-precipitation High temperature water-gas shift Inductively coupled plasma atomic emission spectroscopy Liquid flow controller Low temperature water-gas shift Molten carbonate fuel cell Mass flow controller Mixed metal oxides Non-local density functional theory Nitrate precursor decomposition Oxygen storage capacity Phosphoric acid fuel cell Principal component analysis Proton exchange membrane fuel cell or polymer electrolyte membrane fuel cell Pore size distribution Root mean square Solid oxide fuel cell Swiss-Norwegian beam line Steady-state isotopic transient kinetic analysis Scanning transmission electron microscopy Thermal conductivity detector

xii

List of symbols and abbreviations

TEM TGA TOF TOS TPR TPO US VPI WGS XANES XAFS XAS XLBA XPS XRD

Transmission electron microscopy Thermogravimetric analysis Turnover frequency Time on stream Temperature programmed reduction Temperature programmed oxidation Ultrasonic field Effective core-hole lifetime Water-gas shift X-ray absorption near edge structure X-ray absorption fine structure X-ray absorption spectroscopy X-ray line broadening analysis X-ray photoelectron spectroscopy X-ray diffraction

xiii

Introduction

1.
1.1.

Introduction
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 carboncontaining 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 hightemperature fuel cells such as SOFC and MCFC. Hydrogen is the preferred fuel for lowtemperature 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

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

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, poisonresistant 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

The aim of the present work was to gain insight into certain aspects related to copperbased catalysts used for clean fuel applications and small-scale hydrogen production systems, including preparation, characterization and catalytic performance. The watergas 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

2.
2.1.

Theory and literature


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 [3638]. 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

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.

10000 equilibrium constant [ - ]

1000

100

10

1 100

150

200

250

300

350

400

450

500

temperature [ C ]

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

CO conversion whereas the high H2 concentration expected from conventional steam reforming lowers the WGS equilibrium conversion [34].

1,00

0,80 CO conversion [ - ]

0,60

0,40

CO-H2O
0,20

CO-H2O-CO2-H2
0,00 100

150

200

250

300

350

400

450

500

temperature [ C ]

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

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

Table 1. WGS catalyst requirements for mobile and stationary applications [34].

WGS catalyst attribute Volume reduction Weight reduction Cost Rapid response Selectivity No reduction required Oxidation tolerant Condensation tolerant Nonpyrophoric Attrition resistance Pressure drop Poison tolerant

Mobile application Critical <0.1 l kW-1 Critical <0.1 kg kW-1 Critical <$1 kW-1 Critical <15 s Critical Critical Critical Important Important Critical Important Desired

Stationary application Not as constrained Not as constrained Not as critical Load following Important Important Important Important Eliminate purging No constraint Important 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

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.

Improvement of WGS catalyst materials


Metal / Metal Oxide

Alloying Bimetallic systems Alkali promoter

Doping Mixed oxides

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

10

Theory and literature

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]

CeO2

red . ( H 2 / CO) ox. ( H 2O / CO2 )

CeO2x

(2)

11

Theory and literature

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 coprecipitation, since burying of active components in the bulk is avoided. Thus, the optimum metal loading may be lower [133], presuming a good metal dispersion.

12

Experimental

3.
3.1.

Experimental
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)23H2O, 0.1382 mole Zn(NO3)26H2O and 0.0750 mole Al(NO3)39H2O 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)36H2O in 0.6 ml water and 0.274 g Ce(NO3)36H2O 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)23H2O, 59.8 g Zn(NO3)26H2O and 62.5 g Al(NO3)39H2O 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.

13

Experimental

3.2.2. X-ray diffraction (XRD) XRD spectra for crystallite size estimation and phase identification were recorded on Siemens diffractometers D-5000 (monochromatic CuK-radiation) and D-5005 (dichromatic CuK+-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 CuK 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].

14

Experimental

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.

Weight chamber

Purge gas

Reactant gas

Thermocouple

Furnace

Outlet 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 15

Experimental

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.

16

Experimental

Thermocouple

Gas in

Gas out ~A + Liquid

Water cooling

Sample cell Soller slits X-rays Fluorescent X-rays

Kapton window

Water cooling 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

17

Experimental

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 GCanalysis 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 5 (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.

18

Carbonyl trap Vent


CHPC 150

CO MFC

HC

MFC

Vent Reactor I
<1000 oC

CO2

By-pass

MFC

H2

Reactor II
<500 oC

MFC

Vaporizer

O2

MFC

Vent Air Micro heat exchanger Water tank


S V LFC

Vent

Vent

Experimental

N2 PC

Water tank Vent

He

Calib

CHPC 269

GC Vent

Figure 8. Schematic drawing of the experimental setup used for testing the WGS performance of the catalyst materials.

19

20 Experimental

Figure 9. Pictures of the experimental setup used for testing the WGS performance of the catalyst materials.

Experimental

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:

X CO =

CO2 100 % (CO + CO2 )

(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:
(CO0 CO) 100 % (CO + CO2 CO2,0 )

X 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 shortterm deactivation behaviour. The approach to equilibrium is calculated with the exit gas concentrations, corresponding to a certain CO conversion, and the equilibrium constant (Keq):
CO2 H 2 100 % K eq CO H 2O

Equilibrium approach =

(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)

21

Experimental

were diluted with inert SiC (300 600 m, 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 m, 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) B)

SiC
Catalyst Quartz wool Quartz sinter

Catalyst + SiC Quartz wool Inner steal tube with mesh on top

Figure 10. Schematic drawing of the tubular fixed bed reactor configurations: A)

Stainless steel reactor, B) quartz-type reactor.

22

Results and discussion

4.
4.1.

Results and discussion


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 Harkin-Jura Original DFT BJH method 0.0598 0.0347 0.1288 HCP-1b 0.0958 0.1710 0.2678 HCP-1c 0.0554 0.0837 0.1893 HCP-2a 0.2233 0.2247 0.3345 HCP-2b 0.2126 0.1630 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.

23

Results and discussion

0,012 HCP-1a pore volume [ cm/g STP ] HCP-1b 0,009 HCP-1c HCP-2a 0,006 HCP-2b

0,003

0 1 10 100 pore width [ nm ] 1000

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.

0,008

HCP-1a
pore volume [ cm/g STP ]

HCP-1b
0,006

HCP-1c HCP-2a HCP-2b

0,004

0,002

0 1 10 pore width [ nm ] 100 1000

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].

24

Results and discussion

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 K = 1.5418 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.

25

Results and discussion

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 m, 0,15 g) were mixed with SiC (50 150 m, average diameter 88 m) 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.

26

Results and discussion

100 rep 1 CO conversion [ % ] 80 60 40 20 0 190 rep 2 rep 3 rep 4 rep 5

210

230

250 temperature [ C ]

270

290

310

Figure 13. Reproducibility of WGS measurements on Cu-Ce-Zr MMO catalysts (< 200

m) diluted with SiC (50 150 m, average diameter 88 m) 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 m, while the SiC used exhibited particles mainly in the range 50 150 m. This fraction of small catalyst particles (< 50 m 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 m, 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

27

Results and discussion

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 m. 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 m)-fraction show 13 19 % lower values than the (50 200 m)-fraction, when scaled up to 0.10 g with the integral model. An experiment with 0.10 g of the (< 50 m)-fraction also resulted in a lower CO conversion than for the (50 200 m)-fraction. This suggests that the (< 50 m)-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 m 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 m was therefore introduced and used in subsequent experiments.

28

Results and discussion

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.

29

Results and discussion

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-Znbased 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 pretreatments 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, B2O3containing and B2O3-free Cu-Zn-Al MMO catalysts could be improved by co-

30

Results and discussion

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 CuZn-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

31

Results and discussion

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 TGATPR 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.

0,00

300

-0,02 Norm. deriv. weight [mg/(mgCu*min) ]

250 Temperature [ C ]

-0,04

200

-0,06

Cu-aC-350-ox-red Cu-350 Temperature

150

-0,08

100

-0,10 0 25 50 75 Time [ min ] 100 125

50 150

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.

32

Results and discussion

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) Cu mass frac. [-]

6.74 mg/mg-sample

ICP-AES TGA-TPR - without ox. treat. - with ox. treat.

0.26
Precursor weight loss

0.44 0.23
Cu dispersion [%]

(equivalent to Cu mass frac. of 0.21)

4.99 mg/mg-sample ( 0.4 0.8 %)

TGA-N2O - without ox. treat. - with ox. treat. - with ox. treat.
Effect of nitrate precursor on N2O titration

4.8 8.9 7.9 based on TGA-TPR Cu content based on ICP-AES Cu content

Without ox. treat. With ox. treat.

0.26 10-2 0.26 10-2

mg-O/mg-sample mg-O/mg-sample

Correction of Cu particle size estimated by Scherrer equation

Passivation degree [%] Without ox. treat. With ox. treat. 32 59

Particle size [nm] 16 19

Cu dispersion from XRD [%] 5.6 4.6

33

Results and discussion

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

34

Results and discussion

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.

35

Results and discussion

2,00 Ni-Fe cat norm. absorption [ a.u. ] NiO 1,50 Ni foil

1,00

0,50

0,00 8,31

8,33

8,35 photon energy [ keV ]

8,37

8,39

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:

dP = 3

1 dC (1 x p )

(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

36

Results and discussion

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 reducedpassivated Ni-Fe catalyst samples at the Ni K-edge.
Ni metal core [nm]

4.5

(with XRD, after passivation, [182])

Ni metal (%) 2 reference compounds

NiO (%)

NiAl2O4 (%)

Calcined Reduced-passivated Passivation degree [%] Particle size [nm]

7 70 30 5.4

93 30

3 reference compounds

Calcined Reduced-passivated Passivation degree [%] Particle size [nm]

0 74 8 4.7

62 6

38 20

37

Results and discussion

4.2.3. Preparation and characterization of nanocrystalline, high-surface area CuCe-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 singlephase 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.

38

Results and discussion

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].

surface area ratio A/A(x=0 or 0.12) [-]

1,6

1,2

0,8

0,4

BET [Lam] stv [Lam] m-BET [Lam] BET [Shen] stv [Shen] m-BET [Shen]
0 0,1 0,2 0,3 0,4 0,5 Cu mole fraction x [-] 0,6 0,7 0,8

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

39

Results and discussion

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).

40

Results and discussion

A)

B)

Atomic Metal % Cu K Zr L Ce L 10.52 7.73 20.68 frac. 0.27 0.20 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 K1. D) Ce L1. E) Zr K1. F) O K1. G) EDS spectrum of the mapping area. The size bar corresponds to 40 nm for all images.

41

Results and discussion

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 CuCe- 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.

42

Results and discussion

Additional data:

Figure 19 shows the TPR profile of CuZn-CP.

time [ min ] 0 0 normalised derivative weight [ mg/(mgCu min) ] -0,01 -0,02 240 -0,03 rep 1 -0,04 -0,05 rep 2 temp 90 190 140 25 50 75 100 125 150 390 340 290 temperature [ C ]

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

43

Results and discussion

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 CNFfree 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 coprecipitation 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 [189191]) 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 coprecipitation with urea. Pt-impregnated Cu-Ce-Zr and Cu-Zn-Al MMO catalysts were

44

Results and discussion

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 CuZn-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

45

Results and discussion

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.

46

Concluding remarks

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

47

Concluding remarks

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 prereduction, 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

48

Concluding remarks

temperatures, e.g. by decreasing the CuO particle size and hence increasing the surfaceto-volume ratio, then Cu-Zn-Al mixed oxide catalysts may exhibit an increased lowtemperature 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-ZnAl 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 Cubased 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

49

References

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 [56,203207]. 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 [211213], 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 metalbased catalysts may then give information about the true potential of these two classes of catalysts.

50

References

References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] http://www.shell.com/renewables. RENERGI programme established by The Research Council of Norway: http://www.forskningsradet.no/renergi. http://www.shell.com/hydrogen. http://www.hytrec.no. R. Farrauto, S. Hwang, L. Shore, W. Ruettinger, J. Lampert, T. Giroux, Y. Liu, O. Ilinich, Annu. Rev. Mater. Res. 33 (2003) 1. C. Song, Catal. Today 77 (2002) 17. J.N. Armor, Catal. Lett. 101 (2005) 131. J.M. Ogden, M.M. Steinbugler, T.G. Kreutz, J. Power Sources 79 (1999) 143. B.F. Hagh, J. Power Sources 130 (2004) 85. A. Ersoz, H. Olgun, S. Ozdogan, C. Gungor, F. Akgun, M. Trs, J. Power Sources 118 (2003) 384. J.R. Lattner, M.P. Harold, Appl. Catal. B 56 (2005) 149. D.L. Trimm, Z. Ilsen nsan, Catal. Rev. 43 (2001) 31. B. Emonts, J. Bgild Hansen, T. Grube, B. Hhlein, R. Peters, H. Schmidt, D. Stolten, A. Tschauder, J. Power Sources 106 (2002) 333. R. Holland, G. Schubak, M. Bradley, K. OConnor, B. Peppley, US Patent US6572837, 2003. J.D. Holladay, Y. Wang, E. Jones, Chem. Rev. 104 (2004) 4767. I. Aartun, Microstructured reactors for hydrogen production, Ph.D. thesis, Norwegian University of Science and Technology NTNU, Trondheim, Norway, 2005. [17] [18] [19] [20] [21] [22] I. Aartun, B. Silberova, H. Venvik, P. Pfeifer, O. Grke, K. Schubert, A. Holmen, Catal. Today 105 (2005) 469. J.R. Lattner, M.P. Harold, Int. J. Hydrogen Energy 29 (2004) 393. W. Ruettinger, O. Ilinich, R.J. Farrauto, J. Power Sources 118 (2003) 61. D.G. Lffler, K. Taylor, D. Mason, J. Power Sources 117 (2003) 84. C.D. Dudfield, R. Chen, P.L. Adcock, Int. J. Hydrogen Energy 26 (2001) 763. 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.

51

References

[23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34]

J. Han, I.-S. Kim, K.-S. Choi, Int. J. Hydrogen Energy 27 (2002) 1043. O. Grke, P. Pfeifer, K. Schubert, Appl. Catal. A 263 (2004) 11. T. Giroux, S. Hwang, Y. Liu, W. Ruettinger, L. Shore, Appl. Catal. B 56 (2005) 95. I.R. Wheeldon, B.A. Peppley, H. Wieland, US Patent US2003215374, 2003. A. Karim, J. Bravo, D. Gorm, T. Conant, A. Datye, Catal. Today 110 (2005) 86. A.F. Ghenciu, Curr. Opin. Solid State Mater. Sci. 6 (2002) 389. B. Silberova, H.J. Venvik, J.C. Walmsley, A. Holmen, Catal. Today 100 (2005) 457. B. Silberova, H.J. Venvik, A. Holmen, Catal. Today 99 (2005) 69. B. Arstad, H. Venvik, H. Klette, J. Walmsley, W.M. Tucho, R. Holmestad, A. Holmen, R. Bredesen, Catal. Today, accepted. D. Andreeva, Gold Bull. 35 (2002) 82. L. Mond and C. Langer, British Patent 12 (1888) 608. 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] [36]

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. 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] [38] [39]

K. Atwood, M.R. Arnold, E.G. Appel, Ind. Eng. Chem. 42 (1950) 1600. I. Fishtik, R. Datta, Surf. Sci. 512 (2002) 229. 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] [41]

C. Rhodes, G.J. Hutchings and A.M. Ward, Catal. Today 23 (1995) 43. K. Kochloefl, in: Handbook of Heterogeneous Catalysis, eds. G. Ertl, H. Knzinger, J. Weitkamp, VCH, Ludwigshafen, Vol. 4, Chapter 3.3, 1997, pp. 1831-1843.

52

References

[42] [43] [44] [45]

J. Barbier, D. Duprez, Appl. Catal. B 4 (1994) 105. R.F. Mann, J.C. Amphlett, B. Peppley, C.P. Thurgood, Int. J. Chem. React. Eng. 2 (2004) A5. N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H. Ribeiro, J. Catal. 217 (2003) 233. (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] [47] [48] [49]

R.L. Keiski, O. Desponds, Y.-F. Chang, G.A. Somorjai, Appl. Catal. A 101 (1993) 317. E. Fiolitakis, H. Hofmann, J. Catal. 80 (1983) 328. C. Callaghan, I. Fishtik, R. Datta, M. Carpenter, M. Chmielewski, A. Lugo, Surf. Sci. 541 (2003) 21. C.V. Ovesen, B.S. Clausen, B.S. Hammershi, G. Steffensen, T. Askgaard, I. Chorkendorff, J.K. Nrskov, P.B. Rasmussen, P. Stoltze, P. Taylor, J. Catal. 158 (1996) 170.

[50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60]

D.S. Newsome, Catal. Rev. Sci. Eng. 21 (1980) 275. Y. Li, Q. Fu and M. Flytzani-Stephanopoulos, Appl. Catal., B 27 (2000) 179. X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055. S.L. Swartz, M.M. Seabaugh, C.T. Holt, W.J. Dawson, Fuel Cells Bull. 4 (2001) 7. T. Utaka, T. Takeguchi, R. Kikuchi, K. Eguchi, Appl. Catal. A 246 (2003) 117. H. Meland, T. Johannessen, B. Arstad, H.J. Venvik, M. Rnning, A. Holmen, Stud. Surf. Sci. Catal., accepted. L.A. Kristiansen, US Patent US4308176, 1981. W.F. Ruettinger, O.M. Ilinich, R.J. Farrauto, US Patent US2004082669, 2004. O.M. Ilinich, W.F. Ruettinger, R.T. Mentz, R.J. Farrauto, US Patent US2004082471, 2004. P. Panagiotopoulou, D.I. Kondarides, Catal. Today 112 (2006) 49. F. Boccuzzi, A. Chiorino, M. Manzoli, D. Andreeva, T. Tabakova, J. Catal. 188 (1999) 176.

53

References

[61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80]

A. Basinska, J. Stoch, F. Domka, Pol. J. Environ. Stud. 12 (2003) 395. H. Sakurai, A. Ueda, T. Kobayashi, M. Haruta, Chem. Commun. 3 (1997) 271. J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157. T. Utaka, T. Okanishi, T. Takeguchi, R. Kikuchi, K. Educhi, Appl. Catal. A 245 (2003) 343. A. Basiska, L. Kpiski, F. Domka, Appl. Catal. A 183 (1999) 143. J.C. Frost, Nature 334 (1988) 577. A. Basiska, W.K. Jwiak, J. Gralski, F. Domka, Appl. Catal. A 190 (2000) 107. Q. Fu, W. Deng, H. Saltsburg, M. Flytzani-Stephanopoulos, Appl. Catal. B 56 (2005) 57. T. Tabakova, F. Boccuzzi, M. Manzoli, J.W. Sobczak, V. Idakiev, D. Andreeva, Appl. Catal. B 49 (2004) 73. D. Andreeva, V. Idakiev, T. Tabakova, R. Giovanoli, Bulg. Chem. Comm. 30 (1998) 59. F. Boccuzzi, A. Chiorino, M. Manzoli, D. Andreeva, T. Tabakova, L. Ilieva, V. Idakiev, Catal. Today 75 (2002) 169. T. Tabakova, V. Idakiev, D. Andreeva, I. Mitov, Appl. Catal. A 202 (2000) 91. G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G. Thomas, B.H. Davis, Appl. Catal. A 258 (2004) 203. P. Panagiotopoulou, A. Christodoulakis, D.I. Kondarides, S. Boghosian, J. Catal. 240 (2006) 114. S. Hilaire, X. Wang, T. Luo, R.J. Gorte, J. Wagner, Appl. Catal. A 258 (2004) 271. M. Saito, J. Wu, K. Tomoda, I. Takahara, K. Murata, Catal. Lett. 83 (2002) 1. M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003) 11. L.-T. Weng, B. Delmon, Appl. Catal. A 81 (1992) 141. R. Di Monte, J. Kapar, J. Mater. Chem. 15 (2005) 633. S. Colussi, C. de Leitenburg, G. Dolcetti, A. Trovatelli, J. Alloys Comp. 374 (2004) 387.

54

References

[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] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99]

W. Ruettinger, X. Liu, R.J. Farrauto, Appl. Catal. B 65 (2006) 135. A. Basiska, F. Domka, Appl. Catal. A 179 (1999) 241. K. Sekizawa, S.-i. Yano, K. Eguchi, H. Arai, Appl. Catal. A 169 (1998) 291. E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347. X. Wang, R.J. Gorte, Appl. Catal. A 247 (2003) 157. S. Zhao, R.J. Gorte, Catal. Lett. 92 (2004) 75. C.-S. Chen, W.-H. Cheng, S.-S. Lin, Appl. Catal. A 257 (2004) 97. S.Y. Choung, M. Ferrandon, T. Krause, Catal. Today 99 (2005) 257. A. Basiska, F. Domka, Catal. Lett. 43 (1997) 59. S.-i. Ito, H. Tanaka, Y. Minemura, S. Kameoka, K. Tomishige, K. Kunimori, Appl. Catal. A 273 (2004) 295. J. Nakamura, J.M. Campbell, C.T. Campbell, J. Chem. Soc., Faraday Trans. 86 (1990) 2725. M. Maack, H. Friis-Jensen, S. Sckerl, J.H. Larsen, I. Chorkendorff, Top. Catal. 22 (2003) 151. J. Wold, L.A. Kristiansen, Patent WO 8504597 PCT NO8500016, 1985. N. Pavlenko, P.P. Kostrobij, Yu. Suchorski, R. Imbihl, Surf. Sci. 489 (2001) 29. A. Houteit, H. Mahzoul, P. Ehrburger, P. Bernhardt, P. Lgar, F. Garin, Appl. Catal. A 306 (2006) 22. Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Catalytic Science Series Vol. 2, ed. G.J. Hutchings, Imperial College Press, London, 2002. D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107. 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.

55

References

[105] N.Y. Usachev, I.A. Gorevaya, E.P. Belanova, A.V. Kazakov, O.K. Atalyan, 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. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal. Today 100 (2005) 249. [117] J. Kapar, 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. SolGel Sci. Tech. 26 (2003) 879. [126] Y. Zhang, S. Andersson, M. Muhammed, Appl. Catal. B 6 (1995) 325.

56

References

[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. Lematre, 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, Users Manual, Siemens. [140] Diffracplus Win-crysize, Crystallite size and microstrain, Users 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, Users Manual, Micrometrics Instrument Corporation.

57

References

[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. Rnning, 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. Fernndez-Garca, Catal. Rev. 44 (2002) 59. [166] B.S. Clausen, J.K. Nrskov, Top. Catal. 10 (2000) 221.

58

References

[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. Prez-Ramrez, 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. Knzinger, 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. Sjblom, E.A. Blekkan, Micropor. Mesopor. Mater. 86 (2005) 198. [174] R.J. Berger, J. Prez-Ramrez, F. Kapteijn, J.A. Moulijn, Appl. Catal. A 227 (2002) 321. [175] R.J. Berger, J. Prez-Ramrez, 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, . 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.

59

References

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. Garca-Bordej, I. Kvande, D. Chen, M. Rnning, Adv. Mater. 18 (2006) 1589. [196] O. Smiljanic, T. Dellero, A. Serventi, G. Lebrun, B.L. Stansfield, J.P. Dodelet, M. Trudeau, S. Dsilets, 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.

60

References

[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. Gro, Top. Catal. 37 (2006) 29.

61

List of appendices

List of appendices
I M. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Relating catalyst structure and composition to the water-gas shift activity of Cu-Znbased mixed-oxide catalysts, Catalysis Today 100 (2005), 249-254. II F. Huber, Z. Yu, S. Lgdberg, M. Rnning, 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. Rnning, 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. Rnning, H. Venvik, A. Holmen, Comparison of CuCe-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 CuCe-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift, Manuscript in preparation.

62

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.

Catalysis Today 100 (2005) 249254 www.elsevier.com/locate/cattod

Relating catalyst structure and composition to the watergas shift activity of CuZn-based mixed-oxide catalysts
Magnus Rnninga,*, Florian Hubera, Hilde Melanda, Hilde Venvikb, De Chena, Anders Holmena
a

Norwegian University of Science and Technology (NTNU), Department of Chemical Engineering, N-7491 Trondheim, Norway b SINTEF Materials and Chemistry, N-7465 Trondheim, Norway Available online 28 December 2004

Abstract In order to investigate the effect of cerium oxide on CuZn-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 watergas 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 watergas 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 CuZn precursors leads to a more stable watergas shift catalyst. # 2004 Elsevier B.V. All rights reserved.
Keywords: Watergas shift; Copper; Cerium oxide; XANES; XRD; Thermogravimetric analysis

1. Introduction The watergas 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 component because of its proven activity [2,3]. However, there is a need for catalysts with even higher activity and stability compared to the traditional CuOZnOAl2O3 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 [47]. 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: ronning@chemeng.ntnu.no (M. Rnning). 0920-5861/$ see front matter # 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.cattod.2004.09.059

[7,911]. The promoting metal in the form of e.g. Pt, Au or Cu signicantly 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 CuZnOAl2O3 catalysts derived from layered hydroxide materials of the hydrotalcite structure [13]. Mixed-oxide catalysts derived from hydrotalcite-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].

250

M. Rnning et al. / Catalysis Today 100 (2005) 249254

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, ltration, 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 ltered 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 CuOZnOAl2O3 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 CuOZnOAl2O3 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-aC350 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, ow rate 80 ml/min at ambient temperature). Temperature-programmed reduction (TPR) using 7 vol.% H2 or CO in Argon (heating rate: 2 K/min, ow 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 N2Odecomposition of the reduced catalyst samples were
Table 1 Chemical composition and physical properties for the catalyst samples Sample ICP-AESa catalyst composition mass fractionb Cu Cu-350 Ce-aC-350 Cu-400 Ce-bC-400 0.29 0.26 0.27 0.25 Zn 0.39 0.35 0.40 0.37 Al 0.08 0.08 0.09 0.08 Ce 0.00 0.07 0.00 0.06 O 0.24 0.24 0.24 0.24 72.4 76.2 BETc (m2/g)

corrected for bulk oxidation of copper according to a method proposed by Sato et al. [17]. XRD spectra for crystallite size estimation and phase identication were recorded on Siemens diffractometers D5000 (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) reection 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 SwissNorwegian 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 1.4 104 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

H2O/CO2 lossd (%)

Cu dispersione (%)

XRDf Cu crystal size (nm) Dispersion (%) 3.8 5.6 3.5 3.8

5.7 3.2 1.7 1.7

6.5 4.8 5.4 8.2

23 16 25 23

a ICP-AES: inductively coupled plasma-atomic emission spectroscopy with an estimated detection limit of 0.010.03 mg/g. The elementary analysis was performed on calcined samples. b Normalised mass fractions, i.e. only CuO, ZnO, Al2O3 and CeO2 taken into account. c Performed on the calcined samples, prior to reduction. d Thermogravimetric measurements performed in Ar on the calcined samples up to the reduction temperature (260 8C) resulted in a weight loss assigned to adsorbed water and/or surface carbonates [6]. e Dispersion based on oxygen chemisorption (from N2O) of the reduced catalyst samples taking into account copper bulk oxidation [17]. f Performed on the reduced and passivated samples. The copper crystallite size is estimated using the Scherrer equation for the Cu(1 1 1) reection taking into account the thickness of the passivation layer. The copper dispersion is calculated from the crystallite size assuming a cubic particle shape with one face in contact with the support [18].

M. Rnning et al. / Catalysis Today 100 (2005) 249254

251

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%; ow rate 30 ml/min at ambient temperature) by heating at a rate of 6 K/min from room temperature to 350 8C. XANES proles 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 conditions. The XANES data were energy calibrated, pre-edge background subtracted (linear t) and normalised. Principal component analysis (PCA) was applied for identication 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 tting procedure to determine the fraction of each phase present [21]. The initial watergas shift activity was tested in an externally heated tubular xed-bed reactor with on-line GC analysis over a wide temperature range (200300 8C) [15]. In order to minimize temperature gradients the catalyst samples were diluted with inert SiC (300600 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 ow rate were chosen to perform the activity measurements within the kinetic regime and away from the thermodynamic equilibrium of the watergas shift reaction [23]. The catalysts (0.15 0.25 mg, 200400 mm) were pre-reduced in situ at 260 8C for 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 ow (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.

Fig. 1. XRD spectra of the (a) calcined and (b) reducedpassivated 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.

3. Results and discussion The chemical composition and some physical properties of the catalyst samples are presented in Table 1. Phase identication was performed by XRD both before and after calcination. The nominal catalyst loadings were conrmed by inductively coupled plasma (ICP-AES). The copper content was also conrmed by the TPR-TGA analysis except for CeaC-350. For this catalyst, the extent of reduction exceeded the maximum copper equivalent indicating a reduction of components other than copper oxide. This effect is object for more studies and will not be discussed here in further detail. XRD results conrm 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 signicantly more intense in the samples containing ceria, for the calcined (Fig. 1(a)) as well as the reducedpassivated 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 298 indicates either low CeO2 crystallinity or small crystallites. It has been stated in the literature that ZnO is not an active reaction site in the watergas shift reaction [2,3]. The XRD data suggest that addition of CeO2 to the copperzinc 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) reection (see Table 1). The Cu dispersion obtained from oxygen chemisorption show Cu dispersions of 48% (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 (3.55.5%) 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 chemisorption have been used since this method gives a more direct, and thus presumably more accurate, estimate of the active

252

M. Rnning et al. / Catalysis Today 100 (2005) 249254

surface than the XLBA results from the reducedpassivated 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 copperbased 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, ow rate: 30 ml/min at ambient temperature. (a) Cu K-edge proles 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 reected in a pronounced anisotropy in particle shape seen when analysing the particle size in different crystallographic directions [15]. Zn and Ce K-edge XANES proles (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 frequencies (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 sufcient contact between both phases. The XRD spectra of the reducedpassivated 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 sufcient interface between copper and ceria. Ceria included after calcination seems to reduce the chemisorption-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. Rnning et al. / Catalysis Today 100 (2005) 249254

253

Fig. 3. Temperature-programmed reduction with 7% H2 in Argon using a thermogravimetric device. Heating rate: 2 K/min, ow 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 prole 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 prole of Ce-aC-350 (Fig. 3). It is evident that the preparation procedure for introducing CeO2 is important for the effect of CeO2. As mentioned above, uncertainties still exist concerning the contribution

of ceria to the measured dispersion based on oxygen chemisorption by means of N2O surface titration. Comparing 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

Fig. 4. (a) Turnover frequency (TOF) and (b) short-term deactivation behaviour of the catalyst samples. TOF is calculated from oxygen chemisorption (from N2O) and corrected for loss of H2O/CO2. The deactivation measurements are performed at 300 8C and are presented as normalised reaction rates for the disappearance of CO vs. time on stream (TOS). The normalised reaction rates are not corrected for the H2O/CO2 uptake of the calcined samples, but as the corresponding catalysts exhibit a quite similar weight loss, this effect is negligible. The step in the deactivation curve for Cu-400 after 20 h stems from running out of nitrogen. This has no further effect on the deactivation behaviour of the catalyst.

254

M. Rnning et al. / Catalysis Today 100 (2005) 249254

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 inuence on the deactivation behaviour.

References
[1] Fuel Cell Handbook, 5th ed., US Department of Energy, NETL, 2000. [2] C.V. Ovesen, B.S. Clausen, B.S. Hammershi, G. Steffensen, T. Askgaard, I. Chorkendorff, J.K. Nrskov, 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 (13) (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) 2733. [12] M. Fernandez-Garca, E. Gomez Rebollo, A. Guerrero Ruiz, J.C. Conesa, J. Soria, J. Catal. 172 (1997) 146. ` [13] F. Cavani, F. Triro, A. Vaccari, Catal. Today 11 (1991) 173. [14] M.J.L. Gines, N. Amadeo, M. Laborde, C.R. Apestegua, 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) 118122. [21] T. Ressler, J. Wong, J. Roos, I.L. Smith, Environ. Sci. Technol. 34 (2000) 950. [22] R.J. Berger, J. Perez-Ramrez, 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 Verlagsgesellschaft mbH, 1997(Chapter 9). [24] J.-D. Grunwaldt, A.M. Molenbroek, N.-Y. Topse, H. Topse, 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.

4. Conclusions A comparative study on the effect of cerium oxide on the watergas shift activity of CuZn-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 chemisorption-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 conditions. 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 SwissNorwegian Beam Lines (SNBL) at the ESRF for their assistance. Elin Nilsen (Department of Materials Technology, NTNU) and Egil Haans (Department of Chemical Engineering, NTNU) are gratefully acknowledged for their assistance with XRD and TGA, respectively.

Paper II

Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts


Catalysis Letters, in press.

Remarks on the passivation of reduced Cu-, Ni-, Fe-, Co-based catalysts Florian Huber1, Zhixin Yu1, Sara Lgdberg2, Magnus Rnning1, De Chen1, Hilde Venvik1,*, Anders Holmen1
1

Department of Chemical Engineering, Norwegian University of Science and

Technology (NTNU), N-7491 Trondheim, Norway


2

Department of Chemical Engineering and Technology, Royal Institute of Technology

(KTH), SE-100 44 Stockholm, Sweden * Corresponding author; E-mail: Hilde.Venvik@chemeng.ntnu.no, 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

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 [34]. 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]

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. Cu350 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. CebC-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)26H2O. 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].

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 C, 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 C, 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 C after 1 h in Ar) was determined prior to reduction (at 260 C 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 C, 1 h, Ar), reduced (260 C, 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 + O2 Cu2O (2)

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 (O2passivated) copper catalysts were recorded on Siemens diffractometers D-5000 (monochromatic CuK-radiation) and D-5005 (dichromatic CuK+-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 = 21.67 using the Scherrer equation and LaB6 as a standard for correction of instrumental line broadening (Be)

dP =

K B 2 Be2 cos()

(3)

with = 1.5418 , the wavelength for Cu K, 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 Storster et al. [34] and references therein, most of the reducible cobalt oxide is reduced after this procedure. Subsequently, the sample was

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, preedge 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, -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

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:

dP = 3

1 dC (1 x p )

(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

(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 prereduction with CO.

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, -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 (= FeOFe2O3) 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

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

10

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 wellknown 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 ratelimiting step during bulk oxidation with O2. In principle, this is also the case for passivation with N2O [27,38,42].

11

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

12

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 CNFencapsulated 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 reducedpassivated 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

13

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

14

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 Storster et al. [34], about twothirds 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-

15

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 steamreforming-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

16

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

17

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 NTNUSINTEF 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 Haans (Department of Chemical Engineering, NTNU) are gratefully acknowledged for their assistance with the XRD devices and the TGA, respectively. Hilde Meland and Cathrine Brin 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] . Borg, M. Rnning, S. Storster, 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. Topse, H. Topse, B.S. Clausen, J. Catal. 194 (2000) 452. [7] M.M. Gnter, B. Bems, R. Schlgl, T. Ressler, J. Synch. Rad. 8 (2001) 619.

18

[8] P.L. Hansen, J.B. Wagner, S. Helveg, J.R. Rostrup-Nielsen, B.S. Clausen, H. Topse, 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. Rnning, 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. Vrlstad, M. Rnning, E. Ochoa-Fernndez, 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. Grnert, O. Hinrichsen, A. Birkner, S. Rabe, K. Merz, M. Driess, C. Wll, 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.

19

[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. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal. Today 100 (2005), 249. [34] (a) S. Storster, . Borg, E.A. Blekkan, A. Holmen, J. Catal. 231 (2005) 405; (b) S. Storster, B. Ttdal, 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. Lematre, 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. Nrskov, Top. Catal. 10 (2000) 221. [44] I. Kvande, D. Chen, M. Rnning, 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.

20

[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. Ttdal, 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).

21

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 C, 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 C, 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 -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 O2passivated sample and is therefore not shown in the figure.

22

Figure 1

50 49
N2O decomposition water, carbonates

300 250 200


Cu reduction

weight loss [ mg ]

47 46

150 100

45 44 43 42 41 0 100 200 300 400 500 600

Cu-350 temperature program programme temperature 700 800 900

50 0

time [min]

temperature [ C ]
23

48

Figure 2

(a)

4000 3000 2000 1000 0 30 40 50 60


2[]

Ce-bC-400

signal [a.u.]

Cu-400

Cu-350

70

80

90

(b)

6000 5000 4000 3000 2000 1000 0 25 30 40


signal [a.u.]

Ce-bC-400

Cu-400

Cu-350

50
2[]

60

70

24

Figure 3

normalized weight increase [ - ]

0,08

0,06

0,04

0,02

Ce-bC-400-H2 Ce-bC-400-CO

0 0 2 4 6 8

t^(1/2) [ min^(1/2) ]

25

Figure 4

Y norm. absorption [ a.u.]

1.5

1.0

0.5
NiO Calcined Red-pass Used Ni foil 2.25 8.32

4.5
Z

8.38 8.36 8.34

photon energy [ keV ]


X

26

Figure 5

1.5

norm. absorption [ a.u.]

1.0

0.5
Fe3O4 -Fe2O3 Calcined Red-pass Used Fe foil 7.1

7.2 7.15 photon energy [ keV ]

27

Figure 6

norm. absorption [a.u.]

1.0

0.5

8.32

8.34 photon energy [keV]

8.36

8.38

28

Figure 7

calcined reduced-(O2)passivated signal [a.u.]

2[]

29

Table 1. Chemical composition and physical properties of the copper catalyst samples. Cu crystallite size Sample H2O/CO2 weight lossa Total Cu mass fraction in the samples [-] [ wt.% ] Cu-350 Cu-400 Ce-bC-400
a

Cu dispersiond

Cu passivated with N2Oe [%]

Cu passivated with O2e [%]

(with O2 passivation) Cu core with XRD


f

Core + pass.layer (XRD + TGA)g [ nm ]

ICP-AESb

TGA-TPRc

[%]

[ nm ]

5.7 1.7 1.7

0.29 0.27 0.25

0.28 0.26 0.25

6.5 5.4 8.2

10 7 11

40 45 40

19 20 20

23 25 24

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. 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. 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. 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. Normalized with amount of Cu determined by TGA-TPR prior to passivation, with percentage values being based on mole fraction. Performed on the reduced and passivated samples. The copper crystallite size is estimated using the Scherrer equation for the Cu (111) reflection. The thickness of the passivation layer is taken into account assuming a cubic particle shape.

e f g

30

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 Calcined Reduced-passivated Used

Ni metal [ % ] 7 70 98

NiO [ % ] 93 30 2

Fe K-edge: Sample Calcined Reduced-passivated Used Fe metal [ % ] 0 57 82 -Fe2O3 [ % ] 65 39 10 Fe3O4 [ % ] 35 4 8

31

Table 3. Thermodynamics of the metal oxidation by O2 and N2O, based on the

standard enthalpies of formation for the bulk metals. Standard reaction enthalpy [ kJ/mole ]a N2O (per mole N2O) Cu Cu2O Ni NiO Fe Fe2O3 Fe Fe3O4 Co CoO Co Co3O4 -250 -318 -354 -358 -316 -303 O2 (per mole O2) -337 -472 -543 -408 -468 -442

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).

32

Paper III

Preparation and characterization of nanocrystalline, high-surface area Cu-Ce-Zr mixed oxide catalysts from homogeneous co-precipitation
Manuscript in preparation.

Preparation and characterization of nanocrystalline, high-surface area Cu-Ce-Zr mixed oxide catalysts from homogeneous co-precipitation Florian Huber1, Hilde Venvik1,*, Magnus Rnning1, John Walmsley2, Anders Holmen1
1

Department of Chemical Engineering, Norwegian University of Science and

Technology (NTNU), N-7491 Trondheim, Norway


2

SINTEF Materials and Chemistry, N-7465 Trondheim, Norway

* Corresponding author; E-mail: Hilde.Venvik@chemeng.ntnu.no, 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.

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]:

CeO2

red . ( H 2 / CO) ox. ( H 2O / CO2 )

CeO2x

(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

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]:
+ CO( NH 2 ) 2 + 2 H 2O 2 NH 4 + CO32

(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

[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 posttreatment (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

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

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 =

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.

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 surfacecomposition dependence. Comparing the surface-composition dependence for molebased 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

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.

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.

10

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 glycolwater 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.

11

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 CuK+-

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 CuK 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

12

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

13

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 coprecipitation 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. Coprecipitation 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

14

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 , 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 , 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

15

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 2 [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, 2 = 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.

16

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 pretreatment. 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 (with error deviation: 5-15 ) [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 & Nrskov [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-dataset EXAFS analysis [56] including asymmetric pair distribution functions for nanosized particle clusters [54], could further minimize correlation effects.

17

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 watergas 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 & Kapar [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 2-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

18

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. Temperatureprogrammed 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

19

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-

20

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

21

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.

22

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 (CuCO3Cu(OH)2), azurite (2CuCO3Cu(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 waterethanol 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

23

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

24

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.

25

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 150C), as well as BoEG-s100 (drying in static air at 100C) to BoEG-s100-d250 (drying in flowing air at 250C, 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 150C 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 BoEGs100-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 Obridges). 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

26

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 100C (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-s100d250 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 precalcined 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 Bo2C 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

27

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 coprecipitation (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

28

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

29

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] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Trovarelli (Ed.), Catalysis by Ceria and Related Materials, Catalytic Science Series Vol. 2 (Ed.: G.J. Hutchings), Imperial College Press, London, 2002. R. Di Monte, J. Kapar, J. Mater. Chem. 15 (2005) 633-648. Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179-191. E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347-357. X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062. D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107-112. Y. Liu, T. Hayakawa, T. Tsunoda, K. Suzuki, S. Hamakawa, K. Murata, R. Shiozaki, T. Ishii, M. Kumagai, Top. Catal. 22 (2003) 205-213. X. Zhang, P. Shi, J. Mol. Catal. A 194 (2003) 99-105. X.R. Zhang, P. Shi, J. Zhao, M. Zhao, C. Liu, Fuel Process. Tech. 1680 (2003) 110. A. Szizybalski, Ph.D. Thesis, Technical University Berlin, Germany, 2005.

30

[11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32]

P. Bera, S. Mitra, S. Sampath, M.S. Hegde, Chem. Comm. 10 (2001) 927-928. T.-J. Huang, Y.-C. Kung, Catal. Lett. 85 (2003) 49-55. 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. N.Y. Usachev, I.A. Gorevaya, E.P. Belanova, A.V. Kazakov, O.K. Atalyan, V.V. Kharlamov, Russ. Chem. Bull., Int. Ed. 53 (2004) 538-546. G. Avgouropoulos, T. Ioannides, H.K. Matralis, J. Batista, S. Hocevar, Catal. Lett. 73 (2001) 33-40. G. Avgouropoulos, T. Ioannides, H. Matralis, Appl. Catal. B 56 (2005) 87-93. X. Tang, B. Zhang, Y. Li, Y. Xu, Q. Xin, W. Shen, Catal. Today 93-95 (2004) 191-198. P. Bera, S.T. Aruna, K.C. Patil, M.S. Hegde, J. Catal. 186 (1999) 36-44. Lj. Kundakovic, M. Flytzani-Stephanopoulos, J. Catal. 179 (1998) 203-221. Y-j. Wang, X-f. Dong, W-m. Lin, Dianyuan Jishu 26 (2002) 134-136. S. Hocevar, J. Batista, J. Levec, J. Catal. 184 (1999) 39-48. C. de Leitenburg, D. Goi, A. Primavera, A. Trovarelli, G. Dolcetti, Appl. Catal. B 11 (1996) 29-35. E.E. Ortelli, J.M. Weigel, A. Wokaun, Catal. Lett. 54 (1998) 41-48. W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 83 (2002) 33-35. C. Lu, W.L. Worrell, J.M. Vohs, R.J. Gorte, J. Electrochem. Soc. 150 (2003) 1357-1359. S. McIntosh, J.M. Vohs, R.J. Gorte, J. Electrochem. Soc. 150 (2003) 1305-1312. Y-j. Wang, X-f. Jie, X-f. Dong, W-m. Lin, Dianyuan Jishu 26 (2002) 43-46. L. Jalowiecki-Duhamel, A. Ponchel, C. Lamonier, Int. J. Hydrogen Energy 24 (1999) 1083-1092. M. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal. Today 100 (2005) 249-254. J. Kapar, P. Fornasiero, J. Solid State Chem. 171 (2003) 19-29. M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003) 11-13. M. Ozawa, J. Alloy Compd. 275-277 (1998) 886-890.

31

[33] [34] [35] [36]

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. S. Bedrane, C. Descorme, D. Duprez, Catal. Today 75 (2002) 401-405. A. Norman, V. Perrichon, Phys. Chem. Chem. Phys. 5 (2003) 3557-3564. 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] [38]

M.P. Kapoor, A. Raj, Y. Matsumura, Micropor. Mesopor. Mater. 44-45 (2001) 565-572. 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] [40] [41] [42]

M. Hirano, T. Miwa, M. Inagaki, J. Solid State Chem. 158 (2001) 112-117. J.A. Wang, M.A. Valenzuela, S. Castillo, J. Salmones, M. Moran-Pineda, J. SolGel Sci. Tech. 26 (2003) 879-882. Y. Zhang, S. Andersson, M. Muhammed, Appl. Catal. B 6 (1995) 325-337. (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] [44] [45] [46] [47] [48] [49]

C. Lamonier, A. Bennani, A. D`Huysser, A. Aboukais, G. Wrobel, J. Chem. Soc., Faraday Trans. 92 (1996) 131-136. M. Kurtz, H. Wilmer, T. Genger, O. Hinrichsen, M. Muhler, Catal. Lett. 86 (2003) 77-80. S. Golunski, R. Rajaram, N. Hodge, G.J. Hutchings, C.J. Kiely, Catal. Today 72 (2002) 107-113. Y. Liu, T. Hayakawa, K. Suzuki, S. Hamakawa, T. Tsunoda, T. Ishii, M. Kumagai, Appl. Catal. A 223 (2002) 137-145. Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki, K. Eguchi, Appl. Catal. A 238 (2002) 11-18. E. Matijevic, Chem. Mater. 5 (1993) 412-426. G.J.A.A. Soler-Illia, R.J. Candal, A.E. Regazzoni, M.A. Blesa, Chem. Mater. 9 (1997) 184-191.

32

[50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66]

M. Adachi-Pagano, C. Forano, J.-P. Besse, J. Mater. Chem. 13 (2003) 1988-1993. T. Shishido, Y. Yamamoto, H. Morioka, K. Takaki, K. Takehira, Appl. Catal. A 263 (2004) 249-253. W.H.R. Shaw, J.J. Bordeaux, J. Am. Chem. Soc. 77 (1955) 4729-4733. R.B. Greegor, F.W. Lytle, J. Catal. 63 (1980) 476-486. B.S. Clausen, J.K. Nrskov, Top. Catal. 10 (2000) 221-230. M. Borowski, J. Phys. IV 7 (1997) 259-260. A.I. Frenkel, C.W. Hills, R.G. Nuzzo, J. Phys. Chem. B 105 (2001) 12689-12703. M.V. Twigg, M.S. Spencer, Appl. Catal. A 212 (2001) 161-174. A. Wootsch, C. Descorme, D. Duprez, Proc. of the 13th Int. Cong. on Catal., Paris, France, July 2004. M. Yada, H. Kitamura, A. Ichinose, M. Machida, T. Kijima, Angew. Chem. Int. Ed. 38 (1999) 3506-3510. T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 14 (1997) 105-115. T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 15 (1998) 107-114. D.-H. Tsai, T.-J. Huang, Appl. Catal. A 223 (2002) 1-9. X. Zheng, X. Zhang, X. Wang, S. Wang, S. Wu, Appl. Catal. A 295 (2005) 142149. F.B. Noronha, E.C. Fendley, R.R. Soares, W.E. Alvarez, D.E. Resasco, Chem. Eng. J. 82 (2001) 21-31. M.K. Dongare, V. Ramaswamy, C.S. Gopinath, A.V. Ramaswamy, S. Scheurell, M. Bruechner, E. Kemnitz, J. Catal. 199 (2001) 209. A. Mastalir, B. Frank, A. Szizybalski, H. Soerijanto, A. Deshpande, M. Niederberger, R. Schomcker, R. Schlgl, T. Ressler, J. Catal. 230 (2005) 464475.

[67] [68] [69] [70] [71]

J.C. Vargas, E. Vanhaecke, A.C. Roger, A. Kiennemann, Stud. Surf. Sci. Catal. 147 (2004) 115-120. F. Schueth, Chem. Mater. 13 (2001) 3184-3195. J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157-161. E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373-380. P.B. Balbuena, K.E. Gubbins, Fluid Phase Equilib. 76 (1992) 21-35.

33

[72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89]

DFT Plus for Windows, version 3.00, Users Manual, Micrometrics Instrument Corporation. W.D. Harkins, G. Jura, J. Am. Chem. Soc. 66 (1944) 1366-1373. F. Rouquerol, J. Rouquerol, K. Sing, Adsorption by Powders and Porous Solids, Academic Press, London, 1999. S. Chytil, W.R. Glomm, E. Vollebekk, H. Bergem, J. Walmsley, J. Sjblom, E.A. Blekkan, Micropor. Mesopor. Mater. 86 (2005) 198-206. Diffracplus Profile, Profile fitting program, Users Manual, Siemens. Diffracplus Win-crysize, Crystallite size and microstrain, Users Manual, Bruker Analytical X-Ray Systems. B.E. Warren, in: B. Chalmes, R. King (Eds.), Progress in Metal Physics Vol. 8, Pergamon Press, London, 1959, pp. 147-202. T. Ressler, J. Synch. Rad. 5 (1998) 118-122. N. Binsted, EXCURV98: CCLRC Daresbury Laboratory computer program, 1998. F.H. Firsching, J. Mohammadzadel, J. Chem. Eng. Data 31 (1986) 40-42. F. Zhang, S.-P. Yang, H.-M. Chen, X.-B. Yu, Ceram. Int. 30 (2004) 997-1002. Y. Nagai, T. Yamamoto, T. Tanaka, S. Yoshida, T. Nonaka, T. Okamoto, A. Suda, M. Sugiura, Catal. Today 74 (2002) 225-234. D.C. Koningsberger, B.L. Mojet, G.E. van Dorssen, D.E. Ramaker, Top. Catal. 10 (2000) 143-155. V. Ramaswamy, M. Bhagwat, D. Srinivas, A.V. Ramaswamy, Catal. Today 97 (2004) 63-70. A.S. Deshpande, N. Pinna, P. Beato, M. Antonietti, M. Niederberger, Chem. Mater. 16 (2004) 2599-2604. S.H. Overbury, D.R. Huntley, D.R. Mullins, G.N. Glavee, Catal. Lett. 51 (1998) 133-138. Florian Huber, Zhixin Yu, Sara Lgdberg, Magnus Rnning, Hilde Venvik, Anders Holmen, submitted. S. Brunauer, L.S. Deming, W.S. Deming, E. Teller, J. Am. Chem. Soc. 62 (1940) 1723-1732.

34

[90] [91] [92] [93] [94] [95]

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. A.F. Hollemann, E.Wiberg, Lehrbuch der Anorganischen Chemie, 101st ed. (Walter de Gruyter, Berlin, 1995). R.J. Candal, A.E. Regazzoni, M.A. Blesa, J. Mater. Chem. 2 (1992) 657-661. P. McFadyen, E. Matijevic, J. Inorg. Nucl. Chem. 35 (1973) 1883-1893. J.-P. Andreassen, M.J. Hounslow, AIChE J. 50 (2004) 2772-2782. (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. Krger, U. Hoffmann, U. Kunz, ECCE-1 Proceedings Vol. 2 (1997) 1507-1510.

[96] [97] [98] [99]

L. Pei, K.-I. Kurumada, M. Tanigaki, M. Hiro, K. Susa, J. Mater. Sci. 39 (2004) 663-665. C.J. Brinker, G.W. Scherer, Sol-Gel Science: The Physics and Chemistry of solgel processing (Academic Press, London, 1990). B.-Q. Xu, J.-M. Wei, Y.-T. Yu, J.-L. Li, Q.-M. Zhu, Top. Catal. 22 (2003) 77-85. 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.

35

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.

36

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 = 39 -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 K1. D) Ce L1. E) Zr K1. F) O K1. 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 BoEGs100-d250 was dried at 250 C in air.

37

Figure 1

2 surface area ratio A/A(x = 1) [ - ] BET [Kap] 1,8 stv [Kap] m-BET [Kap] BET [Hir] 1,6 stv [Hir] m-BET [Hir] 1,4

1,2

1 0 0,2 0,4 0,6 0,8 1 1,2 Ce mole fraction x [ - ]

38

Figure 2

A) T

B)
M

39

Figure 3 A)

B)

40

Figure 4

(d) signal [a. u.]

(c) (b)

(a)

5 20

30

40

50
2 theta

60

70

80

41

Figure 5 A)
25 relative frequency [ % ] 20 15 10 5 0 0 2 4 6 8 10 crystal size estimate [ nm ]

B)
100 cumulative frequency [ % ]

80

60

40

20

0 0 2 4 6 8 10 crystal size estimate [ nm ]

C)
2

RMS strain [ 1e-2, - ]

1,6

1,2

0,8

0,4

0 0 2 4 6 8 10 crystal size estimate [ nm ]

42

Figure 6

2,00

norm. absorption [ - ]

1,50

1,00 Bo-d120 0,50 CuO Cu2O Cu-foil 0,00 8,96 8,98 9,00 9,02 9,04 9,06

photon energy [ keV ]

43

44 B) A)
FT k3 (k) k3 (k)

Figure 7

Figure 8 A) B) Atomic [%] Cu K Zr L Ce L 10.14 9.91 17.88 Molar frac. [-] 0.27 0.26 0.47

C)

D)

E)

F)

G)

45

Figure 9

0 derivate weight [ mg/min ]

-0,1

-0,2

-0,3

TPR1 TPR2
-0,4 60 100 140 180 220 260 temperature [ C ]

46

Figure 10

220 nitrogen adsorbed [ cm/g STP ]


pore volume [ cm/g STP ]

0,010

0,008

180 140 100 60 20 0

0,006

0,004

0,002

0,000 1 10 pore width [ nm ] 100

0,2

0,4

0,6

0,8

relative pressure [ - ]

47

Figure 11

0,012

BoEG-s100
pore volume [ cm/g STP ] 0,01 0,008 0,006 0,004 0,002 0 1 10 pore width [ nm ] 100

BoEG-s100-d250

48

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 [ m2/g ] stv pore volume XRD crystal size [ nm ]

[ m2/mm3 ] [ cm3/g ] [ cm3/cm3 ]

Bc HPo HPo-2C Bo-d120b Bo-2C Bo Bo-dry Bo-redox Bo-d150 BoEth BoEth-dry BoEG-s100 BoEG-s100-dry BoEG-s100-d250 Bo-s98

121.13 89.24 89.26 145.20 163.79 154.42 141.69 109.34 110.49

0.86 0.62 0.62 1.01 1.15 1.08 0.99

0.1347 0.1288 0.1352 0.2720 0.2632 0.2678 0.2464 0.2344

0.96 0.90 0.95 1.90 1.84 1.87 1.72 3.8 3.4 3.3 3.1 3.6 3.1

0.78

0.1943

1.37

5.1 4.2

170.65 120.76 115.29

1.19 0.85

0.3345 0.2201 0.1893

2.34 1.54

3.1 2.8 3.5 3.6

a b

Estimated error of BET smaller than 5 % of the total value. 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.

49

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 -1) and the reference compound CuO (k3 weighting, fitting interval: k = 3-12 -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 E0 [ eV ] R1 []

0.94
N1 [-] 2 [ 2 ]

Bo-d120 CuO CuOa CeO2a

-10.5 1.0 -15.2 0.8

1.92 0.01 1.96 0.01 1.96 2.34

2.5 0.5 3.9 0.2 4.0 8.0

0.003 0.002 0.008 0.001

Crystallographic data obtained from literature for CuO [101] and CeO2 [102].

50

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. Bc HPo Bo-d120 Bo Bo-TPR 1 Bo-TPR 2 BoEth BoEG-s100

0.23 0.10 0.22 0.23 0.23(5) 0.16a 0.20 0.19 0.22

0.54 0.63 0.55 0.54 0.54

0.23 0.27 0.23 0.23 0.22(6)

0.58 0.55

0.23 0.23

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 %).

51

Paper IV

Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift
Manuscript submitted.

Comparison of Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift Florian Huber, Hilde Meland, Magnus Rnning, Hilde Venvik*, Anders Holmen

Department of Chemical Engineering, Norwegian University of Science and Technology (NTNU), N-7491 Trondheim, Norway * Corresponding author; E-mail: Hilde.Venvik@chemeng.ntnu.no, 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 CuCe-Zr mixed metal oxide (MMO) catalysts were prepared both by co-precipitation and

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]. CuCeUCP 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 (ICPAES), 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 coprecipitated and flame spray pyrolysed catalysts, respectively. The Cu dispersion of CuZn-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

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-ZnAl 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 CuZnCP/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-

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-CeLa 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-CeZr 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-tovolume 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.

Acknowledgements This work was supported by the Research Council of Norway and Statoil ASA through the Gas Technology Center NTNU-SINTEF. Cathrine Brin Nilsen (Depart. of Chem. Eng., NTNU) and Bjrnar 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] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] D. Andreeva, V. Idakiev, T. Tabakova, L. Ilieva, P. Falaras, A. Bourlinos, A. Travlos, Catal. Today 72 (2002) 51. A. Trovarelli, Catalysis by Ceria and Related Materials, Catalytic Science Series, Vol. 2 (Imperial College Press, London, 2002). M. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal. Today 100 (2005) 249. M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003) 11. R. Di Monte, J. Kapar, J. Mater. Chem. 15 (2005) 633. L.A. Kristiansen, US Patent US4308176, 1981. F. Huber, Z. Yu, J. Walmsley, D. Chen, H. Venvik, A. Holmen, Appl. Catal. B, in press. H. Meland, T. Johannessen, B. Arstad, H.J. Venvik, M. Rnning, A. Holmen, Stud. Surf. Sci. Catal., accepted. F. Huber, Z. Yu, S. Lgdberg, M. Rnning, D. Chen, H. Venvik, A. Holmen, Catal. Lett., in press. N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H. Ribeiro, J. Catal. 217 (2003) 233-239. M.S. Spencer, Catal. Lett. 32 (1995) 9-13. R.F. Mann, J.C. Amphlett, B. Peppley, C.P. Thurgood, Int. J. Chem. React. Eng. 2 (2004) A5. R.L. Keiski, O. Desponds, Y.F. Chang, G.A. Somorjai, Appl. Catal. A 101 (1993) 317-338.

[14] [15]

Y. Choi, H.G. Stenger, J. Power Sources 124 (2003) 432-439. C.V. Ovesen, B.S. Clausen, B.S. Hammershi, G. Steffensen, T. Askgaard, I. Chorkendorff, J.K. Nrskov, P.B. Rasmussen, P. Stoltze, P. Taylor, J. Catal. 158 (1996) 170-180.

[16] [17]

X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062. Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179-191.

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-FSPnorm corresponds to CuCe-FSP normalized with the ratio of the Cu content in CuZnFSP (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.

Figure 1 A)
15 CuZn-FSP CO conversion [ % ] 12 CuCe-FSP CuCe-FSP_norm

0 180

200

220

240

260

280

300

320

temperature [ C ]

B)
100 CuCe-UCP_CO+H2O CO conversion [ % ] 80 CuCe-UCP_reformate CuCe-UCP_CO+H2O_norm CuZn-CP_CO+H2O CuZn-CP_reformate CuCe-UCP_reformate_norm

60

40

20

0 160

210

260 temperature [ C ]

310

360

C)
3 CuCe-UCP CuZn-CP ln(CO conversion) [ - ] 2 CuZn-FSP CuCe-FSP 1

0 0,0016 -1

0,0018

0,002

0,0022

0,0024

1/T [ 1/K ]

Figure 2 A)
1,2 normalized CO conversion [ - ] 1 0,8 0,6 0,4

CuZn-FSP
0,2 0
0 2 4 6 TOS [ h ] 8 10 12

CuCe-FSP

B)
1,2 normalized CO conversion [ - ] 1 0,8 0,6 CuCe-UCP_reformate 0,4 0,2 0 0 4 8 TOS [ h ] 12 16 20 CuZn-CP_reformate CuCe-UCP_CO+H2O CuZn-CP_CO+H2O

10

Table 1. Chemical composition, structural characteristics and apparent WGS activation energies (Ea) of the catalysts. catalyst composition molar metal fraction
a

BET b

surface volume

XRD d crystal size CO+H2O

Ea reformate

Cu [-] CuZn-FSP CuCe-FSP CuZn-CP CuCe-UCP


a

Ce or Zn [-]
0.43(6) 0.54 0.40 0.54

Zr or Al [-]
0.23(7) 0.23 0.33 0.23

[m2/g]
160 93 80 164

[m2/mm3]
0.86 0.65 0.41 1.15

[nm]
n.d. n.d. 14 2 37 48

[kJ/mole]

0.32(7) 0.23 0.27 0.23

44 f 32 f

83 33

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
c

Performed on the calcined samples, after drying at 150 C under vaccum for 1.5 h. 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 -Al2O3 (3650 kg/m3). 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 CuCeUCP estimated by software-based X-ray line broadening analysis [7]. 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.

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
a

Department of Chemical Engineering, Norwegian University of Science and

Technology (NTNU), N-7491 Trondheim, Norway


b

SINTEF Materials and Chemistry, N-7465 Trondheim, Norway

* Corresponding author; E-mail: Hilde.Venvik@chemeng.ntnu.no, 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.

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], watergas 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]:
+ CO ( NH 2 ) 2 + 2 H 2O 2 NH 4 + CO32

(1)

The decomposition rate strongly depends on the reaction temperature [25,27].

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 , 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

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

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 CuK+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 CuK 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 temperatureprogrammed oxidation mode (TPO) was applied to check the metal oxide content in the

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 m) 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

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:

X CO =

CO2 100 % (CO + CO2 )

(2)

For the simulated reformate feed, the calculated CO conversion must take into account the initial composition of the feed gas:
(CO0 CO) 100 % (CO + CO2 CO2,0 )

X 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

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:
CO2 H 2 100 % K eq CO H 2O

Equilibrium approach =

(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

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.

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

10

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

11

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 CNFfree 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 CNFfree 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

12

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 CNFfree 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 ratedetermining step. Frseth 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.

13

The impact of CNF on the catalyst activity and stability could be related to sizematching [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.

14

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] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17]

Y. Liu, T. Hayakawa, T. Tsunoda, K. Suzuki, S. Hamakawa, K. Murata, R. Shiozaki, T. Ishii, M. Kumagai, Top. Catal. 22 (2003) 205-213. X. Zhang, P. Shi, J. Mol. Catal. A 194 (2003) 99-105. X.R. Zhang, P. Shi, J. Zhao, M. Zhao, C. Liu, Fuel Process. Tech. 1680 (2003) 110. A. Szizybalski, Ph.D. Thesis, Technical University Berlin, Germany, 2005. Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179-191. E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347-357. X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062. P. Bera, S. Mitra, S. Sampath, M.S. Hegde, Chem. Comm. 10 (2001) 927-928. T.-J. Huang, Y.-C. Kung, Catal. Lett. 85 (2003) 49-55. D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107-112. 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. N.Y. Usachev, I.A. Gorevaya, E.P. Belanova, A.V. Kazakov, O.K. Atalyan, V.V. Kharlamov, Russ. Chem. Bull., Int. Ed. 53 (2004) 538-546. G. Avgouropoulos, T. Ioannides, H. Matralis, Appl. Catal. B 56 (2005) 87-93. E.E. Ortelli, J.M. Weigel, A. Wokaun, Catal. Lett. 54 (1998) 41-48. W.-J. Shen, Y. Ichihashi, Y. Matsumura, Catal. Lett. 83 (2002) 33-35. P. Bera, S.T. Aruna, K.C. Patil, M.S. Hegde, J. Catal. 186 (1999) 36-44. M. Rnning, F. Huber, H. Meland, H. Venvik, D. Chen, A. Holmen, Catal. Today 100 (2005) 249-254.

15

[18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35]

M. Saito, K. Tomoda, I. Takahara, M. Kazuhisa, M. Inaba, Catal. Lett. 89 (2003) 11-13. R. Di Monte, J. Kapar, J. Mater. Chem. 15 (2005) 633-648. K. P. De Jong and J.W. Geus, Catal. Rev. Sci. Eng. 42 (2000) 481510. X. Dong, H.-B. Zhang, G.-D. Lin, Y.-Z. Yuan, K.R. Tsai, Catal. Lett. 85 (2003) 237-246. H.-B. Zhang, X. Dong, G.-D. Lin, X.-L. Liang, H.-Y. Li, Chem. Comm. 40 (2005) 5094-5096. E. Matijevic, Chem. Mater. 5 (1993) 412-426. G.J.A.A. Soler-Illia, R.J. Candal, A.E. Regazzoni, M.A. Blesa, Chem. Mater. 9 (1997) 184-191. M. Adachi-Pagano, C. Forano, J.-P. Besse, J. Mater. Chem. 13 (2003) 1988-1993. T. Shishido, Y. Yamamoto, H. Morioka, K. Takaki, K. Takehira, Appl. Catal. A 263 (2004) 249-253. W.H.R. Shaw, J.J. Bordeaux, J. Am. Chem. Soc. 77 (1955) 4729-4733. Z. Yu, D. Chen, B. Ttdal, A. Holmen, J. Phys. Chem. B 109 (2005) 6096-6102. F. Huber, H. Venvik, M. Rnning, J. Walmsley, A. Holmen, Manuscript in preparation. E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373-380. Diffracplus Profile, Profile fitting program, Users Manual, Siemens. Diffracplus Win-crysize, Crystallite size and microstrain, Users Manual, Bruker Analytical X-Ray Systems. B.E. Warren, in: B. Chalmes, R. King (Eds.), Progress in Metal Physics Vol. 8, Pergamon Press, London, 1959, pp. 147-202. R.L. Keiski, O. Desponds, Y.F. Chang, G.A. Somorjai, Appl. Catal. A 101 (1993) 317-338. F. Kapteijn, J.A. Moulijn, Laboratory Reactors, in: G. Ertl, H. Knzinger, 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.

16

[37] [38] [39]

M.L. Toebes, Y. Zhang, J. Hjek, 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. S. Hocevar, J. Batista, J. Levec, J. Catal. 184 (1999) 39-48. (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. Krger, U. Hoffmann, U. Kunz, ECCE-1 Proceedings Vol. 2 (1997) 1507-1510.

[40] [41] [42] [43] [44]

A. Cacciuto, S. Auer, D. Frenkel, Nature 428 (2004) 404-406. M.K. van der Lee, J. van Dillen, J.H. Bitter, K.P. de Jong, J. Am. Chem. Soc. 127 (2005) 13573-13582. 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. V. Frseth, Z. Yu, D. Chen, A. Holmen, Manuscript in preparation. B.-Q. Xu, J.-M. Wei, Y.-T. Yu, J.-L. Li, Q.-M. Zhu, Top. Catal. 22 (2003) 77-85.

17

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.

18

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 %).

19

Figure 1 A)

B)

20

Figure 2

21

Figure 3

(002) P-CNF (111) (200) CNF24 CNF13 CNF07 CNF00 20 30 40 50 60 (220) (311)

22

Figure 4 A)

B)

23

Figure 5

40

CO conversion [ % ]

30

20

CNF00
10

CNF07 CNF13 CNF24

0 160

180

200

220

240

260

Temperature [ C ]

24

Figure 6

1,2 Normalized CO conversion [ - ] 1 0,8 0,6 0,4 0,2 0 0 3 6 TOS [ h ] 9 12 15

CNF00 CNF07 CNF13 CNF24

25

Figure 7

50

CNF00
CO conversion [ % ] 40

CNF13 CNF24

30

20

10

0 160

200

240

280

320

360

Temperature [ C ]

26

Figure 8

ln(CO conv.) [ - ]

2,5

2 CNF00 1,5 CNF13 CNF24 1 0,0018

0,00185

0,0019

0,00195

0,002

0,00205

1/T [ 1/K ]

27

Figure 9

1,2 Normalized CO conversion [ - ] 1 0,8 0,6

CNF00
0,4 0,2 0 0 2 4 6 8 TOS [ h ] 10 12 14 16

CNF13 CNF24

28

Table 1. Structural characteristics of the dried and calcined CNF and Cu-Ce-Zr MMO

nanocomposites.
XRD crystal sizea [ nm ]

BET [ m2/g ]

Pore volume [ cm3/g ]

P-CNF CNF00 CNF07 CNF13 CNF24b

172 164 n.d. 164 139

0.326 0.317 n.d. 0.307 0.262 2.4 (3.4) 2.8 2.8 3.1 (3.1)

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.

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.

29

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 Temperature [ C ] CO conv. [%]

Rel. CO conv. normalized with metal oxide content CNF07/CNF00 [-] CNF13/CNF00 [-] CNF24/CNF00 [-]

165 180

10.46 14.25 32

0.95 0.95 31

1.22 1.22 31

1.01 1.07 34

Ea [ kJ/mole ]

(165 210 C)

30

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 Temperature [ C ] CO conv. [%]

Rel. CO conv. normalized with metal oxide content CNF13/CNF00 [-] CNF24/CNF00 [-]

225 250 275

6.23 8.97 9.66 33

1.19 1.30 1.31 38

1.10 1.27 1.25 39

Ea [ kJ/mole ]

(225 275 C)

31

Paper VI

The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for water-gas shift
Manuscript in preparation.

The effect of platinum in Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts for watergas shift Florian Huber1, John Walmsley2, Hilde Venvik1,*, Anders Holmen1
1

Department of Chemical Engineering, Norwegian University of Science and

Technology (NTNU), N-7491 Trondheim, Norway


2

SINTEF Materials and Chemistry, N-7465 Trondheim, Norway

* Corresponding author; E-mail: Hilde.Venvik@chemeng.ntnu.no, Tel: +47-73592831, Fax: +47-73595047 Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts were prepared by homogeneous coprecipitation 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 lowtemperature 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

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 CuCe-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-ZnAl 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 Han were used as metal precursors. Tetraamineplatinum(II)-nitrate (99.995 %) was purchased from Aldrich Chem. Co. Urea ( 99.5 %) was purchased

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 Ptsalt 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).

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 CuK+-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.

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 m and 50 125 m 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 Hepressurized 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.

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 prereduction 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:

X CO =

CO2 100 % (CO + CO2 )

(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:
(CO0 CO) 100 % (CO + CO2 CO2,0 )

X CO =

(2)

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-

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-Albased 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-Zrbased 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

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 CuZn-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

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 CuCe-Zr mixed oxide catalyst under the given conditions. Hungra 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 pretreatment. 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.

10

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.

11

Utaka et al. [14] also observed a decreased activity of pre-reduced, Pt-impregnated CuZn-Al mixed oxide catalysts for oxygen-assisted WGS as compared to pre-reduced CuZn-Al. Huang & Sachtler [21] reported that the addition of Ru, Rh, Pd or Pt to NaYsupported 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 ZnAlPt 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 Al2O3supported, 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

12

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 CuZn-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. Hungra et al. [15] reported that the Pd-Cu alloy formed in an Al2O3supported 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

13

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.

14

4. Conclusions

Cu-Ce-Zr and Cu-Zn-Al mixed oxide catalysts were prepared by homogeneous coprecipitation 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 CuZn-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 CuZn-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

15

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] [2]

D. Andreeva, Gold Bull. 35 (2002) 82-88. 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] [4] [5] [6] [7] [8] [9] [10]

R. Farrauto, S. Hwang, L. Shore, W. Ruettinger, J. Lampert, T. Giroux, Y. Liu, O. Ilinich, Annu. Rev. Mater. Res. 33 (2003) 1-27. Y. Choi, H.G. Stenger, J. Power Sources 124 (2003) 432-439. 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. D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90-102. C. Rhodes, G.J. Hutchings and A.M. Ward, Catal. Today 23 (1995) 43-58. D.S. Newsome, Catal. Rev. Sci. Eng. 21 (1980) 275-318. Y. Li, Q. Fu and M. Flytzani-Stephanopoulos, Appl. Catal., B 27 (2000) 179-191. X. Qi, M. Flytzani-Stephanopoulos, Ind. Eng. Chem. Res. 43 (2004) 3055-3062.

16

[11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32]

N.A. Koryabkina, A.A. Phatak, W.F. Ruettinger, R.J. Farrauto, F.H. Ribeiro, J. Catal. 217 (2003) 233-239. E.S. Bickford, S. Velu, C. Song, Catal. Today 99 (2005) 347-357. E.S. Bickford, S. Velu, C. Song, Preprints of American Chemical Society Symposia, Division of Fuel Chemistry 49 (2004) 649-651. T. Utaka, K. Sekizawa, K. Eguchi, Appl. Catal. A 194-195 (2000) 21-26. A.B. Hungra, A. Iglesias-Juez, A. Martnez-Arias, M. Fernndez-Garca, J.A. Anderson, J.C. Conesa, J. Soria, J. Catal. 206 (2002) 281-294. R.E.R. Colen, M. Kolodziejczyk, B. Delmon, J.H. Block, Surf. Sci. 412/413 (1998) 447-457. N.N. Hoover, B.J. Auten, B.D. Chandler, J. Phys. Chem B 110 (2006) 8606-8612. P.C. Liao, J.J. Carberry, T.H. Fleisch, E.E. Wolf, J. Catal. 74 (1982) 307-316. E.S. Ranganathan, S. Bej, L.T. Thompson, Preprints of American Chemical Society Symposia, Division of Fuel Chemistry 48 (2003) 381-382. S. Zhou, B. Varughese, B. Eichhorn, G. Jackson, K. McIlwrath, Angew. Chem. Int. Ed. 44 (2005) 4539-4543. Y. Huang, W.M.H. Sachtler, J. Catal. 188 (1999) 215-225. A. Gro, Top. Catal. 37 (2006) 29-39. V. Ponec, Appl. Catal. A 222 (2001) 31-45. E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373-380. P.B. Balbuena, K.E. Gubbins, Fluid Phase Equilib. 76 (1992) 21-35. DFT Plus for Windows, version 3.00, Users Manual, Micrometrics Instrument Corporation. F. Huber, H. Venvik, M. Rnning, J. Walmsley, A. Holmen, Appl. Catal. A, submitted. W.C. Conner, J.L. Falconer, Chem. Rev. 95 (1995) 759-788. M.H. Kim, J.R. Ebner, R.M. Friedman, M.A. Vannice, J. Catal. 208 (2002) 381392. J.B. Ko, C.M. Bae, Y.S. Jung, D.H. Kim, Catal. Lett. 105 (2005) 157-161. X. Tang, B. Zhang, Y. Li, Y. Xu, Q. Xin, W. Shen, Catal. Today 93-95 (2004) 191-198. J.-P. Shen, C. Song, Catal. Today 77 (2002) 89-98.

17

[33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52]

C.E. Quincoces, N.E. Amadeo, M.G. Gonzalez, Stud. Surf. Sci. Catal. 111 (1997) 535-541. W.F. Ruettinger, O.M. Ilinich, R.J. Farrauto, US Patent US2004082669, 2004. O.M. Ilinich, W.F. Ruettinger, R.T. Mentz, R.J. Farrauto, US Patent US2004082471, 2004. K.-D. Jung, O.-S. Joo, S.H. Han, Catal. Lett. 68 (2000) 49-54. B.H. Sakakini, J. Tabatabaei, M.J. Watson, K.C. Waugh, J. Mol. Catal. A 162 (2000) 297-306. H. Wilmer, O. Hinrichsen, Catal. Lett. 82 (2002) 117-122. S. Fujita, S. Moribe, Y. Kanamori, M. Kakudate, N. Takezawa, Appl. Catal. A 207 (2001) 121-128. 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. P.L. Hansen, J.B. Wagner, S. Helveg, J.R. Rostrup-Nielsen, B.S. Clausen, H. Topse, Science 295 (2002) 2053-2055. A.A.G. Lima, M. Nele, E.L. Moreno, H.M.C Andrade, Appl. Catal. A 171 (1998) 31-43. C. Rhodes, G.J. Hutchings and A.M. Ward, Catal. Today 23 (1995) 43-58. F. Epron, F. Gauthard, J. Barbier, J. Barbier, Appl. Catal. A 237 (2002) 253-261. A. Ruban, B. Hammer, P. Stoltze, H.L. Skriver, J.K. Nrskov, J. Mol. Catal. A 115 (1997) 421-429. J. Greeley, M. Mavrikakis, Nat. Mater. 3 (2004) 810-815. D. Woodruff (Ed.), Surface alloys and alloy surfaces, The Chemical Physics of Solid Surfaces Series Vol. 10, Elsevier, Amsterdam, 2002. L. Yang, Curr. Top. Catal. 2 (1999) 59-71. A. Christensen, A.V. Ruban, P. Stoltze, K.W. Jacobsen, H.L. Skriver, J.K. Nrskov, F. Besenbacher, Phys. Rev. B 56 (1997) 5822-5834. A.V. Ruban, H.L. Skriver, J.K. Nrskov, Phys. Rev. B 59 (1999) 15990-16000. K. Persson, A. Ersson, S. Colussi, A. Trovarelli, K. Jansson, S.G. Jrs, Proc. of the 12th Nordic Symp. on Catal., Trondheim, Norway, May 2006, 80-81. F. Besenbacher, I. Chorkendorff, B.S. Clausen, B. Hammer, A.M. Molenbroek, J.K. Nrskov, I. Stensgaard, Science 279 (1998) 1913-1915.

18

[53] [54] [55] [56] [57] [58]

J.H. Larsen, I. Chorkendorff, Surf. Sci. Rep. 35 (1999) 163-222. D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90-102. Y. Tanaka, T. Utaka, R. Kikuchi, K. Sasaki, K. Eguchi, Appl. Catal. A 238 (2003) 11-18. O. Hinrichsen, T. Genger, M. Muhler, Stud. Surf. Sci. Catal. 130 (2000) 38253830. (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. D.H. Kim, J.E. Cha, Catal. Lett. 86 (2003) 107-112.

19

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.

20

Figure 1 A)
300 Nitrogen adsorbed [cm /g STP] CeZr CeZr-Pt ZnAl 200 ZnAl-Pt

100

0 0 0,2 0,4 0,6 0,8 1 Relative pressure [-]

B)
0,02 CeZr Pore volume [cm /g STP] 0,016 CeZr-Pt ZnAl 0,012 ZnAl-Pt

0,008

0,004

0 1 10 Pore width [nm] 100 1000

21

Figure 2

PtO2 Pt

ZnAl-Pt Intensity [a. u.]

10000

ZnAl CeZr-Pt CeZr

0 15 20 30 40 50 60 70 80

[]

22

Figure 3 A)

100 nm

B)

200 nm

23

C)

50 nm

24

Figure 4

180 CeZr 140 Signal [ a. u. ] 100 60 20 -20 30 70 110 150 190 230 270 310 Temperature [ C ] CeZr-Pt ZnAl ZnAl-Pt

25

Figure 5

1,1 Normalized CO conversion [ - ]

0,9

0,7

CeZr red CeZr not red CeZr-Pt red CeZr-Pt not red

0,5 0 4 8 TOS [ h ] 12 16

26

Figure 6

24

ZnAl red
20 CO conversion [ % ] 16 12 8 4 0 150

ZnAl not red ZnAl-Pt red ZnAl-Pt not red

170

190

210 temperature [ C ]

230

250

270

27

Figure 7 A)
20

CO conversion [ % ]

16

12

ZnAl red ZnAl not red ZnAl-Pt red ZnAl-Pt not red

0 0 5 10 TOS [ h ] 15 20 25

B)
1,2 Normalized CO conversion [ - ]

0,8

0,4

ZnAl red ZnAl not red ZnAl-Pt red ZnAl-Pt not red

0 0 5 10 TOS [ h ] 15 20 25

28

Figure 8 A)

25

ZnAl red
CO conversion [ % ] 20

ZnAl-Pt red

15

10

0 170

190

210

230

250

270

290

310

temperature [ C ]

B)
24

CO conversion [ % ]

18

12

ZnAl red ZnAl-Pt red

0 0 8 16 TOS [ h ] 24 32 40

29

Table 1. Composition and structural parameters of the mixed oxide catalysts. Catalyst composition metal content Cu Ce/Zn Zr/Al Pt

BET

surface volume

pore volume a

[% (g/g-cat.)]
CeZr CeZr-Pt ZnAl ZnAl-Pt 9.4 9.2 19.4 18.8 69.4 67.2 42.0 40.3 14.0 13.5 7.3 7.1 2.2 0.7

[m2/g] [m2/mm3] [cm3/g] [cm3/cm3]


161 147 74 79 0.40 1.12 0.40 0.30 0.20 0.18 1.08 2.79

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 -Al2O3 (3650 kg/m3).

30

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 prereduction.

CO conversion [ % ] Temperature [ C ] 165 180 195 CeZr CeZr-Pt

pre-reduced not pre-reduced pre-reduced not pre-reduced 6.47 9.27 12.53 6.39 8.39 10.55 5.14 7.94 11.10 5.76 8.36 11.10

31