Catalytic Oxidations in Gas-Phase and Supercritical CO2

Research Collection

Doctoral Thesis

Catalytic oxidations in gas-phase and supercritical CO₂ Investigations of structure-performance relationships Author(s): Kimmerle, Bertram Sebastian Publication Date: 2010 Permanent Link: https://doi.org/10.3929/ethz-a-006096525

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library

Diss. ETH No. 18926

Catalytic Oxidations in Gas-Phase and Supercritical CO2: Investigations of Structure-Performance Relationships ABHANDLUNG zur Erlangung des Titels

DOKTOR DER WISSENSCHAFTEN der ETH ZÜRICH vorgelegt von BERTRAM SEBASTIAN KIMMERLE Dipl.-Ing., Universität Karlsruhe (TH) Geboren am 29.05.1980 In Ludwigsburg (Deutschland)

Angenommen auf Antrag von Prof. Dr. A. Baiker Prof. Dr. A. Wokaun Prof. Dr. J.-D. Grunwaldt

2010

“… ein Stückchen Urwald wenigstens müssen wir haben, und die Lust des Vordringens wollen wir um keinen Preis missen. Und von allen Richtungen, die wir zu diesem Zweck einschlagen konnten, schien mir keine dankbarer und hoffnungsreicher als die Katalyse.“

(W. Ostwald, Über Katalyse, 1901)

ii

iii

Acknowledgements First of all I want to thank my doctoral advisor Prof. Alfons Baiker for the great years in his group. I do not think that there are many places, where I would have been allowed to work that freely on so many different projects. Besides enormously pushing my motivation for science – often by some kind of reverse psychology – he taught me some more general things like composure, the necessity to wait for the right moment to react, and that you should not get corrupted by politics. I am very thankful for that. I hope that your fascination for almost everything will continue to exist and that you find a good road for your future – maybe even without catalysis. I could experience Prof. Jan-Dierk Grunwaldt on his way from being a Privatdozent in our group through a professorship at DTU in Copenhagen to a chair at the KIT in Karlsruhe, and could work together with him the whole time. He acquainted me with heterogeneous catalysis and introduced me to the fascinating world of synchrotron-based experiments. He always managed to bring people of different talents together and create an atmosphere of constructive discussion. His motivation and devotion are phenomenal. Thank you for your patience, your tolerance, for discussions during vast amounts of coffee at synchrotron beam lines by far not only about science, and the incredible possibilities you opened up for me. I thank Prof. Alexander Wokaun for immediately agreeing to be my coexaminer and the effort in clearing up the date of the exam with the department. An important paragraph has to be dedicated to my colleagues. They made me miss ETH after some days of absence. What better could happen than having friends around all day, who care for you, are competent, and motivated enough not only to work together on whatever project seems interesting, but also to try to solve any problem coming your way. Stefan, uncountable discussions about chemistry, engineering, society, theology, metaphysics, gender issues, literature, music, linguistics, and many more

iv

subjects have made – among other things – our friendship so sustainable (I have to use this word at least once in a work about catalysis). Peter, us working together on our shared projects was a great time. You had the invaluable trait to tell me when my ideas were up the pole. The 80’s song of the week will not be forgotten and I still hold the fortress of our political and social views. Niels, I have a deep respect for you approach to experimentation (if more people worked like you, we would probably not read that much crap every day), your – admittedly sometimes unnerving – patience, and your willingness to help. Our courses in contemporary highlevel insulting given together with Dr. Kleist were not passed by many. Wolfgang, you taught me the unchallengeable importance of detailed knowledge about Austria and fringe sports and Erik, thanks for making me feel bad, when I tried to skip sports. Sven is for me the engineer of chemists. Although very busy you always helped me in every way possible. I’d like to thank Atsushi for his time, our discussions what science and chemical engineering should be about, and the professional calm in analyzing situations. And Marek, besides some very nice and funny fondue evenings, I thank you for your profound skepticism – with you one is never overconfident with his own results (“of course, you cannot do anything with it, Bertram … but the pictures are nice”). The following present, temporary and former colleagues are thanked for a lot of enjoyable moments and discussions in special topics: Fabian in high pressure matters, Jean-Michel in ATR-IR measurements, Ronny in everything computer-related (games not being the least important), and Ive Hermans and Matthias in the complex world of radical reactions. I had two students, who worked with me with incredible motivation and the will to learn what they can, Ueli Neuenschwanderand Jonathan Bauer. Thanks you guys, your enthusiasm really brought forward our projects. There are a lot of staff members of the group and the department, without whom this work would not have been possible. I thank Andreas Dutly for his never-ending patience in helping me to solve critical problems concerning gas chromatographs and teaching me their proper maintenance. Max Wohlwend, in addition to build every kind of weird cable or control setup I needed, gave me private lessons in understanding pieces of

v

electronic equipment answering all my questions during a cup of coffee. The workshop guys, Urs Krebs, Roland Mäder, Philippe Trüssel, Jean-Pierre Mächler, and Roland Walker are thanked for their willingness to always help on short-hand and the motivation to build the best spectroscopic cells one can imagine. To Frank Krumeich I am grateful, since he measured loads of – electron-microscopically speaking boring – gold samples without complaining. Erika, thank you for your friendliness, the uncomplicated protocol in your office, and that you opened up a whole world of new sports for me. An important group of people are the physicist, with whom we did many of our synchrotron experiments, namely the group of Prof. Christian Schroer at TU Dresden and the group of Prof. Ronald Frahm at the University of Wuppertal. From Dresden I’d like to thank Christian himself – his knowledge, his motivation, and his almost indestructible good mood are the icing on the cake of a synchrotron beamtime. Pit Boye, Jens Patommel and Sandra Stephan are thanked for days (and nights) of data processing and introduction and help with IGOR (Pit) and Linux (Jens). Many thanks go to Jan Stoetzel from Frahm’s group; doing quick-EXAFS experiments with him is insightful as well as successful. I wished we could work together a little more and hope that – at least sometimes – you had the feeling “einmal mit Profis [zu] arbeiten”. Before coming to a final, more private paragraph, I want to acknowledge all the persons at the different synchrotron facilities, who not only made our experiments possible, but very often dedicated a lot of their (also free) time to optimize our results: Stefan Mangold at the XAS-beamline at ANKA, Karlsruhe, Pieter Glatzel and Christophe Lapras at ID26, ESRF, Grenoble, Wouter van Beek, Olga Safonova and Herman Emerich at the SwissNorwegian beamline, ESRF, Grenoble, Edmund Welter, Mathias Hermann and Adam Webb at beamlines X1 and C at DORIS III, Hasylab, Hamburg, Wolfgang Drube at beamline BW2, DORIS III, Hasylab, Hamburg, Camelia Borca and Beat Meyer at the MicroXAS beamline at the SLS, Villigen, and last but not least Maarten Nachtegaal and Markus Willimann at the SuperXAS beamline, SLS, Villigen.

vi

Finally, I would like to thank my parents, Konrad and Beatrix Kimmerle, who always supported me, took and take me under their wing whenever necessary, and are constantly generous enough to overlook my flaws. And Kathrin, my time at ETH will be always linked to us growing together. Thank you for your support, your willingness to tolerate my weaknesses, and uncountable beautiful moments.

vii

Table of Content Acknowledgements ................................................................. iii Table of Content......................................................................vii Summary ..................................................................................xi Zusammenfassung.................................................................. xiv 1

General Introduction ...................................................... 1

1.1 Heterogeneous catalysis – some thoughts...................................... 1 1.1.1 From the beginning .................................................................. 1 1.1.2 Specialties of heterogeneous catalysis ..................................... 5 1.1.3 Relevance for our world – why the research? .......................... 8 1.2 Gold in Catalysis ............................................................................ 10 1.2.1 History and hysteria ................................................................ 10 1.2.2 Present research ..................................................................... 12 1.3 X-ray absorption spectroscopy for catalysis .................................. 14 1.3.1 Synchrotron radiation............................................................. 14 1.3.1.1 Origin and production ...................................................... 14 1.3.1.2

Insertion devices .............................................................. 18

1.3.2 X-ray absorption spectroscopy (XAS)...................................... 20 1.3.2.1 Extended X-ray absorption fine structure (EXAFS) ........... 22 1.3.2.2

X-ray absorption near edge structure (XANES) ................ 25

1.3.3 Usefulness for catalysis........................................................... 27 1.4 Catalytic partial oxidation of methane .......................................... 28 1.5 Aim of the Thesis ........................................................................... 31

2

Gold-catalyzed alcohol oxidation in supercritical carbon dioxide ......................................................................... 33

2.1 Introduction .................................................................................. 34 2.1.1 Catalysis in supercritical carbon dioxide ................................. 34 2.1.2 Gold-catalyzed alcohol oxidation ........................................... 36

viii

2.2 Experimental ................................................................................. 37 2.2.1 Catalysts Preparation.............................................................. 37 2.2.2 Catalyst Characterization ........................................................ 38 2.2.3 Reaction Setup........................................................................ 39 2.2.4 Phase behavior and IR measurements ................................... 41 2.3 Alcohol oxidation on gold in scCO2................................................ 41 2.3.1 Results .................................................................................... 41 2.3.2 Discussion ............................................................................... 49 2.4 Influence of gold particle size........................................................ 52 2.4.1 Results .................................................................................... 52 2.4.2 Discussion ............................................................................... 56 2.5 Summary ....................................................................................... 56

3

Development of an in situ parallel XAS cell and its application to steady-state CPO of methane ................. 59

3.1 Introduction .................................................................................. 60 3.2 Experimental ................................................................................. 62 3.2.1 Catalysts ................................................................................. 62 3.2.1.1 Pd and Cu catalysts for the ten-fold cell........................... 62 3.2.1.2

Rh and Au/Rh catalysts for CPO of methane .................... 65

3.2.2 Parallel cell setup and experimental procedure ..................... 66 3.2.2.1 General setup ................................................................... 66 3.2.2.2

Ten-fold cell ...................................................................... 67

3.2.2.3

Six-fold cell ....................................................................... 69

3.2.3 Capillary Setup ........................................................................ 71 3.2.4 Parallel gas-phase reactor system (Celero) ............................ 72 3.3 Transformations in the ten-fold cell – proof of principle .............. 73 3.4 CPO in six-fold cell – catalysis in a parallel cell .............................. 79 3.5 CPO in the capillary cell ................................................................. 84 3.6 CPO in the Celero system .............................................................. 87 3.7 Discussion and comparison ........................................................... 89 3.7.1 Remarks to the ten-fold cell ................................................... 89 3.7.2 Catalytic results in six-fold cell................................................ 90

ix

3.7.3 Suggested pathway for CPO of methane ................................ 93 3.7.4 Cell comparison ...................................................................... 95 3.8 Summary ....................................................................................... 95

4

Dynamic processes in the CPO of methane – ignition and oscillations ............................................................ 99

4.1 Introduction to dynamic behavior............................................... 100 4.1.1 Ignition Processes ................................................................. 100 4.1.2 Oscillating Reactions ............................................................. 101 4.2 Experimental ............................................................................... 102 4.2.1 Catalysts ............................................................................... 102 4.2.2 Capillary Setup ...................................................................... 102 4.2.3 X-ray absorption spectroscopy ............................................. 103 4.2.3.1 Experiments with the FReLoN camera ........................... 103 4.2.3.2

XANES and time scans .................................................... 104

4.2.3.3

Quick-EXAFS ................................................................... 105

4.2.4 IR-Thermography .................................................................. 105 4.3 Ignition of CPO ............................................................................ 106 4.3.1 Visualizing the ignition.......................................................... 106 4.3.2 Quick-EXAFS during the ignition ........................................... 111 4.4 Oscillations in the CPO of methane on Pd/Al2O3 ......................... 113 4.5 Summary ..................................................................................... 122

5

Remarks and outlook ................................................. 125

5.1 5.2 5.3

Possibilities for gold in catalysis .................................................. 125 Future research on the CPO of methane .................................... 127 X-ray absorption techniques for catalysis ................................... 129

Literature ............................................................................. 131 List of publications ............................................................... 145 Publications related to this thesis ........................................................... 145 List of other publications ......................................................................... 147

Curriculum vitae ................................................................... 149

x

xi

Summary This doctoral thesis deals with two kinds of oxidation reactions, namely gold-catalyzed oxidation of alcohols in supercritical carbon dioxide and noble-metal-catalyzed partial oxidation of methane in gas phase. Both topics have received growing (re-growing in the case of methane oxidation) attention in the last years – gold catalysis for its novelty (the excitement about nanoparticulate gold being active as catalyst has not abated) and catalytic partial oxidation (CPO) of methane for its potential to be an important process in the future of synthesis gas production. The aim was to gain, by in situ spectroscopy and other analysis methods, a deeper understanding of underlying structure-performance relationships. In Chapter 1 some topics, which are of general importance for this work, are introduced. After some remarks about heterogeneous catalysis in general, history of and present interest in gold catalysis is described. Later, origin and production of synchrotron radiation are sketched, and X-ray absorption techniques are presented. In the last part, as for gold catalysis, former and present interest in the CPO of methane is discussed. Chapter 2 deals with gold-catalyzed alcohol oxidation in supercritical CO2. The standard reaction was the oxidation of benzyl alcohol to benzaldehyde; selectivities close to unity could be achieved. An additional base, which is generally necessary when using gold for alcohol oxidation, was not necessary in CO2, even without a basic support. High-pressure conditions are compared to reactions without a solvent and a variety of parameters such as temperature, amount of CO2 and oxygen pressure is tested. The catalytic activity proved to be higher in an expanded liquid than in a single, supercritical phase, and the variation of added oxygen pressure showed that conversion drops above a certain pressure. A subsequent structuresensitivity study yielded a volcano plot for conversion as function of particle size with a maximum at some 7 nm.

xii

CPO of methane is dealt with in Chapters 3 and 4. The focus in Chapter 3 lies on the development of a novel cell, with which multiple samples can be investigated by in situ X-ray absorption spectroscopy (XAS) in a parallel way. A first version is used to investigate structural changes of several copper and palladium catalysts upon treatment with different gas mixtures. The second version, which contains separate gas supplies for each reaction compartment, was used for steady-state analysis of CPO of methane on differently prepared rhodium and gold-doped rhodium catalysts. The influence of gold is described and the results obtained with the parallel cell are compared to those obtained with a capillary reactor and a parallel gasphase reactor system. A two-zone model for steady-state conditions of CPO of methane at moderate temperature was confirmed (mainly total oxidation in an oxidized zone at the beginning of the catalyst bed and reforming steps in a second, reduced zone), and qualitative agreement between the three very different reactor types showed the applicability of the new cell. In Chapter 4 the focus is shifted from steady-state to dynamic processes, namely ignition of the CPO of methane (on alumina supported Rh/Pt catalysts) – the switch from only combustion to formation of carbon monoxide and hydrogen at a characteristic temperature –, and oscillations around this ignition on alumina supported palladium catalysts. By using a high-speed X-ray area detector in combination with an IRthermography camera the formation and upstream movement of a front of reduction in the originally oxidized catalyst could be filmed. For this purpose the monochromator was set to the whiteline of platinum to be able to detect fast changes in the oxidation state. The process is reversible, i.e. during extinction of CPO of methane the transient between oxidized and reduced noble metal species is moving downstream. The reduction of single particles could be resolved and a reaction model for the ignition process was derived. In a certain range of temperature and gas flow the CPO of methane on Pd/Al2O3 shows periodical oscillation behavior. At constant heater temperature ignition and extinction periods alternate. By using a capillary microreactor, different X-ray absorption techniques and IR-thermography it could be shown that the ignition is quite similar to the one found on Pt/Rh,

xiii

but that stable operation is hindered by a too high overtemperature in the first, still oxidized part of the catalyst bed. The overtemperature causes selfreduction of palladium, which decreases the catalytic activity for total oxidation of methane. Consequently, the produced heat is not sufficient to fuel the reforming steps in the hind part of the catalyst bed and the reaction is extinguished, which causes the catalyst to be re-oxidized – the cycle can start again. In the last part, Chapter 5, I allowed myself to mix my thoughts about some of the used concepts – gold catalysis, partial oxidation of methane and the use of synchrotron-based X-ray absorption techniques in catalysis – with a subjective outlook on how the research in this thesis could be continued.

xiv

Zusammenfassung Die vorliegende Arbeit beschäftigt sich mit zwei Arten von Oxidationen; zum einen mit goldkatalysierter Alkoholoxidation in überkritischem Kohlendioxid, zum anderen mit edelmetallkatalysierter partieller GasphasenOxidation von Methan. Beide Themen rückten in den letzten Jahren immer mehr in den Fokus – bei der Methanoxidation ist es ein erneutes Interesse, sie stand in den 30er- und 40er-Jahren des 20. Jahrhunderts schon einmal im Zentrum des Interesses. Bei der Goldkatalyse liegt das wissenschaftliche Interesse in ihrer Neuheit begründet (die Aufregung darüber, dass nanopartikuläres Gold katalytisch aktiv ist, ist noch nicht abgeebbt), bei der partiellen Oxidation von Methan in ihrem Potential, in der Zukunft der Synthesegas-Produktion eine wichtige Rolle zu spielen. Ziel war es, mit Hilfe von In-Situ-Spektroskopie und anderer Analysemethoden ein gründlicheres Verständnis der zugrundeliegenden Beziehungen zwischen Struktur und katalytischer Leistung zu ermöglichen. In Kapitel 1 werden einige Themen eingeführt, die von Bedeutung für diese Arbeit sind. Nach kurzen Anmerkungen zur heterogenen Katalyse im Allgemeinen folgt eine Beschreibung der Geschichte der Goldkatalyse und des ihr gegenwärtig entgegengebrachten Interesses. Danach werden Ursprung und Erzeugung von Synchrotronstrahlung skizziert und Röntgenabsorptionstechniken vorgestellt. In einem letzten Unterkapitel werden, ähnlich der Goldkatalyse, früheres und heutiges Interesse an der partiellen Oxidation von Methan diskutiert. Kapitel 2 beschäftigt sich mit der goldkatalysierten Alkoholoxidation in überkritischem CO2. Als Standardreaktion diente die Oxidation von Benzylalkohol zu Benzaldehyd. Es wurden Selektivitäten nahe eins erreicht und die im Allgemeinen bei goldkatalysierten Alkoholoxidationen notwendige Zugabe einer Base konnte, auch ohne basisches Trägermaterial, durch die Verwendung von CO2 umgangen werden. Ergebnisse von Hochdruckexperimenten wurden mit denen lösungsmittelfreier Reaktionen

xv

verglichen und Parameter wie Temperatur, CO2-Menge und Sauerstoffdruck wurden variiert. In expandierten Flüssigkeiten zeigte sich eine höhere katalytische Aktivität als in einer einzigen, überkritischen Phase und durch Variation des zugegebenen Sauerstoffdrucks konnte gezeigt werden, dass der Umsatz oberhalb eines bestimmten Drucks fällt. Eine anschliessende Struktursensitivitätsstudie ergab für die Abhängigkeit des Umsatzes von der Goldpartikelgrösse ein Maximum bei ca. 7 nm. Die partielle Oxidation von Methan wird in den Kapiteln 3 und 4 behandelt. Der Schwerpunkt in Kapitel 3 liegt auf der Entwicklung einer neuen Reaktorzelle, mit der mehrere Proben parallel per In-SituRöntgenabsorptionsspektroskopie untersucht werden können. In einer ersten Version der Zelle wurden durch Behandlung mit oxidierenden und reduzierenden Gasatmosphären verursachte Strukturveränderungen verschiedener Kupfer- und Palladiumkatalysatoren untersucht. Eine Weiterentwicklung mit separater Gasversorgung für jeden einzelnen Reaktor wurde für die Analyse des stationären Zustands der partiellen Oxidation von Methan auf verschieden hergestellten Rhodium- und golddotierter Rhodiumkatalysatoren verwendet. Der Goldeinfluss wird beschrieben und die mit der Parallelzelle erhaltenen Resultate werden mit Ergebnissen eines Kapillarreaktorsystems und eines Hochdurchsatz-GasphasenReaktionssystems verglichen. Das Zwei-Zonenmodell des stationären Zustands der katalysierten partiellen Oxidation von Methan bei moderaten Temperaturen wurde bestätigt (hauptsächlich vollständige Oxidation in einer oxidierten Zone am Anfang des Katalysatorbettes und Reformierungsreaktionen in einer zweiten, reduzierten Zone) und die qualitative Übereinstimmung zwischen den drei sehr unterschiedlichen Reaktortypen zeigt die Anwendbarkeit der neu entwickelten Zelle. In Kapitel 4 verlagert sich der Fokus weg vom stationären Zustand hin zu dynamischen Prozessen. Zum einen die Zündung der katalytischen partiellen Oxidation von Methan auf aluminiumoxidgeträgerten Rh/Pt-Katalysatoren, zum anderen Oszillationen um diese Zündung herum auf aluminiumoxidgeträgerten Palladiumkatalysatoren. Als Zündung soll das Umschalten von nur Verbrennung auf die Bildung von Kohlenstoffmonoxid und Wasserstoff verstanden werden.

xvi

Unter Verwendung einer Hochgeschwindigkeits-Röntgenkamera in Kombination mit einer IR-Thermographie-Kamera konnte im ursprünglich oxidierten Katalysatorbett die Bildung und das Wandern einer Reduktionsfront in Richtung des Reaktoreingangs gefilmt werden. Um die schnellen Änderungen des Oxidationszustands detektieren zu können, wurde der Monochromator auf die „Whiteline-Energie“ von Platin eingestellt. Beim Löschen der Reaktion bewegt sich der Übergang zwischen oxidierten und reduzierten Platinspezies wieder in Richtung des Gasflusses aus dem Katalysatorbett hinaus. Durch die Möglichkeit, die Reduktion einzelner Katalysatorpartikel aufzulösen, konnte ein qualitatives Reaktionsmodell für die Zündung abgeleitet werden. In einem bestimmten Temperatur- und Durchflussbereich zeigt die palladiumkatalysierte partielle Oxidation von Methan periodischoszillierendes Verhalten – Zündungs- und Löschungsperioden wechseln sich bei konstanter Heizertemperatur ab. Unter Verwendung eines Kapillarreaktors, verschiedener Röntgenabsorptionstechniken und IRTomographie konnte gezeigt werden, dass die Zündung derjenigen der Rh/Pt-katalysierten Reaktion sehr ähnlich ist. Ein stabiler Betrieb wird aber dadurch verhindert, dass das sich im ersten, noch oxidierten Abschnitt, bildende Temperaturmaximum die Selbstreduktion von Platin bewirkt. Als Folge wird die katalytische Aktivität für die Methanverbrennung drastisch gesenkt und die erzeugte Wärme ist nicht mehr ausreichend, um die Reformierungsschritte im hinteren Teil des Reaktors zu versorgen. Folglich wird die Reaktion gelöscht und das Palladium wird reoxidiert – ein neuer Zyklus kann beginnen. Im letzten Teil, Kapitel 5, habe ich mir erlaubt, meine Gedanken einige der verwendeten Konzepte betreffend – Goldkatalyse, partielle Methanoxidation und die Verwendung von synchrotronstrahlungsbasierten Röntgenabsorptionstechniken in der Katalyse – mit einem subjektiven Ausblick zu erweitern und damit Möglichkeiten der Weiterführung dieser Arbeit aufzuzeigen.

1.1 Heterogeneous catalysis – some thoughts

1

1

General Introduction

1.1 Heterogeneous catalysis – some thoughts 1.1.1 From the beginning When students are asked for a definition of catalysis the regular answer could be summarized as a catalyst being a substance that accelerates a reaction, lowers the activation enthalpy, and leaves the chemical reaction without being changed. Now, all three characteristics are at least qualitatively inexact. Acceleration is only valid for the speed at which equilibrium is reached, and a catalyst only lowers the apparent activation energy by offering new reaction pathways (Figure 1.1), which it has to do to a certain degree to be capable of competing with the respective uncatalyzed reaction1. After a bond has been formed between substrate(s) and catalyst the chemical reaction takes place followed by separation of the product(s). From there probably the greatest opportunity in catalysis arises: The possibility to control selectivity by catalyst design. Addressing the third characteristic – catalytic processes would be much more facile, if catalysts were indeed able to be recovered unchanged. By moving from homogenous to heterogeneous catalysis both the complexity of offered reaction pathways and the risk of losing significant amounts of active material by restructuring in a wide sense or poisoning increase remarkably.

1

For solid-catalyzed gas-phase reactions the activation energy of the catalyzed process has to be at least 65 kJ/mol smaller than for the homogeneous reaction for the reactions rate to be equal (under the assumption that the number of solid-gas collisions is some 12 orders of magnitude smaller than the number for gas-gas collisions).

2

1 General Introduction

Figure 1.1: Reaction pathways for uncatalyzed and catalyzed reactions. The catalyst offers a new reaction pathway in the form of formation of an active complex, chemical reaction and separation. Ea,wc is the activation energy for the uncatalyzed process, Ea,t the true activation energy of the catalyzed process, and Ea,a its apparent activation energy.

Wrong or right – those three characteristics must be remaining marks of the evolution of catalysis. The origins are nebulous, but one of the first catalytic applications was the fermentation of barley to beer, which could be shown up to thousands of years ago. Another example is the small-scale production of sulphuric acid by burning sulphur and nitric acid in humid air, a “process” used in the Middle Ages. In the end of the 18th century it was shown, that this was indeed a catalytic process, nitre being the catalyst, although back then it was not named as such. The name came with J.J. Berzelius, who developed the idea of a “catalytic force” in his “Jahres-Bericht über die Fortschritte der physischen Wissenschaften” (Figure 1.2), in the chapter Pflanzenchemie (plant chemistry) under “Einige Ideen über eine bei der Bildung organischer Verbindungen in der lebenden Natur wirksame, aber bisher nicht bemerkte Kraft” (Some ideas about a force not yet noticed that is effective in the formation of organic compounds in the living nature).

1.1 Heterogeneous catalysis – some thoughts

3

Figure 1.2: Title page of the “Jahres-Bericht über die Fortschritte der physischen Wissenschaften”, in which Berzelius first published his idea of a catalytic force [1].

Berzelius reassessed earlier experiments as examples of catalysis, e.g. Thénard’s metal catalyzed dissociation of hydrogen peroxide, Kirchhof’s studies on acid catalyzed starch dissociation or the ability of platinum to catalyze oxidation reactions, which was investigated under different aspects by Humphrey Davy, Edmund Davy and Döbereiner, culminating in the sentence “Es ist also erwiesen, dass viele […] Körper […] die Eigenschaft besitzen, […] eine Umsetzung [zu] bewirken, ohne dass sie […] selbst Theil nehmen, wenn diess auch mitunter der Fall ist [sic].” (It is thus evidenced, that many bodies exhibit the capacity to cause a conversion without taking

4

1 General Introduction

part themselves, even though this might be the case). Thus, already the creator of the word acknowledged the possibility, that a catalyst does in fact not always leave a chemical reaction unchanged. Ostwald then was the first to base catalytic phenomena on thermodynamic and kinetic (rate of reaction) principles. In his presentation “Über Katalyse” at the “73. Naturforscherversammlug zu Hamburg“ in 1901 [2] he said „Es sind […] unter dem Einflusse von Katalysatoren keine Reaktionen möglich, die nicht auch ohne diesen Einfluss stattfinden könnten, ohne dass eines der Energiegesetze verletzt wird.“ (There are no reactions possible under die influence of catalysts, which would not be possible without this influence, without one of the energy laws being violated.) He summarized his definition of a catalyst in the sentence, which is cited in one or the other form in almost every text book on catalysis: “Ein Katalysator ist jeder Stoff, der, ohne im Endprodukt einer chemischen Reaktion zu erscheinen, ihre Geschwindigkeit verändert.” (A catalyst is every substance that changes the rate of a chemical reaction without appearing in the reaction products.) After the theoretical basis was developed (theory of chemical equilibria by van’t Hoff, rate of reaction, catalytic activity etc.) the principle of catalysis was exploited more and more rationally for industrial processes in the beginning 20th century – the contemporary understanding of chemical process engineering appeared. The first systematic and scientifically based catalytic process to be developed and optimized was the synthesis of ammonia. Haber succeeded in carrying out a successful experiment in 1905, but thought the results not promising enough for commercialization. Only after Nernst investigated and corrected the thermodynamic data that Haber used and reactor and process development by Bosch, iron catalyzed ammonia synthesis was an industrial success; up to then over 10000 catalysts were prepared and over 4000 tested. The process remained almost unaltered till the late 1950s.

1.1 Heterogeneous catalysis – some thoughts

5

1.1.2 Specialties of heterogeneous catalysis Despite its undisputed success and benefits, heterogeneously catalyzed systems exhibit a major complication, namely the existence of phase boundaries, because of which, in contrast to molecularly disperse homogeneous catalyst, macrokinetics come into play. The chemical reaction per se is joined by diffusion and adsorption/desorption steps; the whole circle is depicted in Figure 1.3. Substrates have to diffuse through a boundary layer from the bulk fluid to the catalyst surface (1) into the pores (2). Prior to chemical reaction on the catalyst (4), at least one of the educts has to adsorb on the catalyst surface (3). To keep the surface reaction running products have to desorb (5). The three energetic maxima in Figure 1.1 can thus be interpreted as transition states for adsorption, chemical reaction and desorption. The products eventually have to diffuse back to the bulk fluid (6,7).

Figure 1.3: Steps of a heterogeneously catalyzed reaction: Diffusion through a boundary layer between solid and fluid phase (1) into catalyst pores (2) and subsequent adsorption on the catalyst surface (3). After chemical reaction (4) the product(s) have to desorb (5) and diffuse back out of the pore (6) into the bulk fluid (7). The scheme is adapted from [3].

This model leads to two questions: How can sorption phenomena be involved in kinetic models and where does the reaction occur, i.e. the question of active sites. Already Faraday proposed in 1834, without

6

1 General Introduction

explaining catalytic action, that reactants have to adsorb simultaneously at the surface [4]. A heavily debated point in this topic in the beginning of the 20th century was the existence of an activation barrier for adsorption processes. It was not before Lennard-Jones published his potential curves [5] and Taylor’s publication “The activation energy of adsorption processes” from 1931 [6] that chemisorption was indeed considered an activated process (Figure 1.4).

Figure 1.4: Scheme of an activated chemisorption process. The Lennard-Jones potential curve shows the dissociation energy of an approaching molecule (ED) and the adsorption enthalpy of dissociative adsorption. The activation energy of chemisorption is the difference between zero level and intersection of the chemisorption and physisorption curves. n and m are exponents to describe repulsion and attraction, σ is correlated to a volume, and C describes the depth of the potential minimum.

After Langmuir described his adsorption isotherm model for chemisorptions in 1916 [7], which was extended by people like Temkin and Freundlich, some wider principles for heterogeneously catalyzed reactions could be formulated. The two most prominent ones were the LangmuirHinshelwood and the Eley-Rideal model. The first proposes that the surface

1.1 Heterogeneous catalysis – some thoughts

7

reaction of two chemisorbed species is the rate determining step and that all sorption processes are in equilibrium. The second assumes the reaction occurring between chemisorbed and physisorbed species. Their generalization then came with the Hougen-Watson approach, which is mostly used in chemical reaction engineering. Therein the rate of reaction is defined as the product of a kinetic term and a driving force, divided by an adsorption term to the power of participating active sites (n)2: 𝑟=

(𝑘𝑖𝑛𝑒𝑡𝑖𝑐 𝑡𝑒𝑟𝑚) ∙ (𝑝𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 𝑡𝑒𝑟𝑚) (𝑎𝑑𝑠𝑜𝑟𝑝𝑡𝑖𝑜𝑛 𝑡𝑒𝑟𝑚)

1.1

The question remains, what an “active site” actually is. While chemistry on the well-defined reaction sites of homogeneous catalysts and enzymes (when their structure is known) is relatively well understood, “processes occurring in heterogeneous catalysis are often obscure” [3]. H.S. Taylor’s theory of defect sites showing higher catalytic activity than flat surfaces in 1925 [9] started the idea of active sites. But in what form a catalytic site appears for a certain catalyst in a certain reaction is for many cases still unsolved, and, as Boudart and Djéja-Mariadassou state, “different types of active sites probably always exist, and a molecule may be adsorbed differently on each type and react at a different rate” [10]. Albeit much is known about specific sites today due to modern techniques, a great deal is still to learn and new models of catalytic sites must maybe be thought of, since, for instance, under high performance conditions the active sites “respond in a dynamical fashion to the chemical potential of the reactant phase” [11]. A third prominent concept in heterogeneous catalysis besides adsorption including kinetics and the nature of catalytically active sites was the separation into structure-sensitive and structure-insensitive catalysts, introduced by Boudart [12]. A catalyst is called structure-sensitive, if a 2

This model’s problem is that successful kinetic analyses do not at all guarantee the correctness of the „proven“ mechanism, since the sheer number of fit parameters for more complex reactions offers a wide range of possibilities [8].

8

1 General Introduction

change in particle size or exposed crystallographic face changes the catalytic activity. Boudart chose the low-pressure oxidation of CO on palladium as an example of a structure-insensitive reaction and ammonia synthesis on iron as a structure-sensitive one. This concept proved to be very useful in approaching the goal of catalysis research to understand structureperformance relationships of heterogeneous catalysts and thereby develop novel, designed catalytic systems. 1.1.3 Relevance for our world – why the research? It is estimated that about 90 % of all commercially produced chemical products have at least undergone one catalytic step, of which some 80 % are heterogeneously catalyzed [13]. About one trillion US-dollars are nowadays generated per year by these products. Thus, even small improvements in catalytic processes can lead to dramatic economic advantages. But, let economic aspects aside, or at least taking them as collateral, the challenge is on another level. What do 90 % of all commercially produced chemicals mean? –That almost everything around us is included. And again, most of those products stem from fossil raw materials, competing with the energy sector for finite resources. To say it bluntly: It’s a lot and it’s not lasting. The first part demands for selectivities as high and processes as “green” as possible to produce as little and as harmless waste as possible, the latter for innovative ways of using existing, fossil resources more efficiently and for adaption to new feedstocks. The IPCC climate change report 2007 lists “more selective catalysts for synthesis” [14] as an important means to reduce energy consumption. Weissermehl and Arpe wrote in their introduction to “Industrial Organic Chemistry” that the long range aims for securing industrial raw material for the chemical industry and energy supply must be to firstly extend “the period of use of the fossil raw materials”, and secondly to “[replace] the fossil raw materials in the energy sector” [15]. But changing feedstocks in chemical industry is far from simple, since catalytic processes were tuned for clean oil for decades. Companies like BASF have created Verbund-sites, where processes are designed in a way that in principle every side product is made use of. Other feedstocks like dirtier oil,

1.1 Heterogeneous catalysis – some thoughts

9

not to speak of biomass, pose a great challenge to catalysis research and development. To reach the necessary goals of more selective and efficient processes and adaption to new feedstocks a profound understanding of catalytic mechanisms has to be gained. An understanding of the catchy structureperformance relationships for “real” systems has still not been achieved in a general way. Next to basic and applied research thorough analysis and evaluation has to be performed to assess the achieved steps. Different approaches are available, e.g. integrated process assessment or the E-factor introduced by Sheldon in 1992, which is the ratio of waste mass and product mass [16].3 The final step would be consistent realization of the new processes, bringing the topic back into the regime of economy and politics.

3

Although not in an absolute way, fine chemicals and pharmaceuticals are the most inefficient chemical products, showing E-factor of up to 100. In comparison E-factors for bulk chemicals are in the range of 1 to 5.

10

1 General Introduction

1.2 Gold in Catalysis 1.2.1 History and hysteria The Encyclopædia Britannica starts its entry for gold in the following way: “Chemical element, a dense, lustrous, yellow precious metal […]. Gold has several qualities that have made it exceptionally valuable throughout history. It is attractive in color and brightness, durable to the point of virtual indestructibility, highly malleable, and usually found in nature in a comparatively pure form. The history of gold is unequaled by that of any other metal because of its value in the minds of men from earliest times.” [17] “From earliest times” has up to now been proven back to 4000 B.C., when gold artifacts were used probably as ritual objects in the Balkans. For Middle Europe the earliest gold findings date back to the Bronze Age (about 2000 B.C.). It has been used for cults, as jewelry, as currency (coinage metal), and as perpetual object of greed, be it the mythological chase for the Golden Fleece or the much more mundane conquest of the New World to find the legendary El Dorado. The overwhelming attractive power of gold, besides being easily workable, can be traced back to its inertness – it was found as native metal and stayed shiny. It does not corrode, since Al2O3 has a positive enthalpy of formation (+19.3 kJ/mol [18]), and is as a bulk so inert that it is one of the few metals, which is “neither essential nor toxic” for living beings [19]. Nevertheless, gold exhibits catalytic properties. Haruta is usually mentioned as the first one to bring gold into the scientific discussion as promising catalyst for CO oxidation [20], maybe with reference to Bond [21] and Hutchings [22]. But, the catalytic power of gold was employed much earlier. Berzelius, when introducing his catalytic force in 1836, describes experiments by Thénard, in which gold catalyzes the dissociation of hydrogen peroxide, and a study by Thénard and Dulong, which shows the catalytic formation of water by hydrogen oxidation, although a higher temperature was needed than with platinum and iridium [1]. The possibility of gold gauze and powder being catalytically active was discussed in the

1.2 Gold in Catalysis

11

1920s by different groups, e.g. by Benton and Elgin [23]. Between then and the rise of gold catalysis versus the end of the 20th century, G.C. Bond, after working through an exhaustive review by the Chamber of Mines of South Africa and the International Gold Cooperaction Ltd., sees quite a few mentionable papers and patents, which went almost unnoticed by the catalytic community [24]. It was not before the already mentioned finding of gold being a unexpectedly good catalyst for low-temperature CO oxidation by Haruta that over two decades gold moved to a prominent position in the catalytic community, which is quite evident in Figure 1.5, where the number of publications per year dealing with “gold” and “catalysis” is shown. From 12 publications in 1990 to 780 in 2009 is quite a steep rise. Interesting findings were linked to the unique physical and chemical properties of gold like the relativistic contraction of the 6s-shell combined with a full d-shell, the formation of strong aurophilic bonds or with a value of 343 kJ/mol the rather high sublimation enthalpy [25-26]. 800

Number of publications

700 600 500 400 300 200 100

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

1977

1975

0

Figure 1.5: Number of publications per year dealing with gold catalysis. Data was extracted from ISI Web of Knowledge, published by Thomson Scientific.

12

1 General Introduction

General reviews on gold as heterogeneous catalyst show the attention that gold catalysts have gained during the last two decades [27-31]. Different reactions have been tried – low-temperature CO oxidation (below 0 °C) being by far the most investigated one [28, 31-34]. Metal-support interactions were studied by employing almost every imaginable metal oxide support, combined with different preparation methods like impregnation, precipitation methods, cationic exchange and colloidal routes. Discussion about which oxidation state ought to be exposed by gold sites to be active was carried over from CO oxidation to other reactions like the water-gas shift or selective oxidation of alcohols. There are strong defenders of either Au0 [28, 32, 35-36] or oxidized gold [37-39] being the potent species. Theory and surface science experiments agree on the fact that at least for Au/TiO2 CO oxidation preferentially occurs at defect sites like kinks or edges – which should also to be true for dissociative adsorption of oxygen [36]. Selective oxidations were among others studied for propene [25, 40-42] and other alkenes [43-45], mono-alcohols [46-48] and diols/polyols [49-51]. Gold as catalyst is indeed an academically interesting topic and its industrial usefulness might increase in the next years. But, during the present hype, unfortunately, the keyword “gold” often enough seems to be the only justification for publication4 and it is difficult to restrain oneself from considering the importance of many results published in the name of gold at least as questionable. 1.2.2 Present research The application of gold as catalyst in industrial processes is presently not wide-spread. Examples are the production of vinyl acetate monomer and the production of carboxylic esters from primary alcohols, which was tested in a pilot plant [56]. Nevertheless, Thompson is convinced that “the attractiveness of using gold-based catalysts for commercial applications will increase as data continues to be accumulated on activity, poison resistance 4

As example one might take a series of four publications by Patil et al., all of them from 2004 in renowned journals, whose titles “Epoxidation of styrene by [anhydrous] t-butyl hydroperoxide over [supported gold catalyst]” only differ in the kind of support: Al2O3, Ga2O3, In2O3 and Tl2O3 [52], MgO [53], Yb2O3 [54], and TiO2 [55].

1.2 Gold in Catalysis

13

and durability” [26]. He presented a list of potential applications of gold ranging from chemical processes like the water-gas shift reaction or selective oxidations to clean energy generation by, for instance, using gold as an electrocatalyst for fuel cells. The World Gold Council is another player trying to push gold catalysis into an industrial perspective saying that “gold is set to take its place alongside the other precious metals […] as a key catalyst in a range of industrial processes and uses.” [57] Academic interest besides showing that heterogeneously gold catalyzed processes could be adequate for technical application [26], e.g. vehicle exhaust treatment, removal of volatile organic compounds or reduction of nitrogen oxides, is twofold. To gain a better understanding of the physicochemical behavior of gold and of how to utilize its unique properties for catalytic purposes – often by investigation of ideal systems and model calculations –, and to screen chemical reactions for their catalyzability by gold: oxidation reactions of all kind (alkanes, alkenes, aldehydes, sugars, synthesis of hydrogen peroxide), water-gas shift, hydrogenations (aromatic compounds, alkynes) and dehydrogenation; the list is far from being complete.

14

1 General Introduction

1.3 X-ray absorption spectroscopy for catalysis 1.3.1 Synchrotron radiation 1.3.1.1 Origin and production Synchrotrons are large facilities that nowadays are used to generate soft and hard X-rays, VUV-, and IR-radiation with enormous brilliance for scientific experiments. A typical synchrotron consist of a storage ring, where electrons – or sometimes positrons – are circulated in individual bunches, a linac (linear accelerator) to pre-accelerate the charged particles, a radiofrequency (RF) cavity to compensate for kinetic energy lost by synchrotron radiation, bending magnets and insertion devices (undulators and wigglers) at curved and straight parts of the storage ring, respectively, and beamlines, which are mounted tangentially to the storage ring, where experiments are performed. While at first generation facilities only parasitic synchrotron radiation originating from acceleration experiments was used, second generation synchrotrons were dedicated to its scientific use. Third generation sources like the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, the Swiss Light Source (SLS) in Villigen, Switzerland or PETRA III in Hamburg, Germany contain insertion devices in addition to bending magnets, opening up a vast amount of experimental possibilities for researchers of all kinds of fields. The latest development are fourth generation sources – free electron lasers. For this type of synchrotron radiation generators the principle of stimulated emission is employed, which implies that radiation is no longer generated from a storage ring but on a long linear path with modulating magnetic field. When synchrotron radiation was first discovered at the 70 MeV electron synchrotron at General Electric Research Laboratory (cf. Figure 1.6) in 1947 by Elder, Gurewitsch, Langmuir and Pollock [58], it was rather considered a nuisance than useful, for it was responsible for a large fraction of the energy loss of accelerated particles. The authors wrote that “the radiation is seen as a small spot of brilliant white light by an observer looking into the vacuum tube.” In telling how events unfolded this day Pollock shows the excitement of research:

1.3 X-ray absorption spectroscopy for catalysis

15

"On April 24, Langmuir and I were running the machine and as usual were trying to push the electron gun and its associated pulse transformer to the limit. Some intermittent sparking had occurred and we asked the technician to observe with a mirror around the protective concrete wall. He immediately signaled to turn off the synchrotron as ‘he saw an arc in the tube.’ The vacuum was still excellent, so Langmuir and I came to the end of the wall and observed. At first we thought it might be due to Cherenkov radiation, but it soon became clearer that we were seeing Ivanenko and Pomeranchuk radiation.” [59]

Figure 1.6: 70 MeV electron synchrotron at the General Electric Research Laboratory in Schenectady, NY. The arrow indicates the first sighting of synchrotron radiation in 1947. The picture is taken from [59].

Synchrotron radiation was indeed predicted by Ivanenko and Pomeranchuk in 1944, who derived it “in accordance with the classical electrodynamics, [since] one can easily see that quantum effects do not play here any important role as the dimension of the orbit is very great” [60]. When electrons or positrons – moving at relativistic speed – are forced to

16

1 General Introduction

follow curved trajectories by acceleration fields they emit electromagnetic radiation. Along a circular path the total emitted power of one electron sums up to 𝑃

=

𝑃(𝜆, 𝜓)𝑑𝜆𝑑𝜓 =

2𝑒 𝑐 𝐸 3 𝑅 𝑚𝑐

1.2

and the energy loss per turn and electron is 𝛿𝐸 (𝑘𝑒𝑉 ) = 88.5 where

𝜆 ψ P(𝜆,ψ)

R E e c mc2

𝐸 (𝐺𝑒𝑉) 𝑅 (𝑚)

1.3

wavelength of the emitted radiation azimutal angle power radiated by an electron in a unit wavelength interval centered at λ and in a unit vertical angular aperture centered at ψ bending radius of the electron orbit storage ring energy elementary charge speed of light in vacuum electron rest mass energy

The energy to the power of four in both equations shows that to run a synchrotron radiation facility efficiently it is necessary to go to larger radii. The radiation thereby produced has several advantageous properties [61]: (1) Continuum from the infrared to the X-ray region: The general shape of a radiation spectrum emitted by a bending magnet is shown in Figure 1.7, an important property being the critical wavelength (critical energy) λc. It divides the spectrum in two halves of equal energy and a rapid cutoff can be seen at λ < λc.

1.3 X-ray absorption spectroscopy for catalysis

17

Figure 1.7: Universal spectral distribution curve of the synchrotron radiation emitted by a bending magnet.

(2) High degree of collimation and spectral brilliance: The radiation emitted by charged relativistic particles is confined to a very small cone around the particle’s direction of flight – Figure 1.8 shows a comparison to a slow (classical) electron. The emission angle can be smaller than 0.1 mrad (about 0.006°). This behavior results in an extraordinary brilliance of synchrotron radiation sources (cf. Figure 1.9).

Figure 1.8: Qualitative radiation patterns to be expected from electrons in a circular orbit (a) at low energy and (b) as distorted by relativistic transformation at high energy, β being the fraction of light speed. The Figure is taken from [62].

18

1 General Introduction

Figure 1.9: Brilliance of different X-ray sources. XFEL is the X-ray Free Electron Laser, which is to be built at DESY in Hamburg, Germany. The picture is taken from [63].

(3) Well defined polarization in the orbital plane (4) Well-defined and tunable time structure Since the circulating electrons are grouped in bunches, the radiation is emitted in very short pulses typically below 1 ns. This characteristic is extremely useful when conducting time-resolved measurements. 1.3.1.2 Insertion devices In contrast to bending magnets, which are installed at curvatures of the storage ring and cause emission tangentially to the electron trajectory, insertion devices are installed in straight sections. Wigglers and undulators, as the two types are called, are both periodic magnetic structures, in which electrons are accelerated by periodical deflection of the electron beam (cf. Figure 1.10).

1.3 X-ray absorption spectroscopy for catalysis

19

Figure 1.10: Radiation beams from a bending magnet and from the individual poles of an insertion device. Taken from [64].

In a wiggler the vertically applied magnetic field of the multiple magnet forces the electrons to wiggle around a straight path in the horizontal plane. Their curved trajectory now has a smaller local curvature radius than the one of a conventional bending magnet. Thus, the observed radiation is the incoherent sum of the radiation emitted by each pole, which yields the same overall beam characteristics as those of a bending magnet, but with an Nfold intensity, N being the number of poles (Figure 1.10). Moreover, higher magnetic fields lead to an increased critical energy leading to an enhanced spectral range compared to a bending magnet. As said before an undulator is very similar to a wiggler. The characteristic difference is the ratio of the wiggling angle of the trajectory, α, and the angular aperture of synchrotron radiation, 1/γ (cf. Figure 1.11). For a wiggler this ratio is much larger than one and interference effects between radiation emitted by different poles is negligible. In an undulator the ratio is smaller than one (Figure 1.11), which means that α is smaller or in the range of 1/γ. Therefore, radiation emitted by the same electron at different points along

20

1 General Introduction

the trajectory can interfere constructively for certain wavelengths depending on α, λu, which is the oscillation period of the device, and the angle between the undulator axis and the detector. Due to this coherence phenomenon, the intensity compared to a bending magnet increases with N2. This and a very narrow angular distribution yield brilliances exceeding that of bending magnets and wigglers by several orders of magnitude – at the expense of spectral range.

Figure 1.11: Schemes of undulator and wiggler regimes taken from [65]. α is the wiggling angle of the trajectory, 1/γ the angular aperture, and λu the oscillation period.

1.3.2 X-ray absorption spectroscopy (XAS) Although interactions between X-rays and matter are manifold, we will concentrate on absorption. The energy of X-ray photons is in the range of core level binding energies of atoms. The rate of transition is determined by Fermi’s Golden rule 𝑅 =

2𝜋 ℎ

|〈𝑓|𝐻 |𝑖 〉| 𝛿 𝐸 − 𝐸 − ℎ𝜔 ,

1.4

which sums up the probabilities of all transitions from the initial state 〈𝑖 | (photon and atom) to the final states 〈𝑓| (no photon, ionized atom and leaving electron) with H1 being the dipole Hamiltonian (the delta function describes energy conservation).

1.3 X-ray absorption spectroscopy for catalysis

21

An important quantity when dealing with absorption processes is the linear absorption coefficient μ(E) given by the Beer-Lambert law: 𝜇(𝐸 ) ∙ 𝑑 = 𝑙𝑛

𝐼 𝐼

1.5

where d is the thickness of the sample and I0 and I1 are incident and transmitted radiation intensity, respectively. μ(E) depends strongly on the energy of the incident photon (∝ 𝐸 . ) and on the atomic number (∝ 𝑍 . ). Therefore, absorption of a single atom as function of photon energy is a line with negative slope, which is interrupted by absorption edges corresponding to excitation of core electrons from different shells. Each chemical element has specific edge energies, which is – besides the possibility to penetrate matter – with all its implications the reason for the application of X-ray absorption spectroscopy. In condensed matter the picture changes in so far that now interactions between valence electrons of the condensed atoms occur and a band structure is formed. Core electrons can then be excited to unoccupied local electronic states or to the continuum. Figure 1.12 shows a processed absorption spectrum of the Au L3-edge. A sudden increase in absorption indicates the edge region. Its maximum, if it exists, is called whiteline, since it yielded the brightest spot on photographic plates. It corresponds to the photoionization energy and marks the continuum threshold, i.e. around the edge the empty density of states is probed. Therefore, position and height of the whiteline are indicators for the oxidation state of the investigated metal. The region up to about 40 eV above the edge jump – different values in this order of magnitude can be found in literature – is called X-ray Absorption Near Edge Structure (XANES) and the spectral shape is mostly influenced by multiple scattering on neighboring atoms. The high-energy part of the spectrum is called Extended X-ray Absorption Fine Structure (EXAFS) and is caused by single scattering processes. XANES and EXAFS can, with careful processing and analysis, yield information on the local structure of a specific element.

22

1 General Introduction

Figure 1.12: Normalized and background-corrected X-ray absorption spectrum of gold foil. A rough distinction between XANES and EXAFS region is shown.

1.3.2.1 Extended X-ray absorption fine structure (EXAFS) The EXAFS region of an X-ray absorption spectrum opens up a way to gain information about the local structure of the sample, i.e. coordination and neighboring atoms. When the photoelectron is ejected it can be interpreted as a spherical outgoing wave that interacts with the neighboring atoms and is backscattered resulting in certain interferences between the outgoing and backscattered wave (Figure 1.13).

1.3 X-ray absorption spectroscopy for catalysis

23

Figure 1.13: Scheme of the interference of outgoing and scattered wave that causes the EXAFS phenomenon. The solid lines indicate the outgoing part, and the dashed lines the wave fronts scattered from the surrounding atoms (adapted from [66]).

This interference can be constructive or destructive leading to a modulation in the absorption coefficient around the single atom spectrum. These events are difficult, rather impossible, to explain in a classical way, since it would not be understandable how the behavior of an electron ejected “after” absorption of a photon should influence the absorption process itself. But, in the probabilistic view of quantum mechanics, as written before in formula 1.4, the situation presents itself as a combination of an initial state, final states, and a probability. Hence, the characteristics of the final state influence the probability of the transition, causing a modulation in absorption as function of energy. Or, in other words, EXAFS is the modification of absorption by the backscattered wave. Formula 1.6 is the standard expression for analyzing EXAFS: 𝜒 (𝑘 ) =

𝑁 𝑆 (𝑘 )𝐹 (𝑘 ) ∙ 𝑒 𝑘𝑅

∙𝑒 1.6

∙ 𝑠𝑖𝑛 2𝑘𝑅 + 𝛿 (𝑘) 𝜒(𝑘 ) is defined as the normalized oscillatory part of the absorption coefficient μ and is given by

24

1 General Introduction

𝜒 (𝑘 ) =

(𝜇 − 𝜇 ) 𝜇

1.7

where μ0 is the monotonously declining single-atom absorption. Each term on the right side of equation (1.6) is responsible for bringing a little piece of physical “reality” into the picture, although there are still enough restrictions. Nj is the number of atoms in the j-th coordination sphere. Rj is the average distance from the central atom to an atom in the j-th coordination sphere and k is the photoelectron wave vector, which is related to the photon energy, E, by

𝑘=

2𝑚 ℎ

(𝐸 − 𝐸 ).

1.8

E0 is the binding energy of the photoelectron. Fj(k) can be understood as the real part of the backscattered wave’s amplitude. The first of the two exponential terms accounts for the fact that neighboring atoms are not stationary; due to thermal vibrations and structural disorder the contribution of these atoms will not be in phase. In the term 𝑒 this disorder is 2 assumed to be small with a Gaussian distribution and σ , the well-known Debye-Weller factor, is the root mean square deviation from the average distance Rj. The second exponential term introduces the lifetime of the excited photoelectron state, which is influenced by both core-hole lifetime and the lifetime of the photoelectron itself. This is accounted for by phenomenologically interpreting the lifetime as the probability of the photoelectron wave to travel to the backscattering atom and back without scattering or the core hole being filled. S0 is a damping factor, and finally, the sine term is an expression for the phase shift of the backscattered photoelectron wave with wave number k - 2𝑘𝑅 is the phase shift introduced by traveling to the backscattering atom and back and 𝛿 (𝑘) is due to the varying potential of central and backscattering atom, which the electron feels on its way.

1.3 X-ray absorption spectroscopy for catalysis

25

By further mathematical implementation of physical reality the EXAFS equation can be made almost arbitrarily complex. Nevertheless, the standard formula described above is sufficient for basic fitting, which is crucial if more than qualitative information shall be extracted from recorded spectra. Such a spectrum has to be transformed to a 𝜒(𝑘 )-plot, which can then be Fourier-transformed in R-space to obtain absorption intensities as function of distance from the central atom, which was introduced by Stern, Lytle and Sayers in 1971 [67]. Due to the many variables and the damping of EXAFS modulations at high k-values, good data quality is of utmost importance. Besides, analysis of adequate standards and well-characterized reference samples is often the only way to obtain any useful information at all [66]. 1.3.2.2 X-ray absorption near edge structure (XANES) An assumption in EXAFS theory is that a high kinetic energy photoelectron is weakly backscattered by one of the neighboring atoms in a singlebackscattering process. In contrast, XANES deals with low kinetic energy photoelectrons (10 to 40 eV), which are strongly backscattered by neighboring atoms. These multiple-scattering processes in combination with allowed and not allowed transitions are able to give insight into relative positions of neighboring atoms, i.e. information on coordination geometry, which is not accessible by EXAFS. But this is on the cost of the demand for higher spectral resolution and much more complex calculations. In [68] Bianconi gives an equation for the absorption coefficient over the EXAFS and XANES range: 𝜇 (𝐸 ) = 𝜇 (𝐸) 1 + 𝜒 (𝐸 ) +

𝜒 (𝐸)

1.9

𝜇 (𝐸) is the atomic absorption, 𝜒 (𝐸) describes single-scattering processes (EXAFS), and ∑ 𝜒 (𝐸) sums up over all multiple-scattering processes of order n. While the EXAFS contribution is existent over the whole kinetic energy range of the photoelectron, the contributions of higher order practically disappear at high kinetic energy.

26

1 General Introduction

Several approximations that are very useful in EXAFS theory break down when dealing with XANES. One of them is the already mentioned singlescattering approximation. It fails to work at photoelectron wavelengths larger than the interatomic distance, i.e. if the wave vector 𝑘 is smaller than a critical wave vector 𝑘 =

, normally causing stronger and sharper

resonances than the EXAFS oscillations; the different final-state wave functions are shown Figure 1.14.

Figure 1.14: Sketched final-state wave functions in the core excitation in a diatomic molecule in EXAFS (high energies) and XANES region (lower energy). The dotted lines are the wave functions of the emitted photoelectron (taken from [68]).

1.3 X-ray absorption spectroscopy for catalysis

27

1.3.3 Usefulness for catalysis X-ray absorption spectroscopy offers some features, which can be considered useful for catalysis. It is element-specific, can penetrate matter (energy-dependent) and offers – provided good data quality and proper data processing – the possibility to gain local structural information. Elementspecificity usually allows resolving the surroundings of single elements without interference of other elements (nevertheless they will of course contribute to the background absorption) and the penetration depth, especially of hard X-rays (energies higher than about 12 keV) allows measurements through catalyst beds. In contrary to X-ray diffraction (XRD) the local structure, i.e. the structure of amorphous samples or, for instance, supported nanoparticles, can be measured as well. A drawback for some purposes is the missing surface-sensitivity, which can be introduced by grazing incidence measurements (GI-EXAFS) or Surface-EXAFS (SEXAFS), where Auger emission is measured. The range of possibilities lies between qualitative and fast obtainable information on oxidation states and elaborate elucidations of material structure. Novel methods are developed and more conventional ones are refined to access more specific information and to increase spatial and time-resolution. Examples are Dispersive-EXAFS (DEXAFS), Quick-EXAFS (QEXAFS), or Resonant Inelastic X-ray Scattering (RIXS).

28

1 General Introduction

1.4 Catalytic partial oxidation of methane Besides being the main constituent of natural gas deposits, methane is also extracted with oil and coal. Most of the methane that is released to the atmosphere is of anthropogenic origin. As it is a much more harmful greenhouse gas than carbon dioxide – it has a global warming potential of 72 and 7.6 calculated for 20 and 200 years, respectively –, dumping to the atmosphere should be avoided wherever possible. This is the reason why large amounts are flared on oil rigs all over the world – it can neither be processed on site due to a too small amount and the infrastructure for transportation to a centralized plant is missing nor can it be released as methane. The established process for transformation of methane – besides combustion for energy purposes – is steam-reforming leading to synthesis gas: CH4 + H2O



CO + 3H2

ΔrH° = +206 kJ/mol

CO + H2O



CO2 + H2

ΔrH° = +41.2 kJ/mol

Two different systems exist: catalytic steam reforming in tubular furnaces and autothermal reforming. While in the first case reaction heat has to be delivered by other sources (invented by BASF between 1926 and 1928 to produce hydrogen for ammonia synthesis), the latter produces heat by combustion of part of the feedstock. Methane and steam are fed to reaction tubes, which are heated by combustion of part of the feedstock to render the endothermic reactions possible. The catalyzed reaction occurs at about 850 °C on a nickel catalyst. Modern plants combine two reforming steps, tubular and autothermal reforming, which is “recognized as the most efficient syngas technology for large scale methanol plants at the moment” [69]. The problem with small amounts of methane is the inefficiency of steam reforming and gas-to-liquid (GTL) technologies at small scale. Catalytic partial oxidation (CPO) of methane could be a solution for this problem if run

1.4 Catalytic partial oxidation of methane

29

energy- and cost-efficiently. The first reference to CPO as a means to avoid wasting methane (natural gas) can be dated back to 1929. Liander wrote: Now, there is at the oilfields an abundant supply of natural gases which is utilized to a small extent only. It would seem that these natural gases would form an excellent war material for the manufacture of the initial gases for the ammonia process, and suggestions have been made to that end. [70]. A second aspect is the hydrogen to carbon monoxide ratio. For two largescale processes of growing importance, namely Fischer-Tropsch synthesis of alkenes nCO + 2nH2



Cn H2n + nH2O



CH3OH ,

and methanol synthesis CO + 2H2

this ratio is optimal at two. For classic steam reforming the ratio is equal or higher than three. Although combined reforming processes nowadays allow controlling this ratio in a certain range, they do not reach the value of two, which is achievable by direct partial oxidation: CH4 + 0.5O2  CO + 2H2

ΔrH° = -36 kJ/mol.

The reaction was investigated in the 1930’s and 1940’s, but carbon formation occurred on the metal catalyst, which could not be avoided without increasing explosion hazards or creating new problems. Therefore, and since oil was cheap, partial oxidation of methane “has […] been virtually ignored [during the next] 50 years” [71]. It was not before oil got more expensive that CPO of methane promised to become economically interesting causing a revival of scientific interest. The CPO of methane has been studied on quite a variety of transition metal catalysts with nickel, palladium, platinum, rhodium and iridium being the most prominent ones [72]. Investigated aspects were the activation of methane dissociation (difficulty of CH3-H bond cleavage), binding site preferences, surface residence times of active species, contributions from

30

1 General Introduction

lattice oxygen, and influence of the support – in principle the questions that arise with every heterogeneously catalyzed reaction. Complicating characteristics are the usually very high temperature (up to 1000 °C) and the very short contact time (ms-range). A critical point is the transferability of different results. Since CPO of methane is a reaction with technical potential, each study dealing with this reaction tries to – or probably has to – extrapolate the findings to the technical process. But, this has to be done in an extremely careful way: Different temperature control (adiabatic, isothermal, polytropic), pretreatment of the catalyst, reaction control, and catalyst formulation (packed bed, monolith etc.) can yield a multitude of results. Comparison of different reactor systems and a critical discussion of the respective limitations are therefore in order.

1.5 Aim of the Thesis

31

1.5 Aim of the Thesis Heterogeneously catalyzed oxidation reactions are commonly used in industry to convert cheap substrates into valuable products. Understanding of structure-performance relationships is crucial for the optimization of already applied processes or the development of new competitive ones. As a main objective of this doctoral thesis catalytic oxidations in different phase regimes (gas phase and supercritical fluids) will be investigated and monitored by in situ spectroscopic methods: (i) Gold-catalyzed reactions have received increasing attention during the last years; especially the low-temperature oxidation of CO and the aerobic oxidation of alcohols were widely investigated. Nevertheless, the conditions for optimized Au catalysts have not yet been unraveled. A possible step in this direction as well as the possibility of new, efficient processes shall be explored by the use of supercritical carbon dioxide as solvent in batch and continuous flow experiments. A first step is the transfer of alcohol oxidation from liquid to supercritical media; a second one could be the extension to other oxidation reactions, e.g. the industrially more important epoxidation of alkenes and the oxidation of monoterpenes. Structural questions concerning gold catalysis and mechanistic ones concerning the whole systems shall be tackled by in situ spectroscopic methods. (ii) The catalytic partial oxidation (CPO) of methane could be a means to use abundant natural gas even in small quantities (common side product of oil production) as energy source via conversion to synthesis gas. To gain more information on the actual, still controversial reaction mechanism, a new way of 2-dimensional structural monitoring by XAS was used. The examined system was a fixed-bed capillary reactor with Rh on Al2O3 as catalyst. In the thesis the understanding of this reaction shall be deepened by spectroscopic studies of the reaction dynamics, the transfer of our model to different reactor types, and the development of a new, spectroscopic screening cell to investigate the influence of other transition metals (e.g. gold) on the catalytic behavior.

32

1 General Introduction

1.5 Aim of the Thesis

2

33

Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

The oxidation of benzyl alcohol to benzaldehyde over different supported gold catalysts in supercritical carbon dioxide has been investigated in a highpressure batch reactor. Only molecular oxygen was used as oxidant and no base was needed. Different supports and preparation methods for the catalysts were tested and parameters like reaction temperature, pressure and molar ratios of the components were varied to study the catalytic behavior. Gold colloids deposited on a titania support (1wt%Au/TiO2) yielded a conversion of 16.0% after 3 h and a high selectivity to benzaldehyde of 99% under single-phase conditions. The reaction rate was significantly higher than in a corresponding ‘‘solvent-free’’ reaction without CO2. Even higher rates were found when a CO2-expanded phase was present. Monitoring of the oxidation in a high-pressure view cell via infrared transmission spectroscopy unraveled a slowdown of the reaction rate above 15% conversion. By using supported gold nanoparticles of different size (1.3 to 11.3 nm), it was found that the particle size has a significant effect on catalytic performance – particles with an average diameter of 6.9 nm expose a maximum, with activity dropping for smaller and larger particles. In addition, 1-octanol and geraniol were oxidized as well under similar conditions, yielding conversions of 4% and 10%, respectively, with selectivities towards octanal and geranial of 90% and 30%. Thus, the combined application of gold-based catalysts and supercritical CO2 offers an interesting alternative to the known methods of alcohol oxidation.

34

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

2.1 Introduction 2.1.1 Catalysis in supercritical carbon dioxide The use of supercritical carbon dioxide (scCO2) in industry – the critical parameters being Tc = 304.2 K, pc = 7.38 MPa, and ρc = 10.6 mol/l – is mostly confined to extraction purposes like decaffeination or oil recovery in mature oil fields. Smaller application areas are the production of nanoparticles for pharmaceutical products via processes like GAS (gas anti-solvent), PGSS (particles from gas-saturated solutions) and RESS (rapid expansion of supercritical solutions), and its use as refrigerant for domestic heat pumps. The utilization as supercritical solvent or reagent in chemical reactions for industrial processes is still, despite enormous effort by academia to prove its usefulness, not common. Nevertheless, Pereira and Leib go a little too far by writing in “Perry’s Chemical Engineers’ Handbook” that they “are not aware of any industrial implementation of supercritical conditions in reactors” [73]. Working processes exist, e.g. fluoropolymer synthesis by DuPont or hydrogenation by Thomas Swan, the number of companies employing scCO2 is growing [74], and reactors for catalyzed and uncatalyzed reactions in scCO2 are on the market [75]. However, it is true for the following reasons that industry does still not warmly welcome processes dealing with supercritical fluids. • Reactions, in which supercritical fluids are actually beneficial in a way that a traditional process should be replaced, are up to now very rare. • Although called “green” by scientists, many processes turn out to be economically not feasible, when looking into process design and costing [76-77]. • Engineering demands are major both from a mechanical (redesign of components as high pressure version) and a chemical engineering standpoint (lack of thermodynamic multi-component data for systems in dense CO2) [74]. But, catalyzed reactions in scCO2 were and still are widely investigated to gain a better understanding and hopefully, in the end, open up the

2.1 Introduction

35

possibility for innovative new processes [78-80]. Table 2.1 lists advantages and disadvantages, which are generally thought of when considering scCO2 as solvent for catalysis. Table 2.1: Properties of scCO2 that are generally discussed when thinking about its suitability as solvent in catalytic reactions. Advantages - inert in oxidations - relatively benign - generally immune to free radical chemistry - miscible with gases in all proportions above 304 K - easy tunability of reaction environment - low viscosity - high diffusivity

General properties - aprotic solvent - can be catalytically hydrogenated to CO - low pH (2.85) upon contact with water

Disadvanatages - relatively high critical pressure - low dielectric constant - can be catalytically hydrogenated to CO

Regarding this work scCO2 as solvent was chosen, since it allows performing the oxidation of alcohols in an inert (in respect to oxidation) and non-flammable medium at potentially high reaction rates [78, 81-82]. It is very convenient that by adjusting pressure and temperature the solvent and reactants can be either assembled in a single reaction phase or the reaction can be performed in a CO2-expanded reaction phase, where the oxygen solubility and the mass transport properties are still high [83-84]. The properties of pure carbon dioxide are by now pretty well described. A fundamental equation, from which all thermodynamic parameters – presently up to 1100 K and 800 MPa – can be obtained, was published by Span and Wagner in 1996 [85] involving reduced density and temperature and functions of second and third virial coefficients. But, additional compounds complicate the phase behavior. A p-T diagram of a binary mixture of similar compounds contains a critical line and a binary mixture of compounds with very different critical temperatures develops a three-phase line, shown in Figure 2.1a and b, respectively. In a regular catalytic reaction system at least three components are present with at least two of them

36

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

showing a time-dependence in concentration. Thus, for new systems qualitative phase behavior studies are generally advisable.

Figure 2.1: p-T diagrams of two-component mixtures. a) shows phase equilibria of two components with similar critical temperatures. In (b) the two critical temperatures are very different leading to a three-phase line (S=L=G). CP denotes critical points and UCEP is the upper critical end point, the intersection between critical and three-phase line. The dashed line in b) is the former S=L line of the pure substance.

2.1.2 Gold-catalyzed alcohol oxidation Apart from noble metals like palladium or platinum, gold-based catalysts have also been used for the heterogeneously catalyzed oxidation of alcohols and hydrocarbons [42, 86-87]. In spite of the lower reactivity of gold this approach can yield higher selectivities than achieved over supported Pt or Pd catalysts [50, 87-89]. Alcohol oxidation is typically performed in liquid phase. In case of water-soluble alcohols water can be used [90-91], but in case of gold catalysts the addition of a base is required to facilitate the dehydrogenation step [92-94]. Organic solvents are also applied. However, in the spirit of an environmentally benign process, solvents should be minimized and simple oxidants should be used5. Thus, solvent-free processes [89, 93, 95] in combination with simple and non-toxic oxidants like hydrogen peroxide and oxygen/air [56, 90, 93, 95-99] attracted attention. But the reaction rates are typically lower, particularly since no base can be used [93, 95]. 5

One should be careful in misjudging water a more benign solvent than conventional organic solvents. It is much more cost- and energy-intensive to recover water than organic solvents.

2.2 Experimental

37

An alternative approach is the employment of supercritical carbon dioxide as an environmentally benign solvent, which was recently used for selective alcohol oxidations [100-106]. We thus investigated the combination of the two concepts – gold catalyst and dense CO2. We focused on the evaluation of alcohol oxidation in the absence of an assisting base using supported gold catalysts and molecular oxygen as oxidant. The oxidation of benzyl alcohol was taken as a model reaction. Gold catalysts on different supports (titania, iron oxide and activated carbon) were prepared using different synthesis routes and were tested in a high-pressure batch reactor. With selected catalysts the influence of various reaction parameters and the extension of the approach to other, structurally different alcohols were investigated. To check for structure-sensitivity different sizes of gold nanoparticles supported on titania and ceria were compared in the oxidation of benzyl alcohol to benzaldehyde. A frequent problem in structure-sensitivity studies performed by varying the particle size of the catalytic component is that different samples are not prepared in the same way. By variation of preparation route and pretreatment procedures the isolation of particle-size effects is extremely complicated – if not impossible. Therefore, only one preparation route was used to minimize impurities and residues on the catalytic surface and to avoid morphology alteration [107-108]

2.2 Experimental 2.2.1 Catalysts Preparation The catalyst preparation included deposition of gold colloids [35], impregnation and flame synthesis [109]. 1%Au/TiO2(coll), 1%Au/Fe2O3(coll) and 1%Au/C(coll) were prepared via gold colloids by adding a colloidal solution of gold particles with an average diameter of 2.1 nm (102 ml) to 100 ml of a suspension containing 2 g of the respective support. The support suspension was stirred for 15 min at 750 rpm. Before adsorption of the colloids, the pH of both suspensions was adjusted to pH 2 by addition of sulfuric acid. Used supports were titania P25 (Degussa AG, BET surface area of 49 m2/g) and activated carbon (ABCR, BET surface area of 1350 m2/g).

38

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

After stirring for 15 min, the resulting suspension was filtered and subsequently the filter cake was washed with water until the filtrate did not contain any chloride ions (checked by addition of AgNO3(aq)). The solution of gold colloids was prepared similarly as described in [35] by addition of 3.58 ml 28.84 mM HAuCl4(aq) (ABCR, 51% Au) to a 100 ml flask containing 93.4 ml deionized water, 3 ml 1 M NaOH(aq) and 2 ml 0.95% aqueous solution of tetrakis(hydroxymethyl) phosphonium chloride (THPC) (Fluka, 80% in water). The colloidal solution was stirred for 15 min at 750 rpm before its addition to the support suspension. 1%Au/TiO2(coll) was used asprepared (dried at 80 °C for 15 h) and after calcination at 400 °C for 4 h. For the preparation of Au/TiO2 and Au/CeO2 catalysts for the structuresensitivity study the above described colloidal route was modified for particle-size control. For the standard colloid (mean diameter of 2.1 nm) an equimolar ratio of THPC and the Au precursor were employed. This ratio was changed to n(THPC)/n(Au) = 4 and 0.8 to obtain particles with mean diameters of 1.3 nm and 6.9 nm, respectively. Larger particles with a mean diameter of 11.3 nm could be prepared by addition of 1 ml of the HAuCl4 solution and a stoichometric amount of THPC to the original colloidal suspension. The NaOH/(THPC+Au) molar ratio was kept constant at 3, except for the smallest particles, for which a value of 6 was used. Another titania supported gold catalyst was prepared by impregnation of 1 g titania with 1.79 ml 28.84 mM HAuCl4(aq) and drying at 80 °C for 10 h. The catalyst was used as-prepared (dried at 80 °C for 15 h) and after calcination at 400 °C for 4 h. In addition to 1%Au/Fe2O3(coll) an iron oxide supported gold catalyst was synthesized using flame spray pyrolysis [109110]. For this purpose, a 0.25 M solution of Fe(III)acetylacetonate and a 1 mM solution of HAuCl4, each in a 1 : 1 mixture of methanol and acetic acid were sprayed in a methane/oxygen flame. The parameters were similar to those reported in [109]. 2.2.2 Catalyst Characterization The estimation of Au particle-sizes and their dispersions was performed by scanning transmission electron microscopy (STEM) using a Tecnai F30

2.2 Experimental

39

microscope (FEI, Eindhoven). To improve the z-contrast, high-angle annular dark field detection (HAADF-STEM) was employed. The particles were qualitatively analyzed by energy dispersive X-ray spectroscopy (EDXS), for which the detector was attached to the Tecnai F30 microscope. A real gold loading of 0.5% to 0.8% was determined by AAS for all gold catalysts – catalytic results were corrected by these values. The oxidation state of the particles was additionally determined by X-ray absorption spectroscopy (XAS) in fluorescence mode at the Swiss Norwegian beam line (SNBL) at ESRF (Grenoble, France). The spectra were recorded around the Au L3-edge between 11.90 and 12.00 keV. The raw data were energy-calibrated, background-corrected, normalized and smoothed using the WinXAS 3.1 software [111]. XANES and EXAFS for the particle-size study were recorded at the synchrotron facility ANKA in Karlsruhe, Germany, at the XAS beamline using a Si(111) double-crystal monochromator detuned to about 60% of the maximum intensity. An ionization chamber was used to determine the incoming X-ray intensity. EXAFS spectra were recorded in fluorescence and transmission mode. A five-element Ge solid-state detector (Canberra) was applied to measure the Au Lα fluorescence (fluorescence window from 9.60 keV to 9.75 keV; excitation at the Au L3-edge at 11.919 keV). Samples were pressed as pellets and positioned in the beam at an angle of 45°. EXAFS spectra were processed using Athena [112]. The surface area of the catalysts was calculated from the nitrogen adsorption isotherm according to the BET method. 2.2.3 Reaction Setup The high-pressure batch reactor experiments were conducted in a 100 ml autoclave made of Hastelloy B (Premex Reactor AG, Lengnau); the reactor is depicted in Figure 2.2. Liquid alcohol and the powdered catalyst were put in the reactor, which was then closed and linked to the gas lines. Oxygen (99.5%) was added at 23 °C until the appropriate pressure was reached. Finally, CO2 (99.9%) was added with a compressor (NWA GmbH, Lörrach) from a CO2 gas cylinder equipped with a dip tube. The amount was measured by a mass flow controller (RHE2, RHEONIK Messgeräte GmbH). The autoclave was heated to the reaction temperature (80-120 °C) and, after

40

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

the respective reaction time, cooled down for half an hour by a water cooling system. Samples were taken by dissolving the reactor contents in 50 ml toluene (Fluka, puriss. p.a.).

Figure 2.2: Autoclave made of Hastelloy B. In single-phase experiments the 100 ml reactor was filled with 0.1 g of the respective catalyst and the alcohol before closing. After O2 was added until the appropriate pressure was reached and before heating to the reaction temperature, the desired amount of CO2 was added via a mass flow controller.

After filtration of the catalyst, the reaction mixture was injected in a gas chromatograph (HP 6800 series) equipped with FID detector and HP5 GCcolumn (Agilent Technologies Inc., 25.0 m·x 320 μm·x 0.17 μm). The reactant composition of standard experiments was 7 mmol of alcohol, 14 mmol of oxygen and 1590 mmol CO2. During repetitive experiments the conversion and selectivity could be reproduced with a margin of error of ±6% and ±0.4%, respectively. To check for gold leaching the filtered solution was concentrated and the residue was dissolved in 10 ml aqua regia and diluted with deionized water. The samples were measured in a flame atomic absorption spectrometer (Varian SpectrAA 220FS) equipped with a Gold

2.3 Alcohol oxidation on gold in scCO2

41

UltrAA HC lamp (Varian, ke= 242.8 nm). A gold AAS standard solution was used (Fluka, c(Au) = 1 g/L) for calibration. The detection limit was 1 μg/ml. 2.2.4 Phase behavior and IR measurements Phase behavior studies were performed in a magnetically stirred highpressure view cell with a variable volume of 22 ml to 62 ml [113]. In addition, a spectroscopic view cell was used [114], which was equipped with a digital camera for phase monitoring, ZnSe windows for IR-transmission measurements and a ZnSe ATR crystal at the bottom of the cell. The crystal was coated with 8 mg of 1%Au/TiO2(coll). The Fourier transform infrared (FTIR) spectra were recorded with a Bruker IFS-66 spectrometer, equipped with an MCT detector. A detailed description of the setup can be found in [114]. Benzyl alcohol, oxygen and carbon dioxide were inserted in the same way as described above for the catalytic experiments.

2.3 Alcohol oxidation on gold in scCO2 2.3.1 Results Phase behavior studies using video monitoring in a high-pressure view cell with a sapphire window were used to discriminate the two-phase region and that with a single reaction phase. This allows relating different catalytic performances of the reaction system at different pressures to the respective phase compositions. For this purpose, different mixtures of CO2, O2 and benzyl alcohol were investigated between room temperature and 120 °C and pressures up to 180 bar. Systems with an amount of both alcohol and oxygen 99% at 120 °C. Table 2.3: Conversions, selectivities and reaction pressures of the oxidation of benzyl alcohol to benzaldehyde on 1%Au/TiO2(coll) at different reaction temperature after 3 h (7 mmol alcohol, 14 mmol oxygen and 1591 mmol CO2). Reaction temperature [K]

Reaction pressure [bar]

Conversion [%]

Selectivity to benzaldehyde [%]

353 373 393

133 150 167

9.6 16.0 34.7

93.6 99.0 99.2

In addition to the standard experiment using 1%Au/TiO2(coll) with a reaction time of 3 h, longer time intervals of 6, 12 and 20 h were used (Table 2.4). After 20 h a conversion of 30% was obtained. To check for a possible deactivation, the catalyst was recollected after a 3 h experiment and reused for another reaction; only a conversion of 7% was observed after 3 h. Table 2.4: Conversions and selectivities to benzaldehyde for the selective oxidation of benzyl alcohol on 1%Au/TiO2(coll) under single-phase conditions (100 °C and 150 bar) for different reaction times. Used amounts were 7 mmol benzyl alcohol, 14 mmol O2, 1591 mmol CO2.

a

Reaction time [h]

Conversion [%]

Selectivity to benzaldehyde [%]

3 6 12 20

16.0 16.9 24.1 30.1

99.0 99.0 >99.0 >99.0

3a

9.1

98.3

Addition of 11 mmol of H2O before the reaction

In order to monitor the reaction in a time-resolved manner, the selective oxidation of benzyl alcohol was performed in a high-pressure view cell equipped with IR transmission windows and an ATR-IR crystal on which 8 mg of the catalyst were coated. 2.3 mmol benzyl alcohol, 4.6 mmol O2 and

2.3 Alcohol oxidation on gold in scCO2

47

523 mmol CO2 were used resulting in a total pressure of 150 bar at 100 °C. As can be seen in Figure 2.6 first an induction period of approximately 2 h occurred, similarly as in the oxidation of benzyl alcohol in scCO2 over Pd/Al2O3 [121]. Then a linear increase of conversion followed and a slight deceleration occurred after ca. 15 h. 25

Conversion [%]

20 15 10 5 0 0

5

10

15

20

25

30

35

40

45

50

Time [h] Figure 2.6: Monitoring of the selective oxidation of benzyl alcohol to benzaldehyde in a high-pressure view cell by IR transmission spectroscopy; plot of alcohol conversion over reaction time in hours. Used amounts were 2.3 mmol benzyl alcohol, 4.6 mmol O2 and 523 mmol CO2 at 150 bar and 100 °C. The ATR-IR crystal was coated with 8 mg of 1%Au/TiO2(coll).

A change in the amount of oxygen and by that in the composition of the reaction mixture yielded a lower conversion for both less and more than 0.9 mol% (Table 2.5); the amount of CO2 was also changed to keep the pressure at 150 bar. Although an excess of oxygen is still available in the experiment with an O2 molar concentration of 0.4% (n(O)/n(alcohol) = 2), the conversion dropped from 16% at 0.9 mol% O2 to 12%. Increasing the initial oxygen pressure to 20 and 30 bar also lowered the conversion to 11.7% and 10.9%, respectively. The selectivity towards benzaldehyde ranged in all four cases between 97% and 99%. To check the applicability of 1%Au/TiO2(coll) in the high-pressure reaction system for different alcohols, 1-octanol and geraniol were chosen for oxidation under the same conditions as used for

48

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

the oxidation of benzyl alcohol. 1-Octanol is known to be not very reactive in heterogeneously catalyzed oxidations [93, 98, 122] and geraniol can yield various by-products, since it has several functional groups, that can be oxidized [105]. The conversion for 1-octanol was 4.1% at a selectivity of 90.4% towards octanal. The selective oxidation of geraniol yielded a conversion of 10.9% and selectivity towards citral (combination of the two isomers geranial and neral) of 30.6%. The use of 1%Au/Fe2O3(coll) resulted in a higher conversion of 20.3% at similar selectivity. Note again that the system was not optimized for the highest possible performance. Table 2.5: Conversions and selectivities to benzaldehyde for different molar ratios of oxygen in the selective oxidation of benzyl alcohol on 1%Au/TiO2(coll). Amount of

Molar ratio

Amount

Conversion

Selectivity to

oxygen [mmol]

of oxygen [%]

of CO2 [mol]

[%]

benzaldehyde [%]

7 14 81 122

0.4 0.9 6.0 10.6

1.61 1.59 1.27 1.02

12.0 16.0 11.7 10.9

98.2 99.0 98.6 97.4

The amount of alcohol was kept constant at 7 mmol, the amount of CO2 was adjusted to maintain a reaction pressure of 150 bar.

Whereas hitherto experiments, except the ones at elevated oxygen pressure, were all conducted under single-phase conditions, the study of the reaction performance at different pressures, i.e. different amounts of added carbon dioxide, is accompanied in several cases by phase separations (cf. Figure 2.3). This strongly affected the performance as reflected by the results in Table 2.6. As already mentioned above, the conversion of the system 1%Au/TiO2(coll)-O2-alcohol without CO2 was 7%. When 10 g of CO2 were added, corresponding to a content of 91.6 mol% and a total pressure of 39 bar at 100 °C, the conversion increased to 22.4%, whereas the selectivity remained almost unchanged (95.2% vs. 96.4% upon CO2 addition). Addition of 30 g CO2 (97.0 mol%, 90 bar) resulted in an equally high conversion of 23.6% and a selectivity of 97.0%. When 70 g (150 bar) and 90 g (177 bar) of carbon dioxide were used (98.7 mol% and 99.0 mol%), the conversion

2.3 Alcohol oxidation on gold in scCO2

49

decreased to 16.0% and 16.3%, respectively, whereas the selectivity to benzaldehyde increased to 99% and >99%. Table 2.6: Conversion, selectivity to benzaldehyde and reaction pressure for different molar ratios of carbon dioxide in the selective oxidation of benzyl alcohol on 1%Au/TiO2(coll). Amount of CO2 [mol]

Amount of CO2 [g]

Reaction pressure [bar]

Conversion [%]

Selectivity to benzaldehyde [%]

0 0.23 0.68 1.59 2.05

0 10 30 70 90

3.4 39 90 150 177

7.0 22.4 23.6 16.0 16.3

95.2 96.4 97.0 99.0 98.0

The amounts of alcohol and oxygen were kept constant at 7 and 14 mmol, respectively.

2.3.2 Discussion Our studies demonstrate that supercritical carbon dioxide has a beneficial effect on the reaction rate of the gold catalyzed selective oxidation of benzyl alcohol. Conversions of 16.0% and 10.4% (TOF: 161 h-1 and 170 h-1) for 1%Au/TiO2(coll) and 1%Au/Fe2O3(coll), respectively, under single-phase conditions were higher than those achieved in the reference experiment without CO2 and solvent-free oxidation reactions reported in literature [93, 95]. The selectivities were higher than for other noble metals like platinum or palladium [89, 104], and in several cases close to 100%. Using scCO2 as solvent, aerobic alcohol oxidation with molecular oxygen as sole oxidant could be performed and the addition of an assisting base circumvented [123128]. However, base-catalyzed oxidations typically showed higher reaction rates, particularly in the beginning of the reaction [94, 129]. Despite the fact that carbon dioxide has a beneficial effect on the reaction rate, single-phase conditions were not the most optimal ones. Table 2.6 shows that the rate is highest under conditions where a CO2-expanded reactant phase is present, which is depicted clearer in Table 2.7 and Figure 2.7. When the amount of added CO2 is increased starting from a solvent-free

50

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

reaction, conversion increases to over 20% in biphasic conditions and drops when a single phase is reached. A similar behavior has been observed in other studies [105-106, 130] and can be attributed to the higher specific density of liquid carbon dioxide compared to the supercritical state [1 kg/dm3 to 0.5 kg/dm3]. The resulting higher solubility can be beneficial for two reasons: (1) faster removal of by-products from the catalyst surface (water in the case of alcohol oxidation) and (2) if the rate determining parameter is not mass transport through a phase barrier, but the solubility of the reaction gas [74]. Table 2.7: Conversion and selectivity to benzaldehyde for Au catalysts on different supports. Single-phase conditions were obtained by using 7 mmol benzyl alcohol, 12 mmol O2 and 1591 mmol CO2 (70 g), while only 682 mmol (30 g) were added for a biphasic system. The two figures attached to the table columns are a repetition of Figure 2.3 to visually clarify single-phase and biphasic conditions. Single-phase Catalyst

Biphasic

Conversion

Selectivity to

Conversion

Selectivity to

[%]

benzaldehyde [%]

[%]

benzaldehyde [%]

1%Au/TiO2(coll)

16.0

99.0

23.6

97.0

1%Au/CeO2(coll)

15.0

99.0

24.0

97.0

1%Au/Fe2O3(coll)

10.4

99.0

18.5

99.0

2.3 Alcohol oxidation on gold in scCO2

51

Figure 2.7: Conversion and selectivity to benzaldehyde for Au catalysts on different supports. The biphasic regime is shown on the left. Used amounts were 7 mmol benzyl alcohol, 14 mmol O2. The amount of CO2 is shown on the x-axis.

This observation as well as the fact that 1%Au/TiO2(coll) and 1%Au/Fe2O3(coll) show the same activity in the absence of CO2, but a different one in its presence, indicate that carbon dioxide may affect the properties of the catalytic liquid/solid interface, either by enhanced molecular mass transfer or chemical interaction with the support material. Depending on the hydrophilicity/hydrophobicity of the support this interaction will be weaker or stronger causing differences in sorption behavior and basicity [102-103]. Fl-XANES spectra showed that gold was in the reduced (metallic) state before as well as after the reaction Figure 2.5. The increase of activity with reaction temperature (see Table 2.3) has been expected. But, a simultaneous increase of the selectivity to benzaldehyde was observed. An explanation could be the higher diffusivity of the reactant in the catalyst particle in the supercritical system at higher temperatures, which would allow faster diffusion of the aldehyde into the bulk phase and thus a lower probability for further reaction [103]. The decrease of reaction rate, indicated by reactions of different durations in the autoclave and the experiment in the highpressure view cell, which was monitored by IR-spectroscopy, is not yet

52

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

completely understood. The effect cannot be solely explained by deactivation of the catalyst, since the recycled catalyst still exhibited a certain conversion. Also no indication for changes of the gold particle size or leaching of gold was found. For example, HAADF-STEM pictures of 1%Au/TiO2(coll) before and after reaction at 100 °C and 150 bar for 3 h did not show any significant difference and by atomic absorption spectroscopy no gold could be detected in the filtered solution after reaction. A possible explanation is the influence of water, which, although in certain cases beneficial for CO oxidations [131-132], is detrimental for alcohol oxidations because of its low solubility in the non-polar solvent CO2 [121, 133]. Experiments showed that after the addition of small amounts of water (0.20.8 mol%) the conversion decreased by up to 60% (Table 2.4). This observation could be traced back to different reasons such as blocking of active sites or changes of surface properties. The influence of the added amount of oxygen presents itself in two ways: A lower amount than 0.9% leads to lower conversion, possibly because its concentration at the solidliquid interface is too low. A higher amount of oxygen also leads to lower conversions, which may be explained by the formation of a two phase system, which reduces mass transfer across the phase boundary [104]. Finally, although the best results were achieved for the model compound benzyl alcohol, extension of the combined application of gold-based catalysts and supercritical CO2 seems also interesting for other alcohols, as shown by preliminary experiments with 1-octanol and geraniol. Conversions of 4.1% and 20.3%, respectively, and selectivities of 90.7% and 30.6% can probably be improved by further optimization.

2.4 Influence of gold particle size 2.4.1 Results As described in paragraph 2.1.2 the conclusions of structure-sensitivity studies have to be examined very carefully, since in most cases particles of different sizes are prepared with significant differences in the preparation route. As a result the reliability of these data is questionable. Therefore, we tried to keep everything possible constant during the preparation procedure.

2.4 Influence of gold particle size

53

Gold particles of different size were prepared by the colloidal route described in paragraph 2.2.1; only the molar ratio of base to reducing agent and gold precursor was changed. Figure 2.8 shows the near edge regions (Au L3) of titania supported Au catalysts in the lower part. The lowest, black curve is the spectrum of a gold foil shown as reference. From bottom to top (red to blue) the particle size increases while the loading is kept constant; medians, estimated by electron microscopy, are 1.3, 2.3 and around 12 nm, respectively. They are – as expected – completely reduced. In the upper part of Figure 2.8 the spectrum of the catalyst with the smallest Au particles (1.3 nm, red curve) is plotted with that of the original colloid (black) and the spectrum of the colloid corresponding to a particle diameter of 2.3 nm (green).

Figure 2.8: Lower part (from bottom to top): near edge spectra of gold foil (black) and Au/TiO2 with different sizes of gold particles (1.3 nm - red, 2.1 nm - green, 12 nm blue); upper part: Spectra of colloidal solutions (Au particle size: 1.3 nm - black,2.1 nm - red) and Au/TiO2 (1.3 nm).

54

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

While the not yet deposited small colloid is still partly oxidized, the medium one is already completely reduced. The R-space spectra that were obtained for the Au/TiO2 catalysts by Fourier transformation of the k3weighted EXAFS spectra can be seen in Figure 2.9. The number of Au-Au bonds has its maximum in gold foil (blue). A decrease and shift to shorter bond lengths with decreasing particle size can be observed (green to black). The same results were obtained for gold on ceria.

Figure 2.9: R-space spectra of gold foil (blue) and Au/TiO2 with different sizes of gold particles (11.3 nm - green, 6.9 nm – orange, 2.1 nm - red, 1.3 nm - black).

The supported gold colloids were tested for the oxidation of benzyl alcohol to benzaldehyde under single-phase and biphasic conditions (the same as in paragraph 2.3). The catalyst with a gold particle size of 2.1 nm corresponds to 1%Au/TiO2(coll). The results are shown in Figure 2.10 and Figure 2.11. A plot of conversion over the mean diameter shows a maximum of catalytic activity for the catalyst with gold particles of 6.9 nm size. The activity drops vs. smaller and larger particles. For all tested catalysts the twophase system yielded higher conversions than the single phase.

2.4 Influence of gold particle size

55

Conversion /%

25 20

6.9 nm

2.1 nm

15 10 5 11.3 nm

1.3 nm

0 0

2

4

6

8

Mean diameter /nm

10

12

Figure 2.10: Conversions of 1%Au/TiO2 catalysts. Light gray: single phase 70 g CO2), dark grey: biphasic conditions (30 g CO2). Used amounts were 7 mmol benzyl alcohol, 14 mmol O2 and 1591 mmol (70 g) or 682 mmol (30 g) CO2.

For the case that the conversion is not only normalized to the real gold loadings, but also to the particle size (i.e. number of surface atoms) the increase in activity with increasing particle size is even more pronounced (Figure 2.11). The same results were obtained with the ceria support and in conventional solvents. A closer analysis of the colloids and the sizedependence at ambient pressure can be found in [134]. 1400 1200 6.9 nm

TOF /h-1

1000 800 600

2.1 nm

400 200

11.3 nm

1.3 nm

0 0

2

4

6

8

Mean diameter /nm

10

12

Figure 2.11: Turnover frequencies (conversion normalized for particle surface) of 1%Au/TiO2 catalysts during oxidation of benzyl alcohol. Light gray: single phase 70 g CO2), dark grey: biphasic conditions (30 g CO2). Used amounts were 7 mmol benzyl alcohol, 14 mmol O2 and 1591 mmol (70 g) or 682 mmol (30 g) CO2.

56

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

2.4.2 Discussion By keeping constant as many parameters as possible during the synthesis of the oxide supported Au catalysts with different gold particle sizes, it can be assumed that the only property changed and therefore the reason for different activities in the selective oxidation of alcohols are size and morphology of the adsorbed gold particles. The EXAFS spectra indicate a pronounced structure-performance relation and show that the adsorbed Au particles in as-prepared state are in reduced state as previously shown by XPS [36], even if the original colloid was still partly oxidized. Catalytic experiments showed that – in contrast to the well-established relationship between particle-size and catalytic performance for CO oxidation – larger gold particles are in fact more active for the oxidation of benzyl alcohol. The volcano plots in Figure 2.10 and Figure 2.11 show a range for an optimal particle size around 7 nm.

2.5 Summary The environmentally friendly selective oxidation of benzyl alcohol over supported gold catalysts in supercritical CO2 both under single-phase and two-phase conditions yielded higher conversions than the same reaction without any solvent. Especially the selectivity, with values very close to 100%, underscored the advantageous properties of both gold as catalytic material and dense CO2 as a reaction medium. Furthermore, this reaction was proven to take place with only molecular oxygen as oxidant without requiring the presence of an auxiliary base. The catalytic activity depends on various parameters such as type of support, preparation method, and size and morphology of the gold particles. By preparing catalysts with different gold particle-sizes a strong structure-performance relationship could be shown with the highest catalytic activity around 7 nm. Gold catalysis combined with the application of dense CO2 offers interesting opportunities for aerobic oxidation of a variety of substrates, including several other alcohols, as shown here for 1-octanol and geraniol. However, further investigations are needed to gain deeper insight into the functioning of these

2.5 Summary

57

catalytic systems using in situ spectroscopic methods and structurally tailored catalysts.

58

2 Gold-catalyzed alcohol oxidation in supercritical carbon dioxide

2.5 Summary

3

59

Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

X-Ray absorption spectroscopy using a microreactor array in combination with a two-dimensional X-ray sensor was applied for parallel structural screening of differently prepared supported Pd and Cu particles in the asprepared state and after heat pre-treatments in different gas atmospheres. A second, advanced cell was developed as a parallel reactor for catalytic in situ (XAS) measurements. The cell facilitates the simultaneous catalytic and structural investigation of six catalysts under different feed gas conditions. The same X-ray sensor was used for spectra collection. Gas compositions were measured by on-line mass spectrometry. The potential and limitations of the parallel XAS cell are discussed. The heterogeneously catalyzed partial oxidation (CPO) of methane was chosen as test reaction. Alumina-supported Rh and Au/Rh catalysts with different metal loadings (0.5-2.5 wt%) were applied and prepared via different preparation routes using flame spray pyrolysis (fsp) and colloid adsorption (coll). For comparison the catalysts were investigated in a fixed-bed capillary reactor (also for XAS measurements) and in an eight-fold parallel gas phase reactor using similar reaction conditions (6% CH4-3% O2-He, 250-500 °C). Similar catalytic results were obtained in all three reactor types, confirming the suitability of the parallel XAS cell for catalytic measurements. Analysis of the catalytic data, STEM images and the in situ XANES experiments indicated the following characteristics for the CPO reaction: sufficient heat production by combustion of methane, total conversion of oxygen, and reduction of a certain fraction of the catalyst. The overall catalytic behavior was in line with a two-zone model of the catalyst bed, where catalytic combustion dominates in the front zone and reforming reactions become favored in the second part of the catalyst bed.

60

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

3.1 Introduction Cells for parallel operando analysis of catalytic systems have only been realized for a limited number of techniques, e.g. XRD [135-136] or IRthermography [137-139], the driving force being time, cost efficiency, and the possibility of instant comparison of different materials. Which minimum number of analyses per time qualifies as high throughput method is a subject of discussion and marketing. For X-ray absorption spectroscopy (XAS) the development of parallel spectroscopic cells is beneficial – besides the general attractiveness of high throughput techniques [140] – since synchrotron time is limited and costly. Although serial setups were already designed [141], a cell for the parallel operando screening of heterogeneous catalysts has not yet been presented. In a first step, before looking at catalytic reactions, we developed a cell for optimization and parallelization of the structural characterization of solid materials, particularly under controlled atmospheres, which still remains a challenging task [135-136]. We developed a specially designed microreactor array, in which up to ten different solid materials can be placed and treated under a controlled atmosphere in combination with an X-ray sensitive area detector, previously reported for spatially resolved and energy-dispersive XAS studies [142-143]. To demonstrate the feasibility of this approach, two series of solid materials were investigated. (A) Differently prepared and supported palladium particles (Table 3.1) used as catalysts in the fields of selective alcohol oxidation [144] and total combustion as well as partial oxidation of methane [145-146]. The materials were characterized after preparation and certain pretreatment steps. In addition, their behavior in hydrogen and air was investigated. (B) Cu-based materials (Table 3.2) that are typically applied in methanol synthesis and steam reforming [147]. Since it was previously found that their redox behavior is strongly related to their catalytic activity [148], their reduction in hydrogen was analyzed. Then, an advanced version of this cell was developed allowing specific gas mixtures in every compartment and the monitoring of the catalytic behavior

3.1 Introduction

61

and the oxidation state of six different catalysts during reaction. The partial oxidation of methane (CPO) has been chosen as a test reaction. This reaction, i.e. the formation of synthesis gas from methane has been investigated for decades and can be run more energy efficiently on smaller scales than steam or autothermal reforming [149-151]. This would, as indicated in paragraph 1.3.3, make it a suitable option, e.g. for oil production sites, where the by-product natural gas is still just flared at many locations. Usual catalysts are supported Co, Ni and noble metals like Pt, Rh or Ru [152]. Three different reaction models for the CPO are discussed in the literature, one being the direct partial oxidation (DPO) of methane to synthesis gas (Eq. (1)) [153-155]. The other two are two-zone models with total oxidation (Eq. (2)) [156] or mainly direct partial oxidation [150, 157-158] in the first region of the catalyst bed followed by endothermic steam (Eq. (3)) and/or CO2 reforming (Eq. (4)) in the second region. Other groups propose that both direct partial oxidation and combustion-reforming can occur [151], depending on parameters like oxidation state of the catalyst [159] or the type of support for a certain metal, e.g. rhodium [160]. Successful model calculations were performed for both pathways, DPO [161] and CCR (catalytic combustion-reforming) [162-163]. CH4 + 0.5O2

→

CO + 2H2

ΔrH673K = -25.6 kJ/mol

CH4 + 2O2

→

CO2 + 2H2O

ΔrH673K = -800.2 kJ/mol (2)

CH4 + H2O

→

CO + 3H2

ΔrH673K = +219.2 kJ/mol (3)

CH4 + CO2

→

2CO + 2H2

ΔrH673K = +259.3 kJ/mol (4)

(1)

In a former study [142], where a non-adiabatic capillary reactor [164] containing a fixed-bed of alumina-supported Rh was used, the following behavior was observed during heating: up to a certain ‘‘ignition’’ temperature, which is defined here as the temperature at which CO and/or H2 can be detected in the reactor outlet stream, only carbon dioxide and water were observed as oxidation products and the catalyst was completely oxidized. After ignition carbon monoxide and hydrogen were detected and oxygen was completely consumed. The catalyst was still oxidized in the

62

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

entrance zone of the catalyst bed, whereas towards the exit it was reduced. Thermographic imaging showed a distinct temperature maximum, higher than the oven temperature, at the transition between oxidized and reduced zone of the catalyst bed [165]. The development of a hot spot moving towards the inlet of the reactor during ignition of CPO was already shown theoretically in [162]. During cool-down, i.e. extinction of the CPO reaction, the temperature maximum decreased and moved towards the exit of the catalyst bed, not all oxygen was converted anymore and the whole catalyst bed was in an oxidized state again. The new parallel in situ XAS cell was applied for testing the influence of catalyst preparation method (colloidal route and flame spray pyrolysis), Rh loading and of Au addition on the catalytic performance of aluminasupported Rh in the CPO of methane. For the assessment of the potential and limitations of the cell the results are compared to those collected under similar conditions employing a capillary reactor cell and an eight-fold parallel gas phase reactor system.

3.2 Experimental 3.2.1 Catalysts 3.2.1.1 Pd and Cu catalysts for the ten-fold cell The preparation of the palladium catalysts is given in Table 3.1. Different supported Pd samples were used: flame-made materials, several commercial catalysts for comparison, impregnated Pd/Al2O3 catalysts, oxidized glassy PdZr alloys and specially treated samples. Flame-made catalysts were prepared by flame spray pyrolysis [166] using two different solvents (route 1 and 2). In both cases aluminum sec-butoxide (75% in butanol, Acros) as precursor for the support and Pd acetylacetonate (ABCR) for the noble metal particles were used, once dissolved in xylene (route 1) and once in methanol/acetic acid (route 2). The liquid mixtures were sprayed into a methane/oxygen flame with an oxygen stream. The as-formed particles were collected on a glass-fiber filter placed on a cylinder, which was mounted above the flame by the aid of a vacuum pump (Vacuubrand). For the

3.2 Experimental

63

preparation of the 33%PdO/ZrO2 sample from amorphous alloys, a glassy Pd33Zr67 alloy [167] was exposed to air at 400 °C until full oxidation was accomplished, in the same way as reported in [146]. Pre-reduction of some of the Pd/Al2O3 catalysts was performed in 5% H2-He at 100 °C for 1 h. Note that the catalysts were removed from the reactor after the reduction and stored under air for about one week. Table 3.1: Palladium catalysts studied in the ten-fold parallel cell - Series (A). No. corresponds to the compartment number in the spectroscopic cell. 2, 8 and 9 were diluted with Al2O3. Cat. No. 1 2 3 4 5

6 7 8 9 10

Composition

Preparation Method

More detailed description

5 wt% Pd/Al2O3, Engelhard 40692 (pre-reduction in commercial 5%Pd/Al2O3 Impregnated process, stored in air) Pure palladium, partly oxidized, asPd black Commercial received 5 wt% Pd/Al2O3, Engelhard 40692, Impregnation, 5%Pd/Al2O3 freshly pre-reduced in hydrogen at pre-reduced 100 °C Flame made Flame spray pyrolysis, using xylene 5%Pd/Al2O3 (Route 1) mixtures [110, 166] 5 wt% Pd/Al2O3, Engelhard 40692 (pre-reduction in commercial 5%Pd/Al2O3 Impregnation process, stored in air) Shell-impregnated catalyst, 0.5%Pd/Al2O3 Shell-impregnation Engelhard 4586 (pre-reduced in commercial process, stored in air) 5%Pd/Al2O3 33%Pd/ZrO2

Flame made (Route 2) Oxidation of glassy alloy, afterwards reduced Oxidation of glassy alloy

undiluted 1: 10 undiluted undiluted undiluted

undiluted

Flame spray pyrolysis, using methanol/acetic acid mixtures

undiluted

Pd33Zr67 alloy oxidized in air and reduced in 5%H2/He [146, 167]

1:5

Pd33Zr67 alloy oxidized in air [146, 167] Shell-impregnated catalyst, 0.5%Pd/Al2O3 Shell-impregnation Engelhard 4586 33%Pd/ZrO2

Dilution

1:5 undiluted

The Cu catalysts were prepared by co-precipitation (CP), chemisorption– hydrolysis (CH), deposition–precipitation (DP), flame synthesis (FM), and

64

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

impregnation (I) as shown in Table 3.2. The 33%CuO/ZnO catalyst was prepared by controlled co-precipitation of Cu(NO3)2 and Zn(NO3)2 (precipitation pH = 9.2) followed by filtration through a 0.45 μm cellulose acetate membrane filter [168]. After drying at 90 °C and grinding, 1.0 g of the precursor powder was calcined in an oven with a temperature ramp of 3 °C/min and a final temperature of 250 °C for 10 hours. Chemisorptionhydrolysis was applied to catalysts 2, 4 and 9 (the latter being the same as sample 2) using an aqueous solution of [Cu(NH3)4]2+ and as reported in detail in [169]. The support material of the Au/Cu5Mg3Al2O11 catalyst was prepared by means of flame spray pyrolysis (FSP), gold was deposited by utilizing the deposition-precipitation method; details are described in [170]. The impregnation of 6%Cu/Al2O3 catalyst (sample 8) was achieved by impregnation with a CuCl2-solution in the same way as described in [171]. Table 3.2: Copper catalysts studied in the ten-fold parallel cell - Series (AB). No. corresponds to the compartment number in the spectroscopic cell. 1, 5, 6 and 7 were diluted with Al2O3. Cat. No.

Composition

Preparation Method Co-precipitation (CP)

1

33%Cu/ZnO

2

6%Cu/SiO2

3

1%Au/ Cu3Mg3Al2O11

4

5%Cu/Al2O3

5

Cu(OH)2

6

33%Cu/ZnO

7

5%Cu/ZrO2

Flame made

8

6%Cu/Al2O3

Impregnation

9

6%Cu/SiO2

Chemisorptionhydrolysis (CH)

Chemisorptionhydrolysis (CH) Depositionprecipitation (DP) Chemisorptionhydrolysis (CH) Commercial Co-precipitation (CP)

More detailed description Co-precipitation using Cu(NO3)2 and Zn(NO3)2 according to ref. [5] Chemisorption-hydrolysis according to ref. [6] Deposition-precipitation according to ref. [7] Chemisorption-hydrolysis according to ref. [6] As-received (Fluka) Co-precipitation (same as catalyst 1) Flame spray pyrolysis according to ref. [2] Impregnation of alumina pellets according to ref. [8] Chemisorption-hydrolysis according to ref. [6]

Dilution 1:4 undiluted undiluted undiluted 1 : 10 1:4 1: 3 undiluted undiluted

3.2 Experimental

65

3.2.1.2 Rh and Au/Rh catalysts for CPO of methane The catalysts for the catalytic experiments were prepared either via a colloidal route (coll) using NaBH4 as reductant or via flame spray pyrolysis (fsp). In the latter case, the salts of the desired components, aluminum acetylacetonate (Al(C5H7O2)3) and rhodium acetylacetonate (Rh(C5H7O2)3), were dissolved in a mixture of methanol/acetic acid 1:1 (v/v) and the desired amount of Au was added from an aqueous solution of HAuCl4•3H2O (1 g in 100 ml). The total concentration of metal ions in the solution was 0.40 mol/l. The solution was fed with a syringe pump at 6 ml/min through a nozzle and dispersed with oxygen and ignited by a small annular flame. The as-prepared catalyst was collected from a glass fiber filter placed on top of the apparatus. Further details of the experimental setup are given in [172-173]. Au, Rh or AuRh colloids have been prepared from HAuCl4•3H2O and RhCl3•H2O in aqueous solution mixed in the desired ratio using NaBH4 as reductant. The concentration of the colloids was 25 mg per 75 ml solution. NaBH4 was added in five-fold molar excess with respect to Rh and Au from a freshly prepared solution (0.05 mol/l). The Al2O3-support was suspended in 75 ml H2O and acidified with 3-4 ml of a H2SO4 solution (3 ml concentrated H2SO4 (98%) diluted to 100 ml) and the colloids were subsequently added. The catalyst was filtered off and washed exhaustively with deionized water until no traces of Cl- ions were detected (AgNO3-method). Investigated catalysts were - 2.5 wt% Rh (2.5Rh/Al2O3), - 0.5 wt% Au/2 wt% Rh (0.5Au/2Rh/Al2O3) and - 2 wt% Au/0.5 wt% Rh (2Au/ 0.5Rh/Al2O3) on Al2O3, each prepared as both fsp and coll. All Au/Rh compositions were also prepared as mechanical mixture of the respective Au/Al2O3 and Rh/Al2O3 materials and tested in the eight-fold parallel reactor. 0.5 wt% Rh/Al2O3(fsp) (0.5Rh/Al2O3) was used for comparison with 2Au/0.5Rh/Al2O3(fsp). The composition of the Au and Rh containing catalysts (0.5Au/2Rh/Al2O3 and 2Au/0.5Rh/Al2O3, each prepared by fsp and the colloidal route) was analyzed by inductively coupled plasma optical emission spectroscopy (ICP-OES). The analyses revealed that the samples prepared by flame spray pyrolysis as well as those prepared by the colloidal route showed

66

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

some small but significant loss of the metal contents, resulting in 1.7–1.9% instead of the nominal 2% and 0.4–0.5% instead of the nominal 0.5%. In the following the catalysts will be referred to using the nominal contents. Scanning transmission electron microscopy (STEM) investigations were performed using a Tecnai F30 microscope (FEI, Eindhoven). To improve the zcontrast, high-angle annular dark-field detection (HAADF-STEM) was employed. The particles were qualitatively analyzed by energy dispersive Xray spectroscopy (EDXS), for which the detector was attached to the Tecnai F30 microscope. 3.2.2 Parallel cell setup and experimental procedure 3.2.2.1 General setup The setup consisted essentially of the respective spectroscopic microreactor cell including heater and an area detector with charged couple device (CCD) sensor (Figure 3.1). In the X-ray camera X-ray photons are converted into visible light in a thin single crystal scintillator (thickness of the sensitive layer ca. 10 μm) that is imaged onto a CCD sensor by a microscopic optic [174]. The spatial resolution of the detector is approximately 10 μm [142], but the required resolution was on the 100 μm scale. The incoming intensity I0(E,x,y) was measured by taking an X-ray image without the cell and the transmitted intensity I(E, x, y) by imaging the cell in the beam. The exposure time for each image was 10 s. A typical picture obtained with the 10-fold cell is shown in Figure 3.3. From the recorded data, the absorption can be obtained for each pixel and – by integration over the corresponding compartment area – for each sample individually. For this purpose, the data were first dark-field corrected, i.e. the influence of the CCD dark-current and read-out noise was removed by subtracting an averaged dark image (without X-rays) from each of the images. After this correction, each pixel value (or that of the compartment) is proportional to the intensity. The values for I(E, compartment i) and I0(E, compartment i) for images taken with and without sample then allow calculating the integral absorption for each compartment i and energy by

3.2 Experimental

67

𝜇(𝐸, 𝑖, 𝑧)𝑑𝑧 = − 𝑙𝑛

𝐼(𝐸, 𝑖) . 𝐼 (𝐸, 𝑖)

3.1

Thus, the absorption spectrum for each compartment i can be determined in parallel using this technique. The raw data were calibrated for energy, background-corrected, and normalized using the WINXAS 3.1 software [111].

Figure 3.1: Schematic representation of the experimental arrangement. The parallel XAS cell is positioned between the monochromatic X-ray beam and the X-ray camera, in which a scintillator transforms incoming X-ray photons into visible light, which is led via microscopic optics to a CCD chip.

3.2.2.2 Ten-fold cell The characterization of the materials was performed at a standard EXAFS beamline at the Hamburger Synchrotronstrahlungslabor (HASYLAB at DESY, beamline X1) using an 8 mm x 1 mm large X-ray beam in combination with the X-ray sensitive camera (HASYLAB, 3 mm x 3 mm view with 10 μm resolution) described above. The spectroscopic cell, which is depicted in Figure 3.2 (the whole cell body is shown in Figure 3.4), contained ten sample compartments, the thickness of which was 2 mm for one version of the cell and 5 mm thickness for the other. The gas supply system provided the premixed 5% H2-He and 21% O2-He reaction mixtures. For in situ studies a flow of 20 ml/min was fed over the catalysts using mass flow controllers (Brooks). The catalysts were heated to the respective reaction temperatures by means

68

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

of an oven. The temperature was measured by a thermocouple directly in the stainless steel body. The whole microreactor array was mounted on an x,z,θ-table, which allowed aligning the cell both in horizontal and vertical translation using step motors. Each sample compartment was filled with the corresponding material (cf. Table 3.1 and Table 3.2, sieve fraction 100200 μm).

Figure 3.2:

Picture of the inlet for the ten-fold parallel XAS cell.

A typical transmission image is depicted in Figure 3.3, reflecting the different X-ray transmission intensities of the various materials. Note that 0.5% Pd/Al2O3 in compartments 6 and 10 is much brighter due to lower absorption than 5% Pd/Al2O3 or 5% Pd/ZrO2 in compartments 3 and 8. Materials with a strongly X-ray absorbing matrix were diluted with alumina (γ-Al2O3, Engelhard) to achieve similar X-ray transmissions and thus reasonable XAS spectra as indicated in the last columns of Table 3.1 and Table 3.2.

Figure 3.3: Microreactor array (X-ray transmission image) loaded from left to right with catalysts 1 to 10 as listed in Table 3.1 (taken above the Pd K-edge at 23.350 keV).

From images like those given in Figure 3.3 the X-ray absorption was determined for each pixel of the detector. By using the methodology described above the X-ray absorption spectra could be extracted for all ten

3.2 Experimental

69

compartments. Note that in the present case an X-ray camera with 3 mm x 3 mm image size was used requiring two snapshots for the entire cell. For comparison spectra of the compartments were recorded sequentially, in a conventional manner using a 0.5 mm wide x 1 mm high beam and ionization chambers (length of 10 cm). The in situ cell was placed between the first and second chamber, a reference foil between second and third. Gas composition and pressure were adjusted so that 15% of the X-ray intensity was absorbed in the first and 40% in the second and third ionization chamber. The setup was reported for instance in [175]. The typical beam current of the DORIS III storage ring at HASYLAB was 80 120 mA (operating positron energy at 4.5 GeV). In both cases, a Si(311) and Si(111) double crystal monochromator for step-by-step scanning of the energy around the Pd K-edge and Cu K-edge was used. Higher harmonics were effectively removed by detuning of the crystals to 70% of the maximum intensity. The X-ray energy was scanned in the XANES region around the Pd K- (E = 24320 - 24450 eV) and Cu K-edge (E = 8960 -9040 eV) in steps of 1 eV. 3.2.2.3 Six-fold cell As, contrary to the ten-fold cell, each microreactor compartment of the advanced cell is equipped with its own gas supply and outlet, the dimensions had to be changed. The central part of the parallel cell is the cell body with now only six compartments, the dimensions of each being 0.8 mm x 8mm x 5 mm (width x height x depth); the distance between two compartments is 0.65 mm (Figure 3.4). For experiments in the parallel XAS cell, the six adjacent compartments were filled with about 4 mg of the respective catalysts (again a sieved fraction of 100-200 mm). Quartz wool was used at entrance and exit to fix the powder in the reactor during operation. Each of the compartments had separate lines for gas inlet and outlet, the inlet line being connected to a mass flow controller (Brooks), the outlet via a valve system either to a mass spectrometer (Pfeiffer, OmniStar) or the exhaust gas line (Figure 3.1).

70

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Figure 3.4: View of the parallel XAS cell with cell body. Graphite seals are pressed on the compartments with covers that leave a window of 2 mm x 8 mm for the X-ray beam. Capillaries for gas in- and outlet are connected to each reaction volume separately (outer diameter of capillaries: 1/16 in.). The lower picture shows a top-view of the cell body (4 cm in diameter) containing six reaction compartments.

Standard flow rates were 10 ml/min. The compartments were sealed using a graphite disk and a piston that left a window of 2 mm x 10 mm. The volumetric flows through the parallel XAS cell were checked by a flow meter prior to each experiment. Experiments were performed at the Hamburger Synchrotronstrahlungs-Labor (HASYLAB) at the Deutsches ElektronenSynchrotron (DESY) at beamline C using a Si(111) double-crystal monochromator. For spectra acquisition an 8 mm x 1 mm (width x height) large beam was used. A typical image recorded of this cell by CCD-area

3.2 Experimental

71

detector is shown in Figure 3.5. Data processing was performed as described in paragraph 3.2.2.1

Figure 3.5: X-ray transmission image as obtained by the X-ray camera at 23.22 keV. The following catalysts were loaded: (1) 2Au/0.5Rh/Al2O3 (coll), (2) 0.5Au/2Rh/Al2O3 (coll), (3) 2.5Rh/Al2O3 (fsp), (4) 2Au/0.5Rh/Al2O3 (fsp), (5) 0.5Au/2Rh/Al2O3 (fsp). Note that only five of the six loaded catalysts are visible due to the microscope optics used for this experiment.

3.2.3 Capillary Setup For XAS measurements on the capillary reactor, the X-ray camera was replaced by a second ionization chamber. The quartz capillary (Markröhrchen, Hilgenberg GmbH, 1 mm diameter, wall thickness of 20 μm) with gas in- and outlet containing a packed bed of the respective catalyst (about 4 mg) was mounted on a gas blower; the setup has been described in more detail in [175] and the schematic view in Figure 3.6 is taken from therein. For integral measurements over the catalyst bed the beam size was 8 mm x 1 mm. It was reduced to 0.5 mm x 1mm to be able to scan along the catalyst bed. Spectra were taken at the Rh K-edge (23.190-23.320 eV). The acquisition time for one spectrum was 3 min.

72

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Figure 3.6: Capillary reactor setup consisting of the capillary over a gas blower. Thermocouples measure the temperatures of the heater and directly below the capillary reactor. Image taken from [175].

3.2.4 Parallel gas-phase reactor system (Celero) An eight-fold, software controlled reactor system (Celero, Altamira Instruments) was used for alternative catalytic testing (Figure 3.7). The tubular reactors (5 mm inner diameter, 12 cm length) were filled with 50 mg of the respective catalyst (sieved fraction of 0.1-0.2 mm, diluted with 325 mg Al2O3 of the same fraction). One reactor containing only Al2O3 was always used as a blank test. The gases were led through heated capillaries to ensure an equal space velocity in all reactors. For the CPO of methane a gas mixture (12% CH4, 6% O2, He) was diluted with He to yield an integral flow (over all eight reactors) of 400 ml/min. The gas composition at the outlet was analyzed for each reactor separately by a GC system (Agilent 5890N, 50 m GS-GASPRO column, 320 μm inner diameter) with mass spectrometric (MSD, Agilent 5973N) and thermal conductivity detection (TCD). The reactant mixture and a gas mixture of 1% CO, 1% CO2 and 1% H2 in He were used for the calibration. A detailed description of the setup can be found in [176]. The flow distribution of the reactors was determined by flowing He-diluted nitrogen through the system and analyzing the GC signals: a standard

3.3 Transformations in the ten-fold cell – proof of principle

73

deviation of 0.79% was found among the seven reactors. The reproducibility was determined by analyzing different catalysts multiple times and by comparing different batches; the standard deviations determined were below 2.00% for O2, 4.94% for CO, 1.02% for CH4 and 3.95% for CO2.

Figure 3.7: Eight-fold, software controlled reactor system (Celero, Altamira Instruments). Gases were fed to the reactors via heated capillaries to ensure an equal space velocity in all reactors.

3.3 Transformations in the ten-fold cell – proof of principle The resulting spectra of the palladium based materials are shown in Figure 3.8. The spectra are very similar using the two different approaches (left and right). The X-ray absorption near edge structure among the different catalysts is different. In some of them the Pd particles are completely oxidized (0.5% Pd/Al2O3 in compartments 6 and 10 and the flame made 5% Pd/Al2O3 in compartments 4 and 7), some of them are in a mixed oxidation state and some are quite reduced, due to their pre-treatment (ex situ) in hydrogen (2, 3 and 8). The similarity of the spectra taken from samples in compartments 1 and 5 demonstrates the reproducibility of the parallel screening method.

74

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Figure 3.8: X-ray absorption spectra of 10 Pd catalysts taken with the X-ray sensitive camera (left) and with a 0.5 mm x 0.5 mm large beam (right); catalysts 1 to 10 are from top to bottom as listed in Table 3.1.

Figure 3.9: Extent of reduction of the Pd catalysts (taken with the X-ray camera) after filling the reactor with differently pre-treated catalysts: as-prepared, after reduction in 5% H2-He at room temperature, and after re-oxidation in 21%O2-He at room temperature; numbering as given in Table 3.1.

3.3 Transformations in the ten-fold cell – proof of principle

75

In a next step, the samples were subjected to a 5% H2-He mixture at room temperature. Most of the noble metal particles were reduced, except the flame-made 5% Pd/Al2O3, which had been prepared by flame synthesis route 2. The results are represented in Figure 3.9, where the spectra of a fully oxidized and a fully reduced Pd sample (sample 9 as-prepared and sample 2 after reduction) were taken as basic components for linear combination and thus the determination of the oxidation state. Note that the flame-made materials were also not completely reduced in previous, conventional EXAFS experiments. Scanning of the samples with a 0.5 mm x 0.5 mm beam was performed in this case as well and gave similar results. Similar conclusions can be drawn directly from the XANES spectra in the complete data set as well as from Figure 3.10 and Figure 3.11. Finally, all materials were treated in 21% O2-He to study the re-oxidation behavior (Figure 3.9). Different behaviors can be distinguished in this case as well. The flame-made Pd catalyst (route 2) was re-oxidized, while 0.5% Pd/Al2O3 was only partly reoxidized, and other Pd catalysts even remained in the reduced state. This can be traced back to different Pd particle sizes: at room temperature only the surface layers are re-oxidized and therefore Pd catalysts with high loading and large Pd particle size remain in the reduced state. This phenomenon has been shown to have a strong influence on the catalytic properties [144-146].

Figure 3.10: X-ray absorption spectra of the Pd catalysts taken with the X-ray camera (left side) and with a 0.5 x 0.5 mm large beam (right) after reduction in 5%H2/He; catalyst 1 to 10 are from top to bottom as listed in Table 3.1.

76

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Figure 3.11: X-ray absorption spectra of the Pd catalysts taken with the X-ray camera (left side) and with a 0.5 x 0.5 mm large beam (right) after reduction in 5%H2/He and re-oxidation in 21%O2/He; catalysts 1 to 10 are from top to bottom listed as in Table 3.1.

While the experiments with Pd-based materials were performed at room temperature, catalysts or sensor materials are often pre-treated under different atmospheres at elevated temperatures. To demonstrate the suitability of the suggested strategy for this purpose, the reduction of supported copper catalysts was investigated. Figure 3.12, Figure 3.13 and Figure 3.14 compare the spectra of the copper samples at different temperatures. Comparing samples 1 and 2 Cu/ZnO was almost completely reduced at 175 °C, whereas Cu/SiO2 required higher temperatures to show the same degree of reduction. The reduced state of Cu is indicated by the strong pre-edge feature at 8.979 keV and the two features at 8.987 and 9.02 keV [148, 168].

3.3 Transformations in the ten-fold cell – proof of principle

77

Figure 3.12: XANES spectra of Cu samples 1 to 4 (cf. Table 3.2) taken with the X-ray camera at room temperature, 175 °C, 200 °C, and 220 °C (top to bottom).

Similar results with respect to the reduction behavior were obtained in conventional XAS experiments using a standard in situ cell. Data at all the temperatures were collected for all materials and the degree of reduction could be determined by linear combination analysis using a Cu(II) (asprepared material) and a Cu(0) reference (Cu foil). The results (Figure 3.15) show the different reduction behavior of copper in different metal oxide matrices. Note that compartments 1 to 5 were recorded first, followed by compartments 6 to 9; therefore the latter ones appear slightly more reduced. In a previous study, we recorded the different reduction behavior of various Cu catalysts in sequential in situ XAS experiments [148].

78

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Figure 3.13: XANES spectra of Cu samples 5 to 8 (cf. Table 3.2) taken with the X-ray camera at room temperature, 175 °C, 200 °C, and 220 °C (top to bottom).

Figure 3.14: XANES spectra of Cu sample 9 (cf. Table 3.2) taken with the X-ray camera at room temperature, 175 °C, 200 °C, and 220 °C (top to bottom).

3.4 CPO in six-fold cell – catalysis in a parallel cell

79

Figure 3.15: Extent of reduction of the Cu catalysts at room temperature, 175 °C, 200 °C, and 220 °C (from left to right). For numbering refer to Table 3.2.

3.4 CPO in six-fold cell – catalysis in a parallel cell In a first step the influence of the Rh loading and the preparation route on the catalytic behavior was investigated. Comparing the oxidation behavior of 2.5Rh/Al2O3 (fsp) and 0.5Rh/Al2O3 (fsp) on the basis of the spectra that were extracted from the images obtained with the parallel reactor cell, it emerges that lower metal loading required a higher temperature to reduce the Rh component in the catalyst (Figure 3.16a and d). The reduced state of rhodium can be identified by the two features of about the same absorption intensity at 23.23 keV and 23.25 keV, while oxidized species only show the whiteline feature at 23.23 keV. For 0.5Rh/Al2O3 (fsp) and 2.5Rh/Al2O3 (fsp) the MS signal showed only CO2, H2O, CH4 and O2 while XANES spectra still indicated oxidized Rh species. Reduced Rh species were detected, as soon as the formation of H2 was observed and the oxygen signal completely disappeared. When starting with reduced Rh as in the colloidal case (2.5Rh/Al2O3 (coll), Figure 3.16e), the catalyst showed a small tendency towards oxidation upon heating, indicated by a slight decrease of the second maximum compared to

80

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

the first one in the spectrum at 370 °C. The colloidal catalyst only showed low conversion of methane to CO2. The reliability of the parallel cell – in the sense of reproducibility – was confirmed by obtaining the same results for 2.5Rh/Al2O3 (fsp) in different compartments and for different runs. In order to study the effect of gold addition, different amounts of gold were added to the catalysts using both preparation routes, i.e. colloidal adsorption and flame spray pyrolysis. The absolute noble metal loading was kept constant at 2.5 wt% resulting in the following catalyst compositions: 2.5Rh/Al2O3, 0.5Au/2.0Rh/Al2O3 and 2Au/0.5Rh/Al2O3 (Figure 3.16). In the case of the flame spray materials oxidized Rh species could be observed, which can be reduced at a certain temperature. Reduction and ignition temperature (hydrogen detectable) coincided. This temperature increased with lower Rh content (Figure 3.16a-c); 2.5Rh/Al2O3 (fsp) was reduced between 333 °C and 353 °C (Figure 3.16a), 0.5Au/2Rh/Al2O3 (fsp) between 360 °C and 400 °C (Figure 3.16b), and 2Au/0.5Rh/Al2O3 (fsp) still yielded a spectrum indicative of oxidized Rh species at 400 °C (Figure 3.16c); approximate ranges of ignition temperatures of all investigated catalysts are given in Table 3.3.

3.4 CPO in six-fold cell – catalysis in a parallel cell

81

Figure 3.16: X-ray absorption spectra at the Rh K-edge of the catalysts studied with the parallel XAS cell. (a)-(d) are catalysts prepared by flame spray pyrolysis, (e)-(g) catalysts prepared by the colloidal route, and (h) are reference spectra of Rh0 and Rh3+. The room temperature spectra of the Au/Rh fsp catalysts are identical to the ones of the respective pure Rh catalyst. The thin dash-dotted vertical lines mark the two absorption maxima at 23.23 keV and 23.25 keV. The support (Al2O3) is left out in labeling.

82

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Table 3.3: Ignition temperature ranges of the used catalysts. Ranges were estimated by taking into account results from all three reactor systems (parallel XAS cell, capillary microreactor, and lab-scale eight-fold reactor system). When no temperature interval is given, an ignition of CPO of methane could not be observed in the investigated temperature range. The support was always Al2O3. Catalyst 2.5Rh (fsp) 0.5Rh (fsp) 0.5Au/2Rh (fsp) 2Au/0.5Rh (fsp) 0.5Au/2Rh (fsp), mech. mixture 2Au/0.5Rh (fsp), mech. mixture 2.5Rh (coll) 0.25Au/2.25Rh (coll) 0.5Au/2Rh (coll) 2Au/0.5Rh (coll) 0.5Au/2Rh (coll), mech. mixture 2Au/0.5Rh (coll), mech. mixture

Tentative ign. temp. [°C] 330 - 350 420 - 450 340 - 370 380 - 410 320 - 350 380 - 400 370 - 400 440 - 480 380 - 420 440 - 480

For the colloidal catalysts, however, where Rh was in reduced form at room temperature (Figure 3.16e-g), rhodium was oxidized during heat-up and reaction towards H2O and CO2 was detected. The extent of this oxidation was higher with lower Rh loadings: 2Au/0.5Rh/Al2O3 (coll) was oxidized to a greater extent than 0.5Au/2Rh/Al2O3 (coll) between 350 °C and 360 °C (Figure 3.16g and f). All three colloidal catalysts stayed partly oxidized on temperature increase (up to 400 °C), i.e. the integral spectra obtained with the parallel XAS cell showed no re-reduction, and hydrogen could not be detected by MS. The re-reduction would either be caused by decomposition into metal and oxygen at a sufficiently high overtemperature or by the formation of hydrogen during the ignition of the CPO of methane. STEM images of Rh/Al2O3 (coll) and AuRh/Al2O3 (coll) showed little difference in the appearance of the noble metal particles (Figure 3.17). Agglomerates in the size range between 20 nm and 100 nm were discernible that consisted of primary particles in the sub 10 nm range. EDXS measurements of the agglomerates reflected qualitatively the elemental composition, as determined by ICP-OES given in the experimental part (1.7-1.9 wt% and 0.40.5 wt% instead of the nominal 2 wt% and 0.5 wt%, respectively). STEM

3.4 CPO in six-fold cell – catalysis in a parallel cell

83

images of 2.5Rh/Al2O3 (fsp) and 0.5Au/2Rh/Al2O3 (fsp) are presented in Figure 3.18, the Rh particles exhibiting a size between 1 nm and 2 nm. The Au particles are larger, but smaller than 5 nm.

Figure 3.17: STEM images of Rh and Au/Rh containing catalysts prepared by the colloidal route. The rhodium content decreases from top to bottom. Results of typical EDX measurements of the respective noble metal aggregates are shown on the right side.

84

3 Development of an in situ parallel XAS cell and its application to steady-state CPO of methane

Figure 3.18: STEM images of 2.5Rh/Al2O3 (fsp) and 0.5Au/2.5Rh/Al2O3 (fsp). 1-2 nm spots on the primary support particles are Rh, the larger, brighter ones (