HYDROCARBON AND AMMONIA CHEMISTRY

Transcription

1 HYDROCARBON AND AMMONIA CHEMISTRY ON NOBLE METAL SURFACES CATALYTIC REACTIONS STUDIED ON A MOLECULAR SCALE PROEFSCHRIFT TER VERKRIJGING VAN DE GRAAD VAN DOCTOR AAN DE UNIVERSITEIT LEIDEN, OP GEZAG VAN DE RECTOR MAGNIFICUS DR. D.D. BREIMER, HOOGLERAAR IN DE FACULTEIT DER WISKUNDE EN NATUURWETENSCHAPPEN EN DIE DER GENEESKUNDE, VOLGENS BESLUIT VAN HET COLLEGE VOOR PROMOTIES TE VERDEDIGEN OP WOENSDAG 13 SEPTEMBER 2006 KLOKKE UUR DOOR CORNELIS JOHANNES WESTSTRATE GEBOREN TE NIEUWE-TONGE IN 1979

2 Promotiecommissie promotor: referent: Prof. Dr. B.E. Nieuwenhuys Prof. Dr. G.J. Kramer (TUE) overige leden: Dr. Alessandro Baraldi (Università di Trieste, Italië) Prof. Dr. J. Brouwer Prof. Dr. A.W. Kleyn Prof. Dr. M.T.M. Koper Prof. Dr. R.A. van Santen (TUE) Dr. E. Stobbe (ECN) The research described in this thesis was performed under auspices of NIOK, the Netherlands Institute of Catalysis, and financially supported by the Technology Foundation STW, applied science division of NWO and the technology programme of the Ministry of Economic Affairs, under project number UPC-5037.

3 A scientist in his laboratory is not only a technician: he is also a child placed before natural phenomena which impress him like a fairy tale. MARIE CURIE

4

5 Contents 1 Introduction Catalytic partial oxidation Ammonia oxidation The Surface Science approach Experimental methods and equipment Thermal desorption studies The vacuum system Low energy electron diffraction X-ray photoemission spectroscopy Scanning tunneling microscopy Molecular beam experiments Sample preparation Hydrocarbon oxidation on a stepped Pt surface Introduction n-butane Adsorption, desorption and decomposition The effect of pre-adsorbed oxygen Propene Adsorption, desorption and decomposition The effect of oxygen Steady state oxidation n-butane Propene Saturated vs. unsaturated hydrocarbon oxidation Summary and conclusions Benzene chemistry on Ir(111) Introduction Molecular adsorption Desorption and decomposition v

6 vi CONTENTS 4.4 The effect of pre-adsorbed oxygen Comparison to other surfaces Summary and conclusions Methane and oxygen on Rh(100) Introduction Oxygen adsorption Methane decomposition and oxidation CH x surface chemistry The effect of O ad on methane adsorption High temperature carbon surface chemistry Summary and conclusions Ammonia surface chemistry on Ir(110) Introduction Ammonia surface chemistry Molecular adsorption NH 3 dissociation Hydrogenation of N ad and NH ad An energy scheme Characterization of the oxygen covered surface The influence of O ad on the NH 3 surface chemistry The effect of O ad on N 2 desorption The effect of surface structure NO formation Conclusions Ammonia surface chemistry on Pt(410) Introduction NH 3 surface chemistry Characterization of the oxygen covered surface The effect of O ad on the NH 3 surface chemistry N 2 vs. NO formation Discussion Summary and conclusions Ammonia oxidation on Pt and Ir Introduction Steady state NH 3 decomposition on Ir(110) Steady state NH 3 oxidation on Ir(110) Steady state NH 3 oxidation on Pt(410) Discussion Reaction mechanisms Reactivity of other Pt surfaces

7 8.5.3 The selectivity of NH 3 oxidation: Ir vs. Pt Summary and conclusions General discussion From Surface Science to catalysis Hydrocarbon chemistry on Pt, Ir and Rh Ammonia oxidation A NH 3 decomposition: a kinetic model 141 References 143 Summary 155 Samenvatting 159 Nawoord 163 List of publications 165 Curriculum Vitae 167 vii

8

9 Chapter 1 Introduction Due to the growing world population, and the increasing per capita energy consumption, the demand for energy will continue to increase. Nowadays the largest part of the energy is produced using fossil fuels. Since the amount of available fossil fuels is limited there is an increasing need to find other ways to fulfill the energy needs. Mankind becomes increasingly aware of the fact that its activities influence the ecosphere. Combustion of fossil fuels results in the formation of huge amounts of H 2 O and CO 2, but other substances, for example NO x, SO x and unburned hydrocarbons, are formed as well. Several climate studies indicate [1] that the global temperature is rising. The CO 2 concentration in the atmosphere has increased in the past decades, and as CO 2 is one of the greenhouse gases, this is thought to be one of the causes of a global temperature rise. Although there is still a scientific debate about the exact cause of this temperature rise, several industrialized countries have signed the Kyoto protocol (1997), in which they agreed to reduce the emission of green house gases. This implies a reduction of the use of fossil fuels. Other, sustainable energy sources are therefore needed to provide enough energy. The sun produces more than enough energy to fulfill the energy needs of the world population. Harvesting of this energy (using solar sells, wind power, biomass) is, therefore, a virtually endless source of energy. The state-of-the-art technology to efficiently capture solar energy is not yet suitable for large scale application and lot of research is performed to improve these techniques. The use of renewable energy sources asks for a versatile energy carrier. Hydrogen is one of the promising candidates [2 4]. It can be easily produced from a variety of sources and its energy contents can be released at the point where it is needed. Low temperature H 2 oxidation (for example by using a fuel cell) yields only H 2 O, a harmless product. 1

10 2 CHAPTER 1. INTRODUCTION 1.1 Catalytic partial oxidation The long term goal is hydrogen production from sustainable energy sources, but in the transition period hydrogen production from fossil fuels will be the most important source of H 2. The most commonly applied method to produce H 2 from fossil fuels is by reaction of a carbonaceous starting material, for example coal or methane, with H 2 O (steam reforming). The overall reaction equation is shown in equation (1.1). C x H y (g, l, s) + 2x H 2 O (g) x CO 2 (g) + ( 1 / 2 y + 2x) H 2 (g) (1.1) The catalyst used in this process contains Ni as the active material. The reaction is endothermic, i.e. it requires input of energy. Efficient H 2 production is, therefore, limited to large scale hydrogen production facilities. Hydrogen formation on a small scale should be preferably exothermic, so that the reaction can be self-sustaining, without input of additional energy. Several years ago Hikman and Schmidt [5] reported that H 2 can be produced via catalytic partial oxidation (CPO) of methane (and also other hydrocarbons) using O 2. The overall reaction is shown in equation (1.2). C x H y (g) + x O 2 (g) x CO 2 (g) + 1 / 2 H 2 (g) (1.2) This process requires specific process conditions, to prevent the formation of the thermodynamically most favorable products, H 2 O and CO 2. The advantage of this process is that it is exothermic and that the reaction is self-sustaining after ignition. This makes it a good candidate for application on a small scale, for example in automotive or residential applications. The H 2 that is thus produced can be used to generate electricity using a fuel cell. Due to the specific process conditions (i.e. a high operating temperature) Ni based catalysts are not the best choice. Noble metal catalysts, like Pt, Rh and Ir have shown to be promising candidates as catalyst in this reaction. Next to application as a catalyst in the CPO reaction, noble metal based catalysts are used to catalyze a wide range of reactions, including those involving hydrocarbons. Detailed understanding of the adsorption, decomposition and oxidation of hydrocarbons on these noble metal surfaces is, therefore, useful for a wide range of catalytic systems. In the first part of this thesis several aspects of the catalytic processes taking place during Catalytic Partial Oxidation are discussed. In the Chapters 3 and 4 the adsorption, decomposition and oxidation of larger hydrocarbons (n-butane, propene, benzene) with a Pt and an Ir surface is described, whereas in Chapter 5 the adsorption and decomposition of both methane and oxygen on a Rh surface is described. 1.2 Ammonia oxidation Since the development of the nitrogen fixation process by Haber and Bosch, NH 3 has become one of the major products of the chemical industry. In this catalytic reaction

11 1.3. THE SURFACE SCIENCE APPROACH 3 atmospheric N 2 reacts with H 2 on Fe or Ru based catalysts, forming NH 3. Oxidation of ammonia is an industrially important reaction, which takes place on a large scale. Equation (1.3) shows the overall reaction, in which NH 3 is converted into NO. 4 NH 3 (g) + 5 O 2 (g) 4 NO (g) + 6 H 2 O (g) (1.3) The thus formed NO reacts with O 2 [Eq. (1.4)] and H 2 O [Eq. 1.5)] to form nitric acid, that is used for the production of, among others, fertilizers. 2 NO (g) + O 2 (g) 2 NO 2 (g) (1.4) 3 NO 2 (g) + H 2 O (l) 2 HNO 3 (aq) + NO (g) (1.5) The catalyst used in this so-called Ostwald process is a Pt/Rh gauze. Despite its relevance for the chemical industry several questions regarding the reaction mechanism are still not definitively answered. In the Chapters 7 and 8 results concerning ammonia oxidation on Pt surfaces are presented. The increasingly stringent demands for pollution control also include the reduction of NH 3 emissions. Ammonia that is released in the environment can damage sensitive ecosystems, for example by eutrophication. One of the possible ways to reduce NH 3 emissions is to catalytically convert the ammonia present in the flue gas into harmless substances, i.e. N 2 and H 2 O [see Eq. (1.6)]. 4 NH 3 (g) + 3 O 2 (g) 2 N 2 (g) + 6 H 2 O (g) (1.6) Pt based catalysts are very selective to NO formation, and that is why they are used in the Ostwald process. Removal of NH 3 from waste gas requires a different catalyst, that is selective towards N 2 rather than NO. Van den Broek et al. [6] found, that Ir based catalysts show a good selectivity in the oxidation of NH 3 to N 2 rather than NO. In the Chapters 6 and 8 our results concerning ammonia oxidation on an Ir surface are discussed. 1.3 The Surface Science approach The catalyst used for CPO consist of either a the pure metal gauze or of nanoparticles of the metal on an oxidic support. For ammonia oxidation using the Ostwald process the catalyst is used in the form of a metal gauze as well. In the Surface Science studies presented in this thesis the complexity of these real catalysts is reduced to well-defined, single crystal surfaces. This allows us to study, for example, the differences in reactivity of the different crystal facets, as well as the influence of defects, like steps and kinks, on the reactivity. This approach introduces the socalled material gap, i.e. the model system is a simplification of the real catalyst. Our studies were performed in ultra high vacuum (UHV, mbar), and the results presented here were obtained using low reactant pressures ( 10 7 mbar). Working in an UHV environment offers several significant advantages for the fundamental

12 4 CHAPTER 1. INTRODUCTION study of catalytic processes. Under these conditions the surface contaminants can be removed in situ and the surface remains clean for approx. 1 h afterwards. This allows one to work study clean metal surface, rather than studying a contaminated surface. This approach introduces the pressure gap, i.e. the reactant pressures used during our model experiments are several orders of magnitude lower than the actual conditions during catalysis. Due to these two simplifications extrapolation of the results, obtained using UHV and a single crystal surface, to real catalytic processes under industrial conditions should be looked upon with care. In each case one should consider the possible effects of the gaps on the catalytic process. In Chapter 9 we discuss the possible influence of both the pressure gap and the material gap, including the consequences for extrapolation of our results to more realistic catalytic systems.

13 Chapter 2 Experimental methods and equipment In this Chapter the different experimental techniques that were used to obtain the results reported in this thesis are discussed. The discussion focuses on those aspects relevant for the interpretation of the results and relevant literature references are provided for further reading. 2.1 Thermal desorption studies Irreversible adsorption of a gas on a solid surface occurs only in a limited temperature window. Carbon monoxide, for example, adsorbs at low temperatures on noble metal surfaces, but desorbs upon heating (in vacuum) above a threshold temperature ( 550 K). Other adsorbates like H 2, O 2, NO and NH 3 also adsorb at low temperature. The temperature at which desorption occurs depends on the strength of the adsorbate-surface interaction. In some cases dissociation occurs during adsorption (H 2 on most transition metal surfaces), but in other cases (partial) dissociation occurs during heating of an adsorbed layer (O 2, NO, NH 3 ). The products formed during decomposition eventually desorb as well, but usually at a different temperature than the original molecule. A detailed study of the desorbing molecules as a function of substrate temperature therefore yields information about the surface chemistry of adsorbates. A typical thermal desorption experiment is done in the following way: the surface is exposed to the adsorbate at low surface temperature and the surface is subsequently heated (in vacuum) using a constant heating rate. The molecules that appear in the gas phase during heating are monitored using a mass selective detector. 5

14 6 CHAPTER 2. EXPERIMENTAL METHODS AND EQUIPMENT A general formula to describe the desorption process is the Polanyi-Wigner equation [Eq. (2.1)], r(θ) = dθ dt = ν(θ)θ n e E d(θ)/rt (2.1) in which θ represents the surface coverage, E d represents the desorption barrier, R the gas constant, T the surface temperature, ν the pre-factor β the heating rate and n the order of the coverage dependence. A frequently used approximation, which is only valid for first order desorption processes (n = 1), is the so-called Redhead [7] equation [Eq. (2.2)]. ( ) ] νd T m E d = RT m [ln 3.46 (2.2) β This equation can be used to estimate the desorption barrier (E d, kj mol 1 ) using the temperature at which the desorption maximum is observed (T m, K) and the heating rate (β, K s 1 ) that has been applied. Accurate values of E d for more complicated desorption processes can only be obtained when both the initial surface coverage (θ 0 ) and the heating rate are systematically varied. A detailed discussion about different methods to evaluate thermal desorption data can be found elsewhere [8,9]. A slightly different application of a temperature programmed experiment can be found in temperature programmed oxidation/reduction, in which the adsorbatecovered surface is heated in the presence of an oxidizing agent in the gas phase, such as O 2 or NO. Alternatively a reducing agent, such as CO or H 2, can be used. The desorbing species that are observed provide information about the oxidation or reduction of the adsorbed molecules. An example of this is shown in Chapter 5, in which the reactivity of different oxygen-covered surfaces is compared by heating these surfaces in the presence of CO/H 2. Information about the reaction products during a steady state reaction can be obtained in a similar fashion. In that case the adsorbate-covered surface is heated in the presence of a reactant mixture. Information about the steady state reaction can only be obtained when the heating rate is sufficiently slow, so that transient effects are only marginal. By varying the heating rate one can distinguish between those two processes. Temperature programmed reaction spectroscopy was used to obtain the results presented in Chapters 3, 4 and The vacuum system The UHV system used in this study is equipped with a sputter gun for sample cleaning, VG Microtech Rear View LEED 900 optics, and with X-ray source, an electron gun and an electron energy analyzer (XPS/AES). The quadrupole mass spectrometer (UTI 100C) is located in a separate compartment, which is pumped by a separate pumping system (ion pump and turbo-molecular pump). The QMS compartment is

15 2.2. LOW ENERGY ELECTRON DIFFRACTION 7 connected with the main vacuum chamber by a 2 mm wide (circular) hole, and the sample surface can be placed in close proximity ( 2 mm) of this hole. This experimental layout enhances the sensitivity to desorption from the sample surface with respect to desorption from the heating wires or the edges of the sample. The experimental design also allows the use of relatively high reactant pressures during steady state reaction, while the pressure in the QMS compartment remains below the maximum QMS operating pressure. The mass spectrometer is computer controlled and both sample temperature and up to ten masses could be measured simultaneously. For all the experiments presented in this thesis an ionizer voltage of 70 V was used. The sample holder can be cooled using liquid nitrogen. The sample is spotwelded to Ta wires and is electrically isolated but thermally connected to the liquid N 2 reservoir. The lowest temperature that can be reached in this setup is 100 K. The sample is resistively heated, and the heating rate can be controlled using a Eurotherm (model 2416) temperature controller. Typical heating rates that were used range from 0.5 to 5 K s 1. The temperature of the sample is measured using a chromel-alumel thermocouple spotwelded to the edge of the sample. During the experiments the sample is grounded to prevent charging of the sample. Gases were dosed by backfilling the chamber with the desired gas or gas mixture. Exposures are reported in Langmuir (L), defined as Torr s. The pressure is measured using a standard UHV ion gauge. The base pressure of this system is mbar. This pressure is sustained by a turbo-molecular pump, in combination with an ion pump and a Ti sublimation pump. 2.2 Low energy electron diffraction Single crystal surfaces have a well defined surface structure, in which the atoms are arranged in an ordered lattice. Low energy electrons (<100 ev) that are elastically scattered from the surface region of the single crystal form a diffraction pattern that reflects the structure of the surface unit cell. A more detailed discussion about the theory and applications of low energy electron diffraction (LEED) can be found elsewhere [10]). Some clean single crystal surfaces show a structure that differs from that of the bulk-terminated structure, due to relaxation of the surface. These so-called surface reconstructions usually have their own periodicity, and this can be detected using LEED. When adsorbates interact with a surface the surface structure and the LEED pattern can be influenced. In the simplest case the adsorbate forms an ordered structure on the surface and as a result extra diffraction spots appear in the LEED pattern, characteristic of the periodicity of the adsorbate layer. In a more complex case the adsorbate influences the substrate structure itself and the spots due to the surface structure also change.

16 8 CHAPTER 2. EXPERIMENTAL METHODS AND EQUIPMENT 2.3 X-ray photoemission spectroscopy X-ray photoelectron spectroscopy (XPS, core level spectroscopy) provides information about the chemical nature and concentration of species adsorbed on the surface. Interaction of X-rays with the surface or adsorbate atoms can result in the creation of a core hole. The electron that is formed in this process is emitted. The electrons created in the surface region of the sample can reach the sample surface and are emitted into the vacuum. Their kinetic energy can be measured using an electron energy analyzer. The kinetic energy of the electron is related to the photon energy by E kin = E hν E bind, in which E kin is the kinetic energy of the electron, E hν is the energy of the X-ray photon and E bind is the E between the initial state (electron in the core level) and the final state (core hole, electron in vacuum). Each element has its unique electronic structure, and the kinetic energy of the photoelectron is element-specific. The chemical environment influences the electronic structure of the atom and this results in a chemical shift of the photoemission peaks, so that atoms of the same element in a different chemical environment can be distinguished. Illustrations of this phenomenon can be found in Chapter 7 and Chapter 6, in which the different NH xad species can easily be distinguished from N ad atoms. A more detailed discussion about photoelectron spectroscopy on single crystal surfaces and adsorbates can be found elsewhere [11,12]. The photoemission experiments presented in this thesis have been performed at the SuperESCA beamline of ELETTRA, the synchrotron light source in Trieste, Italy. Details of the experimental setup that can be found elsewhere [13 15]. The main advantages of the use of synchrotron radiation for core level spectroscopy are the high photon flux ( photons s 1 ), the high energy resolution of the X-ray photons and the ability to choose the photon energy. As a result, high resolution core level spectra can be obtained (resolution 0.02 ev) in a short time ( 30 s). During the measurements the sample could be heated up to 700 K, using radiative heating. All spectra reported in this thesis have been recorded at normal emission (X-ray incidence angle 70 o ), using photon energies of 400 ev, 496 ev and 650 ev for the C 1s, N 1s and O 1s spectra, respectively. The core level spectra were corrected by subtraction of a linear background and subsequently evaluated by fitting the peaks with a Doniach- Sunjić function [16], convoluted with a Gaussian function. Core Level Binding Energies of the different species were measured with respect to the Fermi level. The results were also corrected for the photon flux. In our last beamtime [in which Ir(111) was studied] we discovered that due to miscalibration of the photon energy the position of the Fermi edge depended on the excitation energy that is used (differences of 3 ev were found for different excitation energies). For the experiments on Ir(110) and Pt(410) we measured the position of the Fermi level only for one photon energy. As a result the absolute values obtained for the different BE species can deviate significantly from those reported in literature. Comparison with literature in this case can be done using BE s instead of absolute BE s. This approach worked well for all the results presented in this thesis. For the results obtained for Ir(111) the Fermi

17 2.4. SCANNING TUNNELING MICROSCOPY 9 level was frequently measured for all the photon energies used in the experiment, and in that case the absolute BE s are more accurate and can be used directly for comparison with literature values. 2.4 Scanning tunneling microscopy Scanning tunneling microscopy (STM) is based on the quantum mechanical principle of tunneling. When an atomically sharp tip is brought into close proximity (i.e. a few Ångstrom) to a conducting surface electrons can tunnel through the gap between tip and surface. The tunnel current depends strongly on the tip-surface distance. When the tip is mechanically scanned over the surface the tunnel current can be mapped as a function of the position on the sample. When a feedback mechanism is used to keep the tunnel current constant the tip will follow the contours of the surface. In this way the surface can be imaged with atomic resolution. A detailed discussion about the principles and applications of STM can be found elsewhere [17,18]. The microscope used in this study is a commercial variable temperature STM, manufactured by Omicron NanoTechnology GmbH 1. A slightly modified sample holder has been used, as described in Ref. [19], which allows measurement of the sample temperature during STM measurements. The sample could be cooled to 120 K at the STM position. Heating of the sample was achieved by radiative heating. The pictures presented in this thesis were measured in constant current mode, with a typical current of 0.5 na and a tip voltage between V and V. 2.5 Molecular beam experiments This Section provides a brief discussion about the generation and use of a supersonic molecular beam. A detailed discussion can be found elsewhere [20 23]. A molecular beam is formed by supersonic expansion of a gas, at a high pressure, through a small orifice ( 50 µm), the nozzle. During this supersonic expansion the rotational and translational energies of the molecules are cooled, and the energy is converted into translational motion along the direction of flow. The result is a highly directional beam of rotationally cooled molecules, with a narrow velocity distribution [20,21]. The exact distributions of energy depend on the details of the expansion. The kinetic energy of molecules in the beam may be increased by seeding with a lighter (i.e. lower molecular weight) gas, such as helium, and/or heating of the gas prior to the expansion. Heating the nozzle may also be used to increase the rotational and vibrational temperature of the molecules in the beam. The beam is then skimmed and collimated, before striking the sample surface. Molecular beam experiments can be used to determine the initial sticking coefficient (S 0 ), which is defined as the number of incoming molecules that stick on the 1

18 orifices skimmer QMS 10 CHAPTER 2. EXPERIMENTAL METHODS AND EQUIPMENT Normalized M=32 signal intensity A = S 0 t=0: beam flag closed, sample flag open t=30: beam flag open, sample flag open t=600: beam flag open, sample flag closed t=700: beam flag closed, sample flag closed mol. beam system gas in nozzle chopper beam flag sample flag UHV chamber sample chamber axis Time (s) Figure 2.1: O 2 uptake at 300 K, studied using a molecular O 2 beam. A (slightly altered) King and Wells experimental procedure has been applied. The inset shows a scheme of the experimental geometry that was used during the experiments. clean surface divided by the total number of incoming molecules. The most common experimental technique to measure this parameter is the so-called King and Wells (KW) procedure [24]. In the experiments presented in Chapter 5 a variation of this procedure has been applied. A typical measurement is shown in figure 2.1. It shows an O 2 uptake (using a molecular O 2 beam) on Rh(100), at a surface temperature of 300 K. The mass signal of interest (in this case m/e=32) was measured for some time while the beam flag was closed. The initially clean sample was exposed to the molecular beam at t=30 s (beam flag open, beam enters the experimental chamber). The molecules that did not stick to the surface were detected by the QMS (with this measurement the number of incoming molecules that stick on the clean surface is defined). This signal was continuously monitored, until the sticking coefficient had become zero, i.e. all molecules that reach the surface are scattered back into the gas phase. To check whether the saturation coverage was reached, an inert flag was put in front of the sample (at t=600), so that a reference signal was obtained, for which scattering is 100% (this measurement defines the number of incoming molecules). From these two data points S 0, defined as the number of incoming molecules that stick to the clean surface divided by the total number of incoming molecules, can be obtained. When additional information is available [25] about the adsorbate saturation coverage, θ t can be obtained via integration of the QMS signal. Several factors influence the accuracy of the absolute sticking coefficient obtained using the above mentioned method. The beam cross section at the sample

19 2.6. SAMPLE PREPARATION 11 surface should be slightly smaller than the sample surface, to ensure that all molecules detected by the QMS have interacted with the sample surface. The (exposuredependent) pumping speed of the walls of the vacuum system results in a higher apparent sticking coefficient, as these molecules do not appear in the gas phase, even when they have scattered from the sample surface. Methane at room temperature has a low sticking coefficient [26] 2, but by increasing the kinetic energy (using a seeded supersonic molecular beam) of the CH 4 molecules the sticking coefficient is increased [22,26 29]. A typical E kin of 0.71 ev has been used during the experiments presented in this thesis. This energy was obtained by heating the nozzle (diameter 50 µm) to 1073 K, while a 5% methane in He mixture was used. The results described in Chapter 5 were obtained at the Friedrich Alexander Universität in Erlangen-Nürnberg in the group of prof. H.-P. Steinrück, using a supersonic molecular beam setup. It consists of a supersonic molecular beam source and a rotatable, mass selective detector. A schematic picture of the experimental setup is shown in the inset of figure Details of the experimental system can be found elsewhere [30]. During our experiments the QMS was positioned perpendicular with respect to the surface normal at a distance of 90 mm from the sample, while the beam was impinging at normal incidence (see picture 2.1). The sample was spotwelded to two Ta rods and resistively heated, while the sample temperature was monitored using a Chromel-Alumel thermocouple, spotwelded to the side of the crystal. 2.6 Sample preparation The results described in this thesis have been obtained using several different single crystal surfaces, e.g. Pt(111), Pt(410), Ir(110), Ir(111) and Rh(100). The typical diameter of the (disc-shaped) samples is 6-10 mm and the thickness is 1 mm. The Pt(410), Rh(100) and Ir(111) surfaces were obtained from Surface Preparation Laboratories 4 and were cut to within 0.1 o of the desired crystal orientation. A grain size of 0.05 µm was used in the last polishing step. The Ir(110) and the Pt(111) surfaces were prepared within 1 o of the desired orientation, and polished down to a grain size of 1 µm [31]. A typical cleaning procedure in vacuum consists of ion etching and chemical cleaning. Bombardment with Ar + ions with a typical energy between kev (sputtering) removes the first few atomic layers, including contaminants present on the surface. The surface order is restored during a subsequent annealing procedure. During this annealing contaminants present in the bulk tend to diffuse to the surface, and a subsequent sputter-anneal cycles are necessary until the concentration 2 Gee et al. showed that the initial sticking coefficient of thermal methane (74 mev) on a stepped Pt surface is picture reproduced with permission from Ref. [30] 4

20 12 CHAPTER 2. EXPERIMENTAL METHODS AND EQUIPMENT of contaminants in the bulk of the crystal has been sufficiently depleted. Chemical cleaning is an effective method to remove specific contaminants. Heating the surface in the presence of O 2 is a good way to get rid of contaminants that form volatile oxides, carbon being the most important. The remaining O ad can be removed more easily, either by sputtering, by flashing above the O ad desorption temperature, or it can be reacted away via reaction with H 2. Both ion etching and chemical cleaning have been used to clean the Pt, Rh and Ir surfaces. For Pt annealing temperatures between 1000 K and 1100 K have been used, whereas for Rh and Ir the sample were annealed at higher temperatures, between 1200 K and 1500 K. The cleanliness of the samples was checked in different ways. Electron spectroscopies like Auger electron spectroscopy (AES) or XPS (Section 2.3) can detect the concentration and chemical nature of adsorbates on the surface and can therefore be used to check the presence of surface contaminants (except hydrogen). In specific cases one can use thermal desorption (Section 2.1) to detect the presence of remaining adsorbates, for example carbon. When the surface is exposed to O 2 and subsequently heated in vacuum the formation of CO and/or CO 2 indicates that there was still C present on the surface and additional cleaning is needed.

21 Chapter 3 Hydrocarbon oxidation on a stepped Pt surface In this Chapter the decomposition and oxidation of both n-butane and propene on a stepped Pt surface is discussed. The thermal decomposition of these hydrocarbons was studied using thermal desorption spectroscopy. The product formation rate and distribution were studied during steady state oxidation of n-butane and propene a function of surface temperature and reactant ratio. Propene decomposition and oxidation was also studied using high resolution core level spectroscopy (XPS). The high resolution XPS measurements provided information on the nature and concentration of the species present on the surface during decomposition and catalytic oxidation of propene. 3.1 Introduction The experiments discussed in this Chapter were focused on the decomposition and oxidation of two different hydrocarbons on a stepped Pt surface, in a wide temperature range. The two hydrocarbons that have been used (n-butane and propene) serve as model compounds for non-sticky and sticky hydrocarbons, respectively. Surface structure The different crystal facets of Pt have been used frequently as a model for Pt catalysts, especially the low index planes, i.e. (111), (110) and (100). The study of vicinal surfaces is motivated by the notion that defect sites present on a real catalyst particle might be the most active sites for the catalytic reaction. The surface that was used in this study, Pt(410), is an example of such a vicinal surface. The bulk-terminated surface consists of four atom wide {100} terraces and{110} steps (shown in Fig. 3.1). LEED experiments indicate that the surface restructures during the cleaning procedure that has been used. As a result some of the terraces are 13

22 14 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE fcc {410} fcc {111} V, 10 na 15 nm 10 5 a b c nm a 0.1 nm 0.2 nm b 2 4 nm 0.63 nm 6 c nm -0.2V, 10 na nm nm nm LEED picture of clean Pt(410) Figure 3.1: Top left: the bulk-terminated Pt(410) surface, consisting of four atoms wide {100} terraces and {110} steps. Top right: the bulk-terminated Pt(111) surface (used for the STM measurements). Middle and bottom: Structural details of the Pt(410) sample. The STM pictures show a high step density, and the (schematic) LEED pattern shows both the spots due to the bulk-terminated structure (black spots) and those due to reconstruction of {100} facets (grey spots).

23 3.2. N-BUTANE 15 broadened and these (100) terraces are large enough to reconstruct, in a similar way as the (100) surface [32,33]. The LEED pattern of the Pt (410) samples that was observed after cleaning, confirms that the (410) plane is restructured in the clean state. Next to the pattern of the bulk-terminated (410) structure (black dots in the schematic LEED pattern shown in Fig. 3.1, see for example Ref. [34]) we also observed the pattern that typical of the (5 20) type of reconstruction of the late 5d (100) surfaces [32]. We found that the (5 20) pattern disappears when the surface is exposed to either NO or CO, around room temperature. This indicates lifting of the reconstruction, similar to what was observed for Pt(100) [35]. Figure 3.1 shows two STM pictures (20 20 nm and nm respectively). These images show several terraces, with an apparent width of 0.6 nm. The picture shows, that the terrace width shows some variation. Next to the terraces (seen in the bottom right corner of the nm picture) there are also other structures visible, with a higher corrugation (the white line in the center of the nm picture). The height of these features is about twice that of the height observed for the normal steps ( 0.1 nm). The STM pictures show, that most terraces have a width corresponding to 4 atoms, which is expected for the bulk-terminated structure, but also, that surface reconstruction takes place, in which more complicated facets form. The LEED pattern indicates that the (5 20) reconstruction can take place. Large terraces were not observed in our (limited) STM experiments. The STM pictures showed that the terrace width is not constant, which indicates that some of the terraces are a bit broader (i.e. 5 atoms wide), and reconstruction can take place on these terraces. Reference desorption spectra Figure 3.2 shows H 2, O 2 [36], H 2 O and CO desorption spectra from our Pt(410) sample. These desorption spectra are relevant for the interpretation of the results in this Chapter and also for the results reported in Chapter 7. A more detailed discussion of oxygen adsorption and decomposition on this surface can be found there as well. 3.2 n-butane Saturated hydrocarbons interact weakly with most Pt surfaces [22]. Salmeron et al. [37] reported that both n-butane and n-pentane adsorb on Pt(111) at low temperature, but desorb molecularly below room temperature, without dissociation. They found a desorption barrier around 47.7 kj mol 1, in agreement with values found by Xu et al. [38]. According to Roke et al. [39, 40] neither n-butane nor isobutane dissociate on Pt(533), a stepped Pt surface with {111} terraces and {100} steps. Szuromi et al. [41], on the other hand, found n-butane dissociation on Pt(110).

24 16 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE H 2 desorption rate 10 L 5 L 3 L 2 L 1 L 0.5 L O 2 desorption rate x2 3 L 2 L 1.5 L 1 L Temperature (K) Temperature (K) (a) H desorption 2 (b) O desorption 2 CO desorption rate 5 L 3 L 2 L 1.5 L 1 L 0.5 L 0.2 L H 2 O desorption rate 10 L 5 L 3 L 2 L 1 L Temperature (K) Temperature (K) (c) CO desorption (d) H O desorption 2 Figure 3.2: Thermal desorption spectra from Pt(410) [3 K s 1 ]: (a) H 2 desorption after dosing H 2 at 100 K, (b) O 2 desorption after dosing oxygen at 100 K, (c) CO desorption after dosing CO at 300 K and (d) H 2O desorption after dosing water at 100 K.

25 3.2. N-BUTANE Adsorption, desorption and decomposition Figure 3.3 shows the thermal desorption spectra that were obtained for Pt(410) after different exposures of n-butane at 100 K. For low exposures we only observed H 2 desorption, in several steps, between 300 K and 600 K. For higher exposures (5 L, 10 L) molecular desorption was observed as well, in two distinct peaks at 178 K and 156 K respectively. Application of the Redhead approximation [7, 8] using a pre-exponential factor of s 1 (taken from Ref. [42]) yields desorption barriers of 53 and 46 kj mol 1, respectively. Szuromi et al. [41] also found two molecular desorption peaks on Pt(110), with activation barriers of 58.1 and 45.0 kj mol 1 respectively, very comparable to the values that we found. The lower value is also very close to the value reported by Salmeron et al. (47.7 kj mol 1 ) [37] for Pt(111). Higher exposures lead to the formation of n-butane multilayers, which desorb around 110 K. We did not find any indication of desorption of products other than H 2 and n-butane. The H 2 desorption due to n-butane decomposition can be divided into two regions, i.e. below 400 K and above 400 K. C x H y decomposition below 400 K results in H ad on the surface, which desorbs at its normal desorption temperature [see Fig. 3.2(a)]. Most of the n-butane dehydrogenation takes place below 400 K, and most of the H 2 (80%) desorbs in this temperature region. Integration of the total amount of H 2 that desorbs shows, that the carbonaceous species between 400 K and 500 K have a stoichiometry of C 4 H 2. These last hydrogen atoms disappear between 500 K and 600 K. The H 2 desorption spectra obtained after n-butane decomposition resemble qualitatively the results that were obtained for Pt(110) [41]. Since Pt(110) was reported to be the only Pt surface active for n-butane decomposition, we attribute the n-butane dissociation activity of the Pt(410) surface to the presence of {110} steps. The fact that the stepped Pt(533) ({111} terraces and {100} steps) does not show n-butane dissociation activity indicates, that not just the presence of steps, but the step structure is crucial. This points to a unique activity of {110}-like surfaces for saturated hydrocarbon dissociation [41,43 47] The effect of pre-adsorbed oxygen The effect of O ad on n-butane adsorption and decomposition was studied as well. Figure 3.4 shows the results that were obtained after dosing 5 L n-butane on an O ad pre-saturated surface. The molecular n-butane desorption peak is not influenced by the presence of O ad (the peak at 156 K was also observed, but for exposures >5 L) and both the peak positions and the amount of desorption are similar to those in the absence of O ad. A larger influence of O ad was found for the products of n-butane decomposition. H 2 desorption was not observed at all, and instead the formation of H 2 O was observed. H 2 O formation starts at 200 K, and it is reaction limited at these temperatures, because H 2 O ad desorbs below 200 K [see Fig. 3.2(d)]. This indicates that the first C x H y dehydrogenation steps take place around 200 K. This is in line with

26 number of desorbed H s 18 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE (a) H 2 desorption H 2 desorption Temperature (K) 10 L 5 L 1 L n-butane desorption (b) n-butane peak area exposure (L) 20 L 10 L 5 L 1 L Temperature (K) Figure 3.3: Thermal desorption spectra after dosing n-butane at 100 K (3 K s 1 ). Panel (a) shows H 2 desorption, formed during n-butane decomposition (the peak below 200 K is not H 2 desorption, but due to n-butane desorption). The inset shows an analysis of the area of the H 2 desorption spectrum for a saturated surface. Panel (b) shows the molecular desorption, the inset shows the area of the n-butane peak as a function of exposure.

27 3.3. PROPENE 19 n-butane Desorption rate H O 2 CO 2 CO Temperature (K) Figure 3.4: The effect of O ad on n-butane adsorption and decomposition. The figure shows the thermal desorption traces (3 K s 1 ) obtained after 5 L n-butane exposure on an O ad saturated surface. the conclusion of Szuromi et al. [41], who found n-butane dissociation on Pt(110) around the same temperature. A reaction between C x H y and O ad, so-called oxydehydrogenation (rather than C x H y dehydrogenation and subsequent H ad + O ad reaction), cannot be excluded on the basis of the thermal desorption data. CO 2 formation takes place between 300 K and 500 K, while a small amount of CO desorption is observed above 500 K as well. The formation of CO 2 shows that (oxygen-assisted) C-C bond breaking already takes place around 300 K. The CO 2 desorption signal shows two distinct peaks, one at 330 K and another at 450 K. This suggests that the formation of O-containing surface fragments occurs (for example carboxylate species, see Chap. 4), which subsequently decompose forming CO 2 (g). The fact that CO 2 and H 2 O are the major reaction products in this experiment (O ad saturated surface, subsequently exposed to n-butane) indicates that the amount of n-butane that dissociates is rather small when compared to the result of a similar experiment using propene (discussed in Section 3.3.2). As a result there is a large excess of O ad available, which results in H 2 O and CO 2 being the major reaction products. 3.3 Propene Unsaturated hydrocarbons, like ethene and propene, adsorb readily on Pt surfaces. Several authors studied the adsorption, decomposition and oxidation of propene, in most cases on Pt(111) [48 55]. For low surface temperatures propene adsorbs

28 20 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE molecularly on Pt(111), with the π-bond coordinated to the surface. It decomposes upon heating, and propylidyne (C-CH 2 -CH 3 ) has been identified around 300 K. Above 400 K the propylidyne species dehydrogenates, leaving a carbon-covered surface. Gabelnick et al. studied the effect of O ad on the propene adsorption and decomposition. Other authors reported data obtained on other metal surfaces, like Rh(111) [56], Ir(111) [57] and Ni(100) [58,59] Adsorption, desorption and decomposition Thermal desorption Figure 3.5 shows thermal desorption spectra that were obtained after dosing propene at 100 K on Pt(410). For low doses (<2 L) H 2 was the only gas phase product observed during heating, but for higher doses molecular desorption was observed as well. We did not find any indication of the formation of gaseous products other than H 2 and propene. After an exposure of 2 L a small molecular desorption peak was observed at 290 K, whereas for higher exposures a second, much larger, peak was observed at 228 K. Application of the Redhead equation [7, 8] using a pre-exponential factor of ±1 s 1 yields desorption barriers of 82±6 kj mol 1 and 64±4 kj mol 1 respectively. Lee et al. [50] reported a barrier of 58±3 kj mol 1 for propene desorption from Pt(111), close to the value that we found for the propene desorption peak at 228 K. Multilayer desorption was observed around 105 K, after exposures >5 L. The amount of H 2 formed via propene decomposition on the surface reaches a saturation point for exposures between 2 L and 3.5 L. For low coverage H 2 formation is observed between 300 K and 600 K, whereas for higher coverage the H 2 formation takes place between 250 K and 740 K. Hydrogen desorption after H 2 dosage takes place below 400 K [see Fig. 3.2(a)]. The H 2 desorption peaks observed after propene decomposition below 400 K are limited by the H ad combination reaction, whereas the desorption peaks above 400 K are limited by C x H y decomposition processes. The H 2 desorption due to propene decomposition on Pt(410) is slightly different than what was observed for Pt(111) [49, 54]. On Pt(111) most of the H 2 desorbs above 400 K (80%), while we observed (for the propene saturated surface) that most of the H 2 already desorbs below 400 K (65%). Evaluation of the area of the H 2 desorption peaks after propene saturates shows, that the stoichiometry on the surface at 400 K is C 3 H 3. Based on this we suggest, that the surface species around 400 K on Pt(410) is C 2 CH 3 [58] rather than propylidyne (C-CH 2 -CH 3 ). A more detailed discussion about the possible surface intermediates can be found in the next Section. The H 2 desorption spectra between 600 K and 800 K after propene dosage on Pt(111) and Pt(410) are very similar. H 2 formation in this temperature region was only observed after high propene doses. The underlying surface chemistry responsible for these peaks is discussed in more detail in the following Sections.

29 number of desorbed H s 3.3. PROPENE 21 (a) 2 H desorption H 2 desorption Temperature (K) L 2 L 1 L 0.5 L propene desorption (b) propene peak area exposure (L) 4 5 L 4 L 3.5 L 3 L 2 L Temperature (K) Figure 3.5: Thermal desorption after dosing propene on Pt(410) [3 K s 1 ]. Panel (a) shows the H 2 desorption due to propene decomposition and the inset shows an evaluation of the H 2 desorption spectrum of a propene saturated surface. Panel (b) shows molecular propene desorption and the inset shows the amount of molecular propene desorption as a function of exposure.

30 22 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE X-ray Photoelectron Spectroscopy High resolution XPS was used to get insight into the surface chemistry of propene. Details about the experimental procedures and data analysis are described in Chapter 2. Figure 3.6 shows the results that were obtained during heating of a propene saturated surface ( 125 K). The spectra reported above 700 K have been obtained by heating to this temperature and subsequently cool down to record the core level spectrum, heat to the next temperature, etc. The first spectrum in this figure represents the C 1s spectrum of a saturated propene layer at 125 K. The spectrum at this temperature contains a single broad peak, centered at a binding energy (BE) of ev. We were not able to resolve the three chemically different C atoms in the core level spectra, unlike Whelan et al. [58 60]. This is probably caused by the inhomogeneity of the surface, on which multiple propene adsorption sites can be present. As a result the C 1s spectrum contains several slightly different BE species and it is not possible to deconvolute the spectrum into the individual contributions. The analysis of the spectra is also complicated by the fact that different adsorption sites exist, and the decomposition pathway on for example step sites might be different from that on terrace sites. It was, therefore, not possible to analyze the C 1s spectra in detail, unlike Whelan et al. [58, 59] [Ni(100)] and Lee et al. [50] [Pt(111)]. Although a detailed analysis was not possible the spectra contain interesting information. The total C 1s intensity for example contains information about the amount of propene that adsorbs on the surface. The area of the C 1s peak was normalized using the C 1s signal from a CO saturated surface (at 300 K). The C 1s signal was subsequently divided by three, to account for the fact that each propene molecule contains three C atoms. The propene saturation coverage was found to be times the CO saturation coverage. Initially the surface is covered by molecular propene. About 25% of the propene that is initially present desorbs between 200 K and 300 K, resulting in a decrease of the C 1s signal. The spectral shape changes around 250 K, from a single broad peak to a peak with two maxima. This new spectral shape corresponds to the spectra that were reported for C 2 H x CH 3 species on Pt(111) and Ni(100). The shoulder at the low BE side of the peak corresponds to the signal from the -CH 3 group, while the other peak originates from the the -C 2 H x ) moiety of the surface species. The change of the spectral shape between 250 K and 425 K coincides with desorption of 50% of the hydrogen. As was shown previously, the stoichiometry at 425 K is C 3 H 3. The spectrum changes gradually between 425 K and 780 K, in the region where the remaining 50% of hydrogen desorbs. The most prominent change is the disappearance of the low BE shoulder, which was assigned to the -CH 3 group. This indicates that this group starts to dehydrogenate in this temperature region. Around 700 K a single, broad peak is observed, which is assigned to carbonaceous clusters on the surface. Evidence for the formation of such clusters is presented in the next Section. Above 600 K the total C 1s intensity increases. There was no C-source present in the gas phase during heating, so the increase of the C 1s must be caused

31 3.3. PROPENE Temperature (K) C 1 s intensity Temperature (K) 1000 Photoemission intensity Binding Energy (ev) 1200 K 690 K 425 K 130 K Binding Energy (ev) Figure 3.6: C 1s core level spectra taken during heating (0.5 K s 1 ) of an adsorbed propene layer. Some typical spectra are emphasized. The inset (left) shows a top view of the spectra and (right) the total area of the C 1s peaks vs. temperature.

32 24 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE by another process. In the next Section it is shown that huge morphological changes occur in the adsorbate layer between 400 K and 1000 K, involving the formation of 3D carbon clusters and sheets of graphite. These changes will influence the photoelectron yield during the XPS experiment and this is a probable explanation for the observed increase of the C 1s intensity. Above 780 K a new stage starts, in which both the peak shape and position change. All hydrogen has desorbed below this temperature, and changes that occur in the spectra are due to changes of the C xad layer. At 1200 K the spectra shows a sharp peak at ev, and a shoulder at the high BE side. The changes that take place between 780 K and 1200 K are assigned to the transition from an amorphous C xad layer to a graphite layer [50,61]. Scanning Tunneling Microscopy The chemistry of C x H y between 400 K and 870 K is rather complex and cannot be completely understood on the basis of H 2 desorption spectra and C 1s core level spectra. The STM setup described in Chapter 2 was used to study propene chemistry on Pt(111). As was shown in the previous Section the chemistry of propene on Pt(410) above 420 K shows a lot of similarities with that on Pt(111), reported by Lee et al. [50]. This means that our STM results obtained on Pt(111) can provide valuable information about the processes taking place on the surface above 400 K on Pt(410) as well. During these experiments a propene-saturated surface (120 K) was heated to the desired temperature and subsequently cooled down ( 120 K) to measure the STM pictures. Some typical STM pictures, obtained after annealing to different temperatures, are shown in figure 3.7. The C 1s core level spectra at these temperatures are shown as well. Land et al. [62] did similar experiments for ethene adsorption on Pt(111) and obtained very similar pictures. The picture taken at 140 K shows small protrusions, with an average height of 0.04 nm and an average diameter of 0.55 nm. These species are assigned to intact propene molecules. When the surface is heated to room temperature the pictures change. The surface species that are observed after heating to 300 K have an apparent height of 0.1 nm and a diameter of 0.55 nm. Koestner et al. [61] have shown that the propylidyne species forms an ordered layer on Pt(111) at room temperature, characterized by a p(2 2) LEED pattern. The STM picture also seems to indicate a hexagonal ordering of the propylidyne moieties, but the resolution is not high enough to clearly resolve the adsorbate structure. The surface is not completely covered at this point and some empty surface is observed as well. This is in line with thermal desorption data from this surface (shown for Pt(410) in Fig. 3.5), in which molecular propene desorption was observed around 230 K. Heating to higher temperature (600 K) changes the picture completely. The molecular signature has completely disappeared and instead larger clusters are observed. The average diameter of these clusters is 2.7 nm, much larger than that of the molecular species. The apparent average height of the clusters is 0.25 nm, about the height of a step on the Pt surface and similar to the average height observed

33 3.3. PROPENE na, V 25 nm Photoemission intensity nm Binding energy (ev) nm (a) 140 K Photoemission intensity Photoemission intensity Photoemission intensity Binding energy (ev) Binding energy (ev) Binding energy (ev) na, -1 V na, -1 V 0.5 na, -1 V nm nm 30 nm nm 0 5 nm 0 5 nm nm nm nm (b) 300 K (c) 600 K (d) 900 K Figure 3.7: STM pictures [obtained on Pt(111)] and the corresponding C 1s core level spectra [obtained on Pt(410)] after annealing of a propene saturated surface to different temperatures.

34 26 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE by Land et al. (0.22 nm, see Ref. [62]). According to thermal desorption data the species at this temperature contain only a small amount of hydrogen ( 25%), so the clusters can be described as carbonaceous (C x H y ) nano-particles. The clusters are distributed evenly over the whole surface and the rough appearance of the step edges indicates that the clusters also decorate the steps. Next to the clusters a large amount of empty surface is exposed. Heating above 900 K again changes the situation. Graphite forms above this temperature and the surface looks rather smooth in the STM pictures. Hexagonal structures typical of graphite [62] are observed on the terraces (diameter 8 nm) and the step edges also show hexagonal protrusions. The results of Land et al. [62], which were obtained with a higher resolution, show that the Pt atoms at of the step edges rearrange around the graphite patches. In the measurements presented here the resolution was not high enough to clearly distinguish graphite sheets from the Pt substrate, but we suggest that a similar process takes place as well. Images of a larger part of the surface (not shown here) show, that the shape of the step edges is very irregular, and that only a few graphite sheets are present on the terraces. This indicates that graphite formation preferably takes place at the step edges. Decomposition in the presence of propene (g) In another experiment the surface was slowly heated in the presence of propene (1 K s 1, mbar) in the gas phase. Figure 3.8 summarizes the results that were obtained, using thermal desorption, high resolution XPS and STM [Pt(111)]. The XP spectra below 500 K did not differ significantly from those obtained in the absence of propene and are therefore not shown. The H 2 desorption spectrum below 600 K is also very similar to that observed during heating of a propene saturated surface in vacuum. Above 600 K the presence of propene in the gas phase results in an extra H 2 desorption feature, which is accompanied by consumption of propene. This indicates that the hydrogen formation is due to adsorption of additional propene from the gas phase, which immediately decomposes upon adsorption. The intensity of the C 1s peak also indicates that the amount of C increases above 600 K. The amount of C present at 800 K is about three times the amount that was present after saturation with propene at low temperature. The STM results discussed in the previous Section offer an explanation for this phenomenon. The formation of carbonaceous clusters, which start around 600 K, frees a part of the surface, additional propene can adsorb and decompose on these empty parts of the surface, thereby filling the surface completely with carbonaceous deposits. Further heating up to 1200 K results in a large graphite peak in the C 1s spectrum. This graphite layer was very difficult to remove by chemical means, for example oxygen treatment ( mbar O 2, 1000 K), and ion etching (sputtering) was needed to remove the graphite layer. STM pictures of Pt(111) subjected to a similar treatment [shown in the inset of Fig. 3.8(b)] show, that the Pt(111) surface severely restructures during a similar treatment and the step edges become very irregular. A more detailed analysis of the pictures was not possible, due to extensive roughening of the surface.

35 Photoemission intensity nm 3.3. PROPENE 27 (a) propene ( consumption) MS intensity H (desorption) na, -1 V (b) nm 5 nm Temperature (K) Figure 3.8: Heating in propene. Panel (a) shows H 2 production and propene consumption (1 K s 1 ), in particular between 600 K and 800 K. Panel (b) shows the C 1s intensity during a similar XPS experiment (0.5 K s 1 ). The C 1s intensity shows a large increase between 600 K and 800 K. The inset in panel (b) shows the STM picture obtained after a similar treatment on Pt(111).

36 28 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE The effect of oxygen Pre-adsorbed oxygen The effect of O ad on the adsorption and decomposition of propene was also studied, using both thermal desorption techniques and high resolution XPS. The results of these experiments are shown in figure 3.9. For all the experiments presented there the surface was initially saturated with O ad and subsequently exposed to different amounts of propene. During the XPS measurements the surface was pre-covered with a mixed O ad /O 2ad overlayer. We did not find any indication that O 2ad plays a special role other than providing some additional ( 10%) O ad. Panel (a) shows the molecular propene desorption in the presence of O ad. Two distinct desorption features are observed, at 290 K and 260 K. This corresponds to desorption barriers of 83±6 kj mol 1 and 74±5 kj mol 1 respectively (Redhead). The high temperature desorption peak is not affected by the presence of O ad, but the low temperature desorption peak is significantly shifted to higher temperature ( 30 K, 10 kj mol 1 ) in the presence of O ad. This indicates a stronger adsorption of propene in the presence of O ad. The amount of molecularly desorbing propene is also lower in the presence of O ad. The shape of the C 1s XP-spectra during heating in the presence of O ad differs only slightly from those in the absence of oxygen and the spectra are, therefore, not shown. We also did not find any evidence (i.e. a new peak) in the C 1s or O 1s spectra for the formation of C x H y O z species during these experiments. However, during the experiment presented in Section (steady state propene oxidation) we did observe a very small O-containing species in the C 1s spectrum, between 250 K and 400 K (details can be found in Section 3.4.2). Its intensity is rather low, and it is only a minor product under the experimental conditions that were used. The similarity of the spectra in the absence and presence of O ad indicates that O ad does not influence the propene decomposition process directly but it only influences the products of propene dissociation, i.e. H 2 O rather than H ad and CO 2 /CO rather than C. The saturation coverage of propene on the presence of O 2ad /O ad is 70-80% of the saturation coverage in the initially clean surface. It is not clear whether this lower saturation coverage is caused by O ad or O 2, since both species were present during the adsorption of propene during the XPS measurements. The lower molecular propene desorption observed during the TPD experiments (which were done with an O ad but not O 2ad covered surface) suggests that the lower saturation coverage is caused by O ad, but we do not have direct evidence from XPS to confirm this. Gabelnick et al. [49] found that the amount of molecularly desorbing propene from Pt(111) decreases with increasing O 2ad coverage. The panels (b-e) show the desorption spectra of the different gas phase products formed during propene decomposition in the presence of O ad. In these experiments the O ad pre-covered surfaces was exposed to different doses of propene. Next to molecular propene desorption (discussed in the previous paragraph) we observed the desorption of CO 2 (b), H 2 O (c), CO (d) and H 2 (e). The exact product distri-

37 O1s intensity (b) (d) H O desorption intensity 2 H desorption intensity 2 C1s intensity 3.3. PROPENE 29 propene desorption intensity (a) no O ad 5 L 3.5 L 1 L Temperature (K) CO 2 desorption intensity in O L 0.75 L 1 L 5 L (c) in O 2 5 L 1.5 L 1 L 0.5 L CO desorption intensity in O 2 5 L 3.5 L 2 L 1 L 0.5 L (e) no O ad 5 L 3.5 L 2 L 1 L Temperature (K) (f) (g) C1s intensity CxH yad CO ad O ad O 2ad e Temperature (K) Figure 3.9: The effect of O ad on the propene surface chemistry. (a-e) show thermal desorption spectra (3 K s 1 ) for the different products, obtained after various doses of propene on an oxygen pre-covered surface. (f,g) show the data obtained using XPS (0.5 K s 1 ). The results shown in panel (f) was measured in the absence of O 2 in the gas phase (but with an O ad /O 2ad pre-coverage) and those shown in panel (g) in the presence of mbar O 2.

38 30 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE bution depends strongly on the propene dose that was used. We did not find any indication of the formation of other gaseous partial oxidation products. Panel (f) shows the results of a TP-XPS experiment during heating of a surface which was precovered with O ad /O 2ad and subsequently saturated with propene. This experiment corresponds to the TPD traces taken after a propene exposure of 5 L. The O ad (two peaks, at ev and ev respectively, see Chapter 6) signal initially shows a small increase (due to O 2ad (observed at ev) decomposition, below 200 K), but above 200 K O ad starts to be consumed, and all O ad is removed around 350 K. The formation of H 2 O (g) is also observed in this temperature region, as well as a small amount of CO 2 (XPS results correspond to desorption spectra after 5 L propene). The amount of CO 2 that is formed under these conditions is rather small and the C x H y intensity decreases only slightly. A small amount of CO ad is observed between 300 K and 550 K, both in the C 1s (286.4 ev) and the O 1s (533.9 ev) spectrum. In the figure only the CO signal obtained from the C 1s spectrum is shown 1. This CO ad desorbs between 450 K and 550 K (also observed in thermal desorption). The formation of CO ad around 300 K indicates that some C-C bond breaking already takes place at this temperature. During the thermal desorption experiments H 2 desorption was observed as well. In the experiment using a propene exposure of 5 L the amount of H 2 that is formed is rather large. The amount of H 2 formed after propene saturation on an O ad covered surface is only 25% lower than the amount of H 2 formed in the absence of oxygen. These observations show, that the amount of oxygen present on the saturated surface is small compared to the amount of H ad and C that is formed due to the decomposition of a saturated propene layer. For lower propene doses the product distribution is different. The amounts of CO, H 2 and H 2 O decrease, but the amount of CO 2 increases. After a post dosage of 0.5 L propene the amount of H 2 desorption is low and the amount of CO 2 desorption is high, indicating that in that case the relative amount of O ad high. Karseboom et al. [57] found on Ir(111) that oxygenated surface species form when both O ad and propene are present. In our XPS measurements a high propene dose was used, and there was only a very small indication that an oxygen-containing surface species (reported in the next Section) was formed. The XPS measurements were not performed for lower propene doses, and it is very well possible that oxygen-containing species do form under O ad rich conditions. This notion offers an alternative explanation for the low temperature CO 2 formation peaks observed during the thermal desorption experiments. In that case these peaks are due to the oxidation/decomposition of such oxygen containing species (for example carboxylate species, R-CO 2ad ) and oxygen in that case plays a very important role in the activation of the C-C bonds. When the amount of O ad is small compared to the amount of propene (in the case of the XPS measurements) all O ad is removed via reaction with H ad and the formation of oxygen-containing surface species does not take place. 1 The C xh y signal was calibrated using the CO ad saturation coverage at 300 K. This result was divided by 3, so that 1 mole of adsorbed propene is equal to 1 mole of CO ad. The CO ad signal in the figure was not divided by 3.

39 3.3. PROPENE 31 We propose the following model to explain our observations for high propene doses. The adsorbed propene layer starts to decompose around 200 K (the normal propene decomposition temperature), forming H ad, which immediately reacts with O ad to form water. A small amount of O ad reacts with C and forms CO ad on the surface. An alternative route for this CO formation is the initial formation of a C x H y O species, which decomposes into CO and C x H y (see results in Section 3.4.2). When all O ad is consumed the remaining C x H y species decompose in the normal way, into H 2 and carbonaceous clusters/graphite. Under these conditions the presence of O ad has only a small influence on the propene surface chemistry. For low propene dosages the situation is more complex. There is enough O ad present to react with the hydrogen and there is even O ad left to react with the carbon moieties as well. The formation of oxygen containing surface intermediates (for example propoxy or propionate species) is possible. Decomposition of these species and/or C-C bond breaking and subsequent C-oxidation result in the formation of CO 2 already around 350 K. Under these conditions the presence of O ad has a large influence on the propene surface chemistry. Oxidation by O 2 from the gas phase More insight into propene oxidation was obtained by an experiment in which a O ad (-O 2ad )-propene (sat.) layer was heated in the presence of O 2 (g). The temperature programmed oxidation results are shown in figure 3.9(b-d), and TP-XPS results are the is shown in panel (g). For the thermal desorption spectroscopy experiments (labeled in O 2 in the figure) a heating rate of 3 K s 1 and an O 2 pressure of mbar were used, while for the TP-XPS experiments an O 2 pressure of mbar was used (heating rate 0.5 K s 1 ). During the XPS experiments we found that the C x H y decomposition below 450 K proceeds in a similar way as in the absence of O 2 (g). The amount of CO ad formed above 300 K is also similar, indicating that the O 2 from the gas phase cannot reach the surface, i.e. O 2 adsorption/dissociation is blocked by C x H y species. Above 450 K the situation changes drastically. O 2 can reach the surface and C x H y is rapidly removed. Above 550 K the surface is covered with O ad and there is no C x H y present anymore. In the thermal desorption spectra similar behavior was found. Below 600 K the desorbing products are similar to those found in the absence of O 2 2. Both the absence of H 2 desorption (not shown) and H 2 O formation between 400 K and 650 K show, that small amounts of O 2 reach the surface, react with the hydrogen formed in the propene deocmposition process and form H 2 O. Above 600 K the formation of both CO 2 and CO is observed. The amount of CO 2 and CO is much larger than in the absence of O 2. This is in line with the XPS results, where it was shown that 80% of the initially present C is removed between 450 K and 550 K. The 2 Note that the scaling of the desorption traces in this experiment is different from that of the other results and absolute amounts cannot be compared.

40 32 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE temperature difference between the onset of C oxidation in TPD experiment and the XPS experiment is probably caused by the difference in the heating rates used in the different experiments. Gabelnick et al. reported thermal desorption data for co-adsorption of oxygen and propene on Pt(111) [49]. Their observations concerning the effect of O ad are very similar to the results that we found (i.e. CO 2 formation around 350 K, decreasing with increasing propene dose, H 2 production for higher propene doses, H 2 O formation between 300 K and 400 K). This indicates that the surface structure only plays a secondary role during propene decomposition in the presence of oxygen. 3.4 Steady state oxidation The steady state oxidation reaction was studied as well, both for n-butane and propene. During these experiments the surface was slowly heated (and cooled) in the presence of a reaction mixture of n-butane (/propene) and O 2. This experiment was done for several different propene:o 2 ratios. The overall reaction equations for both n-butane and propene oxidation are shown in the following equations. Two possible reaction pathways are shown, e.g. partial oxidation [Eqs. (3.1) and (3.3)] towards CO and H 2, and total oxidation [Eqs. (3.2) and (3.4)], towards CO 2 and H 2 O. 2 C 4 H 10 (g) + 4 O 2 (g) 10 H 2 (g) + 8 CO (g) (3.1) 2 C 4 H 10 (g) + 13 O 2 (g) 5 H 2 O (g) + 8 CO 2 (g) (3.2) 2 C 3 H 6 (g) + 3 O 2 (g) 6 H 2 (g) + 6 CO (g) (3.3) 2 C 3 H 6 (g) + 9 O 2 (g) 6 H 2 O (g) + 6 CO 2 (g) (3.4) These equations show that for total n-butane oxidation a O 2 /n-butane ratio of at least 6.5 is needed, whereas for total propene oxidation a ratio of at least 4.5 is sufficient. An important parameter that needs to be taken into account when the reaction is catalyzed by a solid surface is the reactive sticking coefficient for both reactants. When the O 2 sticking coefficient differs a lot from that of the hydrocarbon the reactant ratio at the surface will be different than the reactant ratio in the gas phase. Another complicating factor is the interaction between the adsorbates. As we have shown in the previous Section the presence of C x H y species can block O 2 adsorption. At high temperature the surface can become covered with a graphite layer, which has a large influence on the accessability of the surface for O 2 as well n-butane Figure 3.10 shows the reaction products during steady state n-butane oxidation. The results that are shown were obtained using an n-butane/o 2 ratio of 1:1 and

41 formation rate (a.u.) 3.4. STEADY STATE OXIDATION ratio 1:1 ratio 1:5 heat 8 8 heat formation rate (a.u.) H H 2 O CO 4 4 CO cool cool Temperature (K) Figure 3.10: Reaction products of n-butane oxidation during heating and cooling in the presence of the reactants in the gas phase. The results for two different reactant ratios are shown.(heating rate 1 K s 1, mbar n-butane, O 2 pressure according to ratio, 1 data point per 2 K). Note that the scale of the axes for the results of the cooling branch are different from those for the heating by a factor of two. 1:5 respectively. The measured signals of H 2 (m/e=2), CO (m/e=28) and CO 2 (m/e=44) are partly due to n-butane. The signal for m/e=43 (the highest peak for n-butane, not shown) was, therefore, used to correct the other mass signals, to remove the contribution of n-butane. The signals of the different products were not corrected for differences in sensitivity of the QMS, so the intensity of different products can not be compared directly, but the intensities of one mass can be used for comparison of the different reactant ratios. For a ratio of 1:1 the first reaction products appear around 600 K. Initially H 2 and CO are the main reaction products, but above 800 K the amounts of both H 2 and CO decrease slightly, while the formation of H 2 O increases. The formation of CO 2 was not observed under these reaction conditions. In the cooling branch the formation rate of CO remains constant down to 600 K, where the reaction stops.

42 34 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE The H 2 production decreases during cooling by 50%, and H 2 O is observed instead. The hysteresis that was observed can be explained by the fact that the surface was initially covered with C, and the first products during the heating are, therefore, CO and H 2. In the cooling branch the amount of C on the surface is probably lower, and O 2 has more access to the surface. As a result H 2 O is formed instead of H 2. For the ratio 1:5 the product distribution is different. The onset of the reaction, around 400 K, is marked by a very large CO 2 formation peak. The H 2 O formation trace also shows a small peak in this temperature region. Above 600 K both H 2 O and CO 2 are the major products, while the formation rates of H 2 and CO are very low. Above 800 K the selectivity changes. The amount of CO 2 drops, the H 2 O formation rate stays constant and the CO formation rate increases, i.e. the C-selectivity changes from CO 2 to CO whereas the H-selectivity doesn t change. In the cooling branch the C-selectivity changes back to CO 2 at the same temperature. Both H 2 O and CO 2 formation rates increase between 600 K and 500 K in the cooling, but they drop rapidly to zero below 500 K. The large CO 2 formation peak that was observed in the heating is absent in the cooling, indicating that it is caused by a non-steady state process. This peak is assigned to oxidation of the carbonaceous species that were initially present. Above 400 K they are quickly oxidized (because there is enough O 2 available), giving rise to a transient CO 2 peak. The selectivity change from CO 2 to CO takes place above 800 K. This temperature corresponds to the recombinative O ad desorption temperature from this surface [see Fig. 3.2(b)]. The change of the C-selectivity, from CO 2 to CO, is, therefore, explained by a decrease of the O ad concentration on the surface. A comparison between the two reactant ratios is complicated by the fact that different reaction products are formed for the different ratios. Since the n-butane pressure was kept constant for both ratios the consumption of n-butane is a good measure for the reactivity. When the thermal behavior of the n-butane signals (not shown) of the two ratios are compared it becomes clear that the reaction rate below 700 K is higher for the ratio 1:5, while it is similar for both ratios above 700 K. We suggest that the presence of C in the first heating branch causes this difference. When the O 2 partial pressure is low it takes more time and a higher temperature to remove this carbon (i.e. the reaction rate depends on the O 2 partial pressure). When the excess carbon is removed (above 700 K) the reaction proceeds in a similar fashion for both ratios, i.e. the adsorption/decomposition of n-butane is rate-limiting Propene Some of the typical results that for propene oxidation are shown in figure The top panel shows the surface coverage (obtained by XPS) during propene oxidation (ratio 1:5), whereas the bottom figures show the gas phase products for the ratios 1:5 and 1:20. Note that the y-scale for both heating and cooling are the same in this figure, but that the y-scale for the different ratios is not the same. The (arbitrary) numbers can again be used, but only for comparison of one mass, not for comparison between

43 formation rate (a.u.) 3.4. STEADY STATE OXIDATION 35 ratio 1:5 C1s intensity (a.u.) x5 Photoem. intensity 420 K 300 K R-C=O (?) CO ad C x H y predicted 'C' R-C=O (?) CO Binding energy (ev) Temperature (K) heat ratio 1:5 heat ratio 1:20 20 formation rate (a.u.) cool H 2 H 2 O CO CO 2 cool Temperature (K) Figure 3.11: Reaction products of propene oxidation during heating and subsequent cooling in the presence of the reactants in the gas phase. Two different reactant ratios are shown (heating rate 1 K s 1, mbar propene,1 data point per 3 K), The O 2 pressure was adjusted according to ratio, while the propene pressure was kept constant. The XPS results shown in the top panel were obtained using a reactant ratio of 1:5.)

44 36 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE the different masses. The signals of H 2 (m/e=2) and CO (m/e=28) are partly due to propene. The signal for m/e=41 (the highest signal for propene, not shown) was used to correct the other mass signals. X-ray Photoelectron Spectroscopy The XPS results shown in the top panel of figure 3.11 were obtained using an O ad pre-covered surface, which was subsequently exposed to the reaction mixture ( mbar propene, mbar O 2 ) and heated up to 600 K (0.5 K s 1 ). The surface was heated further up to 1200 K and slowly cooled (both in the presence of the reaction mixture), but due to experimental limitations it was not possible to measure core level spectra in this temperature region. The behavior that we expect for the C 1s signal between K, based on the results shown in Section 3.3, is depicted as a dotted line in the figure. Below 600 K the thermal behavior of the C x H y species is very similar to what was observed in the presence of propene (Section 3.3). The total C x H y intensity starts to increase above 500 K, due to the start of carbon cluster formation. After heating to 1200 K in the presence of the reactants the XP spectrum indicates that the C xad is present in the form of graphite. The total amount of C is 4 times higher than at the start of the experiment (i.e. a propene saturated surface). The similarity with the experiment in the presence of only propene indicates that the O 2 partial pressure is not high enough to remove all the carbonaceous deposits, and the surface coverage is dominated by C x H y species for all temperatures studied. Below 500 K the presence a small amount of CO ad (4% of the CO saturation coverage) was observed, at a BE of ev. Interestingly we also observed another C 1s species at a BE of ev. Two C 1s spectra showing the both the ev and ev species are depicted in the inset of the top panel. The ev species is already present at 250 K (the lowest temperature used in this experiment) and decomposes upon heating into CO ad, between 300 K and 400 K. The fact that it decomposes into CO ad indicates that this species is a partial oxidation product, i.e. C x H y O. During experiments on Ir(111) [see also Chapter 4] using methanol and ethanol we also found a BE component at the low BE side of the CO ad peak, with a BE of 0.7 ev with respect to the CO ad BE. This species was also assigned to a C x H y O intermediate. Carboxylate species, on the other hand, appeared at the high BE side of CO ad. These observations on Ir(111), combined with the experimental observation that the ev species converts to CO ad rather than to CO 2 (g), indicates that the intermediate on Pt(410) [with a BE of ev] should be assigned to a R-C=O rather than R-CO 2 intermediate. The possible role of oxygenate intermediates is already discussed in the previous Sections concerning the effect of O ad on the propene decomposition. At this point we do not have enough information to be conclusive about the role of the R-C=O intermediate, and further investigations are, therefore, needed.

45 3.4. STEADY STATE OXIDATION 37 Gas phase products The main reaction products observed in the first heating branch, using a pro-pene/o 2 ratio of 1:5, are H 2 and CO. The H 2 peak between 600 K and 800 K occurs in the same temperature region as the H 2 formation peak during heating propene (g), and it is accompanied by an increase of the C 1s intensity observed with XPS. This shows that the O 2 partial pressure is not high enough to remove all the carbon formed during propene decomposition. In the cooling branch both CO and H 2 are the major products, while a small amount of H 2 O is observed as well. The reaction rate in the cooling seems to be slightly lower than in the heating. XPS showed that the surface is covered with a large amount of graphite in the cooling branch, so it is remarkable that the reaction can still proceed on the surface. In a subsequent heating (not shown) the observed H 2 and CO formation rate are similar to those observed in the cooling branch. The formation of (unreactive) graphite has a negative effect on the reaction rate, and the selectivity is influenced as well. The presence of graphite is, however, not detrimental for partial oxidation, in which CO and H 2 are the desired products. The presence of graphite is even useful, because it hinders O 2 adsorption so that the concentration of oxygen on the surface remains low and the partial oxidation products H 2 and H 2 O are formed, instead of the deep oxidation products H 2 O and CO 2. When the propene/o 2 ratio is changed to 1:20 the products of the reaction are different. In the first heating the H 2 formation peak between 600 K and 800 K is still present, indicating that the formation of carbonaceous clusters and decomposition of additional propene takes place even in the presence of a high O 2 partial pressure. In contrast to the ratio 1:5 the situation changes around 750 K. We explain the high CO 2 and CO peaks observed around 750 K to oxidation of the carbonaceous layer formed between 600 K and 800 K. After these transient peaks in CO 2 and CO formation a steady state reaction regime is reached, in which CO 2 and H 2 O are the major products. Similar to what was observed for the steady state oxidation of n-butane the C-selectivity changes around 900 K, from CO 2 to CO, while the H-selectivity remains unaltered. We explain this in the same way as in the case of n-butane oxidation, i.e. selectivity towards CO 2 below the O ad desorption temperature ( 800 K), and selectivity towards CO above this temperature. In the cooling branch the selectivity changes again at the same temperature, from CO to CO 2, and due to the absence of a high C coverage in the cooling branch the reaction can continue down to 400 K. The results obtained during a subsequent heating (not shown) were similar to those obtained during the first heating, indicating that a carbonaceous layer again builds up again below 400 K, and this layer has to be removed before the steady state reaction can take place. H 2 production between 600 K and 800 K, indicative of the dehydrogenation of carbonaceous clusters, was again observed. In contrast to the ratio 1:5 the surface carbon is effectively removed when a ratio of 1:20 is used. As a result the situation after the cooling is similar to the situation before the first heating, in contrast to the situation for the second heating branch using a ratio of

46 38 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE 1:5, where there is a large amount of (unreactive) graphite on the surface after the cooling (seen by XPS). The surface is also more reactive when a higher O 2 partial pressure is used. The H 2 O formation around 800 K in the cooling branch is about ten times higher when a propene/o 2 ratio of 1:20 instead of 1:5 was used Saturated vs. unsaturated hydrocarbon oxidation The results for n-butane and propene are similar in several ways. For both hydrocarbons the selectivity of the reaction depends on the reactant ratio. For low O 2 partial pressures H 2 and CO are formed, while for high O 2 partial pressures the major products are CO 2 and H 2 O. The change of the C-selectivity during oxidation for high O 2 partial pressures changes around 850 K from CO 2 to CO, while the H- selectivity remains unaltered. In both cases this was assigned to the decrease of the O ad concentration, due to O ad desorption. There are also several remarkable differences. Despite the overall reaction stoichiometries [see Eq. (3.2) and (3.4)] the O 2 partial pressure necessary for complete oxidation (to H 2 O and CO 2 ) is much lower for n-butane than for propene. For n-butane a ratio of 1:5 is enough to get complete oxidation, whereas for propene a ratio of 1:20 is needed. The most important difference between n-butane and propene is caused by the fact that propene adsorbs/decomposes readily on the surface, whereas for n-butane the dissociative adsorption probability is much lower and might be limited to specific sites, such as the {110} steps. The high dissociation probability of propene results in a high C ad coverage which, at high surface temperatures, results in the formation of unreactive graphite. Only when a relatively large amount of O 2 is present this can be avoided, and in that case the C x H y is oxidized around 750 K, and the C ad coverage is probably low at the temperatures where graphite formation takes place. For n-butane the dissociation probability is lower, O 2 adsorption and dissociation is, therefore, not inhibited and C ad oxidation proceeds already for lower O 2 partial pressures. In general, it can be said that poisoning by carbon deposition is an important issue when unsaturated hydrocarbons (i.e. sticky hydrocarbons) are used, whereas in the case of saturated hydrocarbons (i.e. non-sticky hydrocarbons) carbon deposition can easily be prevented. The differences between n-butane and propene oxidation show, that the ratio of the reactants in the gas phase does not necessarily translate into surface coverage, and the product selectivity for a specific reactant ratio can, therefore, be very different from the selectivity that follows from the overall reaction equations [see Eq. (3.2) and (3.4)]. 3.5 Summary and conclusions In this Chapter the results of our study of hydrocarbon oxidation on Pt(410) are presented. Both a saturated (n-butane) and an unsaturated (propene) hydrocarbon

47 3.5. SUMMARY AND CONCLUSIONS 39 were studied, representing non- sticky (n-butane) and sticky (propene) hydrocarbons, respectively. n-butane adsorbs both molecularly and dissociatively on this particular Pt surface. The activity of our Pt(410) surface is assigned to the presence of {110} steps, because Pt(110) is the only low index surface that shows activity for dissociative adsorption of low weight alkanes. The presence of O ad does not influence the molecular desorption of n-butane. We suggest that the n-butane dissociation in the presence of O ad proceeds in the same way is on the clean surface, but that the reaction products react with O ad, forming H 2 O, CO and CO 2. The fact that CO 2 rather than CO is the most prominent C-containing product indicates, that only a small amount of n-butane dissociates, so that there is an excess of O ad available. The adsorption and decomposition of propene was studied as well, using several experimental techniques. Thermal desorption of a surface saturated with propene at low temperature ( 100 K) showed both molecular propene desorption an H 2 desorption (due to dissociation of propene). Using X-ray photoelectron spectroscopy it was found that the major part (80%) of the adsorbed propene decomposes rather than desorbs. The C 1s core level spectrum changed significantly at specific temperatures during heating, but due to the heterogeneity of the surface it was not possible to exactly assign the spectra to specific surface intermediates. Scanning Tunneling Microscopy showed (for Pt(111)) that the low temperature (<600 K) adsorbates can be described in terms of molecular species, while the adsorbates above 600 K are more appropriately described as carbonaceous clusters, i.e. larger 3D-clusters with a variable (temperature dependent) C x H y stoichiometry (y/x <1). Above 900 K the formation of graphite is observed. When the surface is heated in the presence of propene (g), the formation of carbonaceous clusters gives rise to additional propene adsorption and decomposition. During the formation of the clusters a part of the metal surface is exposed, and more propene can decompose on these empty sites. The influence of pre-adsorbed oxygen on the propene surface chemistry seems to depend strongly on the O ad /propene ad ratio. High resolution XPS measurements (which were done for oxygen saturated surfaces post-saturated with propene) indicated, that O ad does not have a large influence on the decomposition of propene, but only consumes the dissociation products H ad and C ad (forming H 2 O, CO ad, CO and CO 2 ). The XPS measurements also provide some evidence for the formation of a R-C=O species at low temperature, which decomposes around 300 K forming CO ad. For high propene doses on an oxygen saturated surface the amounts of CO 2 and CO are low, and H 2 formation is observed, next to the formation of H 2 O. This is caused by the fact that a large fraction of the adsorbed propene decomposes, giving rise to a large amount of H ad and C x H y. All O ad is removed via reaction with H ad, and there is no oxygen left to react with the remaining hydrogen and carbonaceous species. When the propene dosage is lowered the formation of R-C=O species is more probable and it is suggested that oxygenate species play crucial role in the propene decomposition mechanism in that case. More (for example XPS) experiments would needed to answer this question. For steady state oxidation of both n-butane and propene it was found that there

48 40 CHAPTER 3. HYDROCARBON OXIDATION ON A STEPPED PT SURFACE is a large influence of the hydrocarbon/o 2 ratio on the activity and selectivity. For relatively low O 2 partial pressures H 2 and CO are the major products, whereas for relatively high O 2 partial pressures CO 2 and H 2 O are the major products. There is also a large difference between n-butane and propene in the sense that the partial pressure of O 2 needed to get complete oxidation (H 2 O and CO 2 ) is much lower for n-butane than for propene. This is explained by the fact that propene decomposes readily on the surface and inhibition by carbonaceous carbon or graphite is an important issue. In the case of n-butane the dissociative sticking probability is lower, and it might also be limited to certain surface sites. Oxygen adsorption and dissociation is, therefore, not hindered and CO 2 and H 2 O formation occurs at relatively low O 2 partial pressures. The carbonaceous deposits that form during propene oxidation can only be effectively removed when a high O 2 partial pressure is used. Graphite formation was observed during propene oxidation, when a low O 2 partial pressure was used. It is not necessarily a negative process, because partial oxidation products H 2 and CO were observed even in the presence of large amounts of graphite on the surface. In that case graphite even plays a beneficial role, by blocking the O 2 adsorption/decomposition, thereby keeping the O ad concentration low. As a result partial oxidation products, like CO and H 2 were formed. The specific structure of the surface seems to be more relevant for the n-butane than for propene. For n-butane the presence of {110} structural elements is necessary for n-butane dissociation. Propene on the other hand dissociates readily on Pt surfaces. A comparison of the thermal desorption spectra that were obtained for Pt(410) with those obtained on Pt(111) show only minor differences. The surface structure therefore seems to play only a minor role for propene decomposition.

49 Chapter 4 Benzene chemistry on Ir(111) In this Chapter the surface chemistry of benzene on Ir(111) is discussed. Molecular adsorption of benzene was found between 170 K and 350 K, and above this temperature decomposition of benzene is observed. An ordered adsorbate structure has been observed upon adsorption around 335 K. Decomposition proceeds via two different pathways, i.e. stepwise removal of H atoms (via C 6 H x species) and CH ad formation, respectively. The decomposition pathway was found to depend on the benzene coverage. The presence of a saturated O ad layer (0.5 ML) weakens the benzene adsorption. Decomposition in the presence of O ad proceeds partly via CH ad formation, and a very small amount of a carboxylate intermediate was observed as well. 4.1 Introduction In this Chapter our results for benzene (C 6 H 6 ) adsorption and decomposition on Ir(111) are reported. The structure of the fcc(111) surface is shown in Figure 3.1. The adsorption and decomposition of benzene has been studied previously, on several metal surfaces. Nieuwenhuys et al. [63], Mack et al. [64] and Netzer et al. [65] studied the benzene chemistry on Ir(111), but only that of a saturated benzene layer, which was prepared around 300 K. Nieuwenhuys et al. found both molecular benzene desorption (a single peak), and H 2 formation (due to benzene decomposition), at temperatures similar to those observed in our experiments. Both Mack et al. [64] and Netzer et al. [65] reported that molecular benzene adsorbs with the molecular plane parallel to the surface. The molecular structure is distorted upon adsorption, and a C 3v symmetry was observed. Koel et al. [66] compared Rh(111) with Pt(111) and Pd(111) and concluded that the mechanism of benzene decomposition is fundamentally different on Rh. They reported that decomposition on Rh starts with C-C bond breaking, forming C 2 H 2 and CH ad intermediates that subsequently dehydrogenate. On Pt(111) and Pd(111) the initial decomposition step is C-H bond breaking, and decomposition proceeds 41

50 42 CHAPTER 4. BENZENE CHEMISTRY ON IR(111) via C 6 H x intermediates. In this study we present a detailed study of the benzene surface chemistry on Ir(111). By using several techniques, including high resolution fast XPS, temperature programmed desorption (TPD) and low energy electron diffraction (LEED) we were able to get a detailed insight into benzene adsorption and decomposition on Ir(111), and about the effect of oxygen thereon. The adsorption of H 2, CO and O 2 on Ir(111) has been studied previously. Several authors [67 69] found, CO desorption between 400 and 600 K, and a CO saturation coverage at room temperature equal to 7 / 12 ML. This value was used to normalize the C 1s signal. Oxygen adsorption on Ir(111) has been studied by Zhdan et al., Cornish et al. and Chan et al. [70 72]. These authors reported an O ad saturation coverage (after dosing O 2 at 300 K) of 0.5 ML, and a c(2 2) LEED pattern was found as well for the O ad covered surface. Desorption of O 2 (via O ad combination) occurs in a broad temperature region, between 700 and 1200 K. During our experiments involving an O ad covered surface O 2 was dosed at 300 K, and the area of the O ad peak was used to normalize the O 1s signal intensities. Hagedorn et al. [73] reported H 2 desorption from the clean Ir(111) surface, between 150 and 350 K. We also performed thermal desorption experiments, using H 2, CO and O 2. The desorption temperatures that were observed corresponded very well to the values reported in literature. We have studied H 2 O desorption (after dosing H 2 O at 100 K) from the Ir(111) surface as well. H 2 O desorption was observed around 175 K, and a small ( 10%) H 2 O desorption peak was observed around 220 K as well. 4.2 Molecular adsorption We used both temperature programmed desorption and high-resolution, fast XPS to study the benzene adsorption, desorption and decomposition in detail. Table 4.2 lists all the BE components that were observed during our XPS measurements, both in the absence and in the presence of O ad. Figure 4.1 shows three different C 1s spectra, measured at different stages during a benzene uptake at 170 K. During uptake we initially observed two different BE components, at ev (I) and ev (II), respectively. The concentration of the ev (II) species on the benzene saturated surface is 10% of the concentration of the ev (I) species. The peak positions shift ( 0.1 ev) to slightly higher values with increasing exposure. For higher exposures (>2.5 L) a new species (III) appears, at ev. This species is assigned to benzene molecules adsorbed on top of the C 6 H 6 -saturated Ir surface (i.e. the beginning of multilayer formation), a process that only takes place at low temperature (<170 K). The temperature at which the uptake was performed is close to the desorption temperature of the multilayer, and this species is, therefore, only observed in the presence of benzene in the gas phase (i.e. during dosing). During heating of this layer in vacuum this species is not observed anymore, indicating that it desorbed when the benzene was pumped out of the vacuum chamber

51 Photoemission intensity 4.2. MOLECULAR ADSORPTION 43 Table 4.1: Binding Energies and TP-desorption maxima found during experiments using benzene (and oxygen). The BE s for CH ad and the carboxylate species were determined by decomposition of acetone, acetic acid and ethanol (which results in CH ad and carboxylate species). species C 1s BE (ev) O 1s BE (ev) des./dec. T (K) benzene (I) benzene (II) multilayer benzene (III) benzene on sat. ox benzene (g) (Ref. [74]) - - CH ad C ad C x H y CO ad carboxylate O ad >700 Coverage (ML C ) benzene (I) benzene (II) multilayer Exposure (L) Binding energy (ev) 8 L 2.3 L 0.25 L Figure 4.1: Three C 1s spectra obtained during benzene uptake on Ir(111), at a surface temperature of 170 K. The inset shows the evaluation of the uptake experiments, using three different fitting components.

52 44 CHAPTER 4. BENZENE CHEMISTRY ON IR(111) after the dosing was stopped. The C 1s signal obtained for a saturated CO layer ( 7 / 12 ML C ) was used to normalize the signal of the different C 1s components. This normalization procedure results in a C coverage of 0.9 ML on the benzene saturated surface, i.e. the coverage of benzene molecules is 0.15 ML. About 90% is benzene(i) and 10% is benzene(ii). We suggest that the two species represent benzene adsorbed on different sites, or with a different adsorption geometry. We cannot exclude that defects on the surface are responsible for the benzene(ii) species, although a defect density of 10% is not very likely. For molecular adsorption on Pt(111) it was found that the benzene molecule lies flat on the surface [75 77]. Two different adsorption sites were observed, i.e. occupation of a threefold hollow site and a bridge site, respectively. Those two different adsorption sites have significantly different desorption barriers [78], i.e. 129 kj mol 1 for desorption from bridge sites and 91 kj mol 1 for desorption from threefold sites. Morin et al. [79] showed, that the energy difference between adsorption on threefold hollow and bridge adsorption on Rh(111) (same group as Ir in the periodic table) is not very large. For the molecular desorption from Ir(111) (see Fig. 4.3) we observed two peaks, one (very small) peak around 410 K, and another (much larger) peak around 370 K. Application of the Redhead equation [7] gives desorption barriers of 117±8 kj mol 1 and 106± 7 kj mol 1, respectively. After a detailed analysis of the C 1s intensities of the two types of benzene (I and II, shown in Fig. 4.3) we tentatively conclude, that the large desorption peak (around 370 K) is indeed due to desorption of benzene (I), whereas the high temperature shoulder (around 410 K) originates from desorption of benzene (II). These values found for the desorption barrier are closest to those observed on other metal surfaces for desorption from bridge sites, and we, therefore, tentatively conclude that molecular benzene (I) adsorbs mainly on bridge sites. The small amount of benzene (II) is assigned to benzene adsorbed on a different site. Adsorbate structure LEED was used to study benzene adsorption, below and around room temperature. Nieuwenhuys et al. [63] and Mack et al. [64] already reported that benzene forms a (poorly ordered) overlayer upon adsorption at room temperature, showing a LEED pattern that is somewhat reminiscent of a (3 3) pattern. Hamm et al. [80] found a well ordered benzene overlayer on Pd(111) upon adsorption between 275 and 325 K, with a ( 19 19)R±23.4 o superstructure, while Van Hove et al. [81, 82] found an ordered benzene/co mixed overlayer on Rh(111), with a c(2 3 4)rect unit cell. Neuber et al. [83] reported that benzene also forms a ( 19 19)R±23.4 o overlayer in the absence of CO ad, while a transition to the c(2 3 4)rect and (3 3) structures was found when the adsorbate layer was exposed to CO. These results stress that even a small CO ad contamination can have a large influence on the ordering of benzene on an fcc(111) surface. Gland et al. [84] also reported a c(2 3 4)rect and a c(2 3 5)rect benzene superstructure on Pt(111), but the

53 4.2. MOLECULAR ADSORPTION 45 Figure 4.2: LEED pattern observed after benzene adsorption at 335 K (sum of five pictures). The structural model that we propose is shown on the righthand side. presence of CO in these ordered structures cannot be excluded. By careful optimization of the benzene uptake temperature we were able to observe a much sharper LEED pattern than those reported before for benzene adsorbed on Ir(111) [63,64]. The picture shown in figure 4.2 was obtained using an electron energy of 61 ev, and a benzene uptake temperature of 335 K. The temperature that was used was as close as possible to the benzene desorption temperature, to ensure a high mobility of the adsorbates, which allows them to establish the most stable adsorbate structure. The LEED patterns observed for lower uptake temperatures look essentially the same, but the spots are more diffuse. The ordering of benzene on a surface is very sensitive to the presence of small amounts of CO ad. The structure that we observed for benzene on Ir(111) is similar to structures obtained on Pt(111) [84], but CO contamination cannot be excluded in the case of Pt(111). In our study we tried to avoid CO contamination as much as possible. In our experiment the sample was cleaned by heating in O 2 (g), and flashed in vacuum to remove the surface contaminants (mainly O ad ) while the background pressure was mbar. The cooling (with the same background pressure) from 600 K to 335 K took place in 120 s. The benzene uptake was done very fast, i.e mbar benzene, 10 s, while the LEED electron gun was switched off. The picture shown in figure 4.2 was taken immediately afterwards. Thermal desorption of CO after this showed, that the maximum CO contamination is ML. We, therefore, conclude that the structure we obtained is not caused by the presence of CO, but only by benzene itself. The right panel in Figure 4.2 shows the structure that we propose, based on (i) adsorption in bridge sites (based on literature for other fcc(111) surfaces) and (ii) the symmetry of the LEED pattern and literature interpretations of almost identical patterns obtained on both Rh(111) [in the presence of CO ad ] and Pt(111). Only one of the three possible domains is shown. The structure reported here was also found for benzene co-adsorbed with CO on Rh(111) using STM [85], and it is usually referred to as a c(2 3 4)rect overlayer.