UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Surface studies of planar model HDN catalysts Leung, Yin-Ling 1998

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1998-345769.pdf [ 8.39MB ]
JSON: 831-1.0061529.json
JSON-LD: 831-1.0061529-ld.json
RDF/XML (Pretty): 831-1.0061529-rdf.xml
RDF/JSON: 831-1.0061529-rdf.json
Turtle: 831-1.0061529-turtle.txt
N-Triples: 831-1.0061529-rdf-ntriples.txt
Original Record: 831-1.0061529-source.json
Full Text

Full Text

Surface Studies of Planar Model HDN Catalysts by Yin-Ling Leung B.Sc, National Taiwan Normal University, 1988 M.Sc., The University of British Columbia, 1992 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF. GRADUATE STUDIES Department of Chemistry We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 1998 © Yin-Ling Leung» 1998 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada DE-6 (2/88) Abstract Model catalysts based on oxidized Mo, and formed on planar supports, have been studied by x-ray photoelectron spectroscopy (XPS), secondary ion mass spectrometry (SIMS), Auger electron spectroscopy (AES) and secondary electron microscopy (SEM). The reason for using planar catalysts is to reduce the effects of uncertainties that may occur, as a result of differential charging, when direct measurements are made on real catalysts. The model catalyst approach is designed to provide new information about the chemistry which takes place during the preparation stages of real catalysts. It is hoped that the observations made here may help to understand behaviors for high-area supported Mo catalysts. Alumina supported molybdena model catalysts, prepared by wet deposition of ammonium molybdate on oxide-coated planar Al, were studied as a function of calcination temperature (200, 350 and 450°C). Characterizations by XPS, SIMS and SEM indicate that the Mo disperses more uniformly on the support when the heating was done at 450°C as opposed to the lower temperatures. Observations of the Al(KLL) Auger electron peak and changes in the corresponding Auger parameter, suggest that the calcination at 450°C yields a new surface bonding. It is hypothesized that direct Mo-O-Al surface linkages are formed at this particular temperature, and that this provides the driving force for the enhanced Mo dispersion observed. Comparative sulfidabilities of these model catalysts (i.e. the uncalcined form and those calcined at 200, 350 and 450°C) were assessed by XPS. The Mo dispersion affects the sulfiding of the samples calcined at lower temperatures, and in particular the amount of Mo(+4) sulfide formed through the treatment with H2S is greater when the Mo is better dispersed on the initial sample. The sulfiding of the 450°C-calcined sample seems particularly influenced by the Mo-support interaction, and the Auger parameter changes in a way that suggests the rupture of its 11 Mo-O-Al linkages. High-area catalysts were treated as similarly as possible to the model catalysts for some initial tests on the hydrodenitrogenation (HDN) reaction for pyridine. The sulfided 450°C-calcined sample provides better HDN activity than the other sulfided samples. XPS study of the sulfided 450°C-calcined model sample indicated the presence of a non-stoichiometric sulfide (MoS2-x) which may be a factor in the reactivity. Reaction pathways associated with the nitridation by NH3 of M0O3 thin films, formed on a Mo substrate, were studied by XPS. Core level and valence spectra are consistent with the M0O3 being reduced, but the degree of reduction depends on the reaction temperature. Heating to 350°C indicates some conversion to Mo(+5.) and "O-rich" Mo(+4) components, while heating to 450°C and then to 700°C give respectively a "N-rich" Mo(+4) form and a Mo(+3) oxynitride as the dominant components. Comparisons are made with treating the original M0O3/M0 sample by cold plasmas formed by H2 and by N 2 . The whole evidence is consistent with the nitridation of M0O3/M0 by NH3 involving initial hydrogenation, with the subsequent elimination of water, reduction and the effective replacement of O by N. The nitridation of M0O3 samples formed on alumina and silica supports have also been characterized by XPS, and comparisons made with behavior for the NH 3 reaction with M0O3/M0. The samples on the oxide supports appear to show easier O-N replacement compared with the M0O3/M0 system. In general, the reduction behavior for MoGVAlOx is similar to that of M0O3/M0, but the metal in MoCVSiOa is more easily reduced (Mo(0) is detected after the reaction with NH3 on the S i02 system but not on A10x). Comparisons of heating rates for the nitridation step from 350 to 450°C were made for the M0O3/M0 and MoGVAlOx samples. Differences between the higher heating rate (100 K/h) and lower heating rate (40 K/h) are incremental but definite. In this work, the lower heating rate appears to help both the O-N iii replacement and the metal reduction. These observations contrast with conclusions reached previously from x-ray diffraction for the evolution of bulk phases during the nitridation process the different heating rates. iv Table of Contents Abstract 1 1 Table of Contents v List of Tables X 1 List of Figures xiii Acknowledgements xix 1. Introduction 1 1.1 Introduction 1 1.2 HDN processes 4 1.2.1 Nitrogen compounds 4 1.2.2 HDN catalysts 7 1.3 Surface characterization of catalysts 8 1.3.1 Sample charging of commercial catalysts 10 1.3.2 Model catalysts : 11 1.4 Preparation of HDN catalysts - literature review 12 1.4.1 Application of Mo 12 1.4.2 Drying process 13 1.4.3 Calcination process 14 1.4.4 Sulfidation process 17 v 1.4.5 Nitridation process 20 1.5 Objectives for this research 21 2. X-ray Photoelectron Spectroscopy 23 2.1 Introduction 23 2.2 XPS spectral features 25 2.3 Quantitative analysis 29 2.3.1 Surface sensitivity and sampling depth 29 2.3.2 Atomic concentration determination 31 2.3.3 Angular dependent measurements 33 2.4 Instrumentation 35 2.4.1 Ultrahigh vacuum (UHV) 35 2.4.2 Sampling handling 37 2.4.3 X-ray sources 40 2.4.4 Energy analyzer 42 2.4.5 Data processing 47 2.5 Sample charging 51 2.5.1 Energy referencing 51 2.5.2 Differential Charging 54 3. Secondary Ion Mass Spectrometry 55 3.1 Introduction 55 3.2 SIMS analyses 57 3.2.1 Static SIMS 57 vi 3.2.2 Dynamic SIMS 58 3.2.3 Imaging SIMS 60 3.3 Matrix effects 60 3.4 Instrumentation 62 3.4.1 Ion sources : 65 Electron impact ion source 65 Gallium liquid metal ion source 65 Duoplasmatron ion source 68 3.4.2 Detectors 68 Quadrupole mass filter 68 Scintillator-photomultiplier 69 4. Study of Calcinated Molybdenum-Aluminum Oxide Model Catalysts 73 4.1 Introduction 73 4.2 Experimental 74 4.3 Results 75 4.3.1 Planar and real catalysts 75 4.3.2 Dispersion of Mo at varied calcination temperatures 77 4.3.3 Interfacial analysis of Mo03/A10x 77 4.4 Discussion 83 5. Sulfidation of Thin Films of M0O3 on a A10x Planar Support 85 5.1 Introduction 85 5.2 Experimental 86 vii 5.2.1 Sulfidation of planar catalysts 86 5.2.2 Catalytic activity studies 86 5.2.3 XPS measurements 88 5.3 Results 90 5.3.1 Sulfidation of MoCVAlOx samples 90 5.3.2 Catalytic activity studies 100 5.4 Discussion 100 6. Study of the Reduction of Mo0 3 Thin Films by NH 3 105 6.1 Introduction 105 6.2 Experimental 106 6.3 Results 108 6.3.1 Plasma treated M0O3 thin film 108 6.3.2 Reaction of M0O3/M0 with NH 3 112 Nls spectra .. 112 Ols spectra 115 N/Mo and O/Mo ratios 117 Mo3d spectra 119 6.3.3 Valence band spectra 122 6.4 Discussion 124 6.5 Concluding remarks 126 7. Studies of the Nitridation of M0O3 Thin-Films on Alumina and Silica Supports 131 7.1 Introduction 131 viii 7.2 Experimental 132 7.3 Results 132 7.3.1 Nitridation of Mo03/A10x 133 Mo-A10x interaction 133 Nitrided Mo03/A10x compared with nitrided M0O3/M0 134 7.3.2 Nitridation of Mo0 3/Si0 2 137 Mo-silica interaction 137 Nitrided Mo0 3/Si0 2 compared with Mo03/Mo 141 7.3.3 Effect of heating rate 144 Mo03/Mo 144 Mo03/A10x 146 7.4 Discussion 147 8. Concluding Remarks 150 8.1 Summary for new results 150 8.2 Future directions 154 8.2.1 Quantification of Mo loadings 154 8.2.2 Further studies on sulfided Mo03/A10x 155 Studies on MoS2.x, S22" and Sn2" 155 Analysis of Ols peak 156 8.2.3 Further studies on Mo nitride catalysts 156 Promoter effects 156 Further studies on nitrided Mo03/A10x and Mo0 3/Si0 2 158 ix Passivation effects 158 HDN activity studies 158 References 160 x List of Tables Table 1.1 Typical nitrogen compounds in light petroleum feedstocks 6 Table 1.2 Some surface and bulk analytical techniques for catalyst characterization 9 Table 4.1 Surface science characterizations for different samples studied in this work 78 Table 5.1 Reported Mo3ds/2 and S2p binding energies relevant to reduced and sulfided Mo0 3 91 Table 5.2 Characterizations for model catalysts before and after sulfidation at 350°C 93 Table 5.3 Binding energies used as starting points for XPS curve fitting analyses for MoGyAlOx samples after sulfidation 95 Table 5.4 Binding energies and relative proportions of different components in Mo3d and S2p spectra from MoCVAlOx samples after sulfidation. These samples have been calcined at different temperatures 96 Table 5.5 Products obtained from sulfided high-area catalysts during the HDN of pyridine.... 101 Table 6.1 Binding energies and relative proportions for the different components identified in the NIs, Ols and Mo3d spectra measured from the M0O3 thin film after initial preparation and after subjecting to various stages of treatment in NH 3 114 xi Table 6.2 Atomic ratios indicated by XPS for the initial M0O3 film and after the different stages of heating in NH 3 118 Table 6.3 Five main Mo components emphasized for the Mo0 3/NH 3 reaction 127 Table 7.1 Atomic ratios and binding energies measured for Mo03/A10x prior to and after nitridation. That formed after nitridation is compared also with the M0O3/M0 system where both have been heated in NH 3 with the high heating rate procedure. 138 Table 7.2 Atomic ratios and binding energies measured for Mo03/SiC<2 prior to and after nitridation. That formed after nitridation is compared also with the M0CVM0 system where both have been heated in NH 3 with the high heating rate procedure. 143 Table 7.3 Atomic ratios and binding energies measured for Mo03/Mo and Mo03/A10x systems after nitriding in NH 3 for different heating rates (high or low) and different final temperatures (450 or 550°C) 145 X l l List of Figures Figure 1.1 A scheme of conversion processes in the oil refining industry 2 Figure 1.2 Simplified diagram of a hydrotreatment process 5 Figure 2.1 Schematics for (a) the photoelectric effect: the emission of photoelectron and the formation of a atomic core hole; the de-excitation of the ion by (b) x-ray fluorescence and (c) Auger emission. Note that the ion is doubly charged after the Auger de-excitation 24 Figure 2.2 (a) A low-resolution XPS spectrum of Ag, and (b) a high-resolution Ag3d spectrum, excited by MgKct 26 Figure 2.3 XPS spectra from MoCVSiOa: (a) low-resolution and (b) Si2p at high-resolution... 28 Figure 2.4 Inelastic mean free path of electrons as a function of kinetic energy inside a solid.. 30 Figure 2.5 Notation for photoelectron intensity from semi-infinite homogeneous sample. ...32 Figure 2.6 Angle dependent measurements: (a) definition of take-off angle 0, the angle between the sample surface and the collection direction; (b) measurements at 0=30° and 90° from an oxidized Si sample. In (b) the substrate layer of Si0 2 are emphasized relative to the elemental form at 0=30° 34 Figure 2.7 A Schematic diagram of the MAX 200 system viewed from top 36 xiii Figure 2.8 The pumping system for the MAX 200 system. 38 Figure 2.9 (a) Transfer devices in the MAX 200 facility, (b) Shows the five degrees of movement possible for sample on the manipulator; there are three linear motions (X, Y, Z) and two rotations (T and R) 39 Figure 2.10 Schematic diagrams of (a) the Mg/Al dual anode x-ray source and (b) the filaments in the x-ray source, (c) Energy distribution of Al x-rays, showing the Koc and KP lines and the high-energy bremsstrahlung radiation 41 Figure 2.11 Schematic diagram for the concentric hemispherical analyzer and input lens system in the MAX 200 system 43 Figure 2.12 Schematic of the relevant energy levels for binding energy measurements 46 Figure 2.13 Shirley non-linear background subtraction applied to a spectrum with overlapping Mo3p and Nls peaks 48 Figure 2.14 Three Mo species identified from a Mo3d spectrum measured from an oxidized metal sample. Each species has two (3d3/2, 3ds/2) spin orbit coupling components with a splitting of magnitude 3.2 eV ; 50 Figure 2.15 Figure 2.15 Mo3d spectra from M0O3 thin films after nitriding at (a) 450°C and (b) 750°C. The 3ds/2 binding energies and proportions for different components from each individual measurement are listed in the tables. In each case independent preparations and measurements are made (trial 1 and 2) 52 xiv Figure 3.1 Secondary ion mass spectrometry: (a) schematic diagram of the major components of a SIMS spectrometer; (b) illustration of phenomena in the SIMS process 56 Figure 3.2 Static SIMS spectra: (a) positive ion and (b) negative ion spectra from an oxidized Al support; (c) positive ion and (d) negative ion spectra from a MoCVAlOx sample 59 Figure 3.3 Ion-induced SEM and SIMS images from a MoCVAlOx sample which had been calcined at 200°C 61 Figure 3.4 Schematic diagram showing the main components of the VG Scientific SIMS spectrometer 63 Figure 3.5 Pumping system for the VG SIMS spectrometer 64 Figure 3.6 Schematic diagram for the AG61 electron impact ion source showing: (a) the ion source assembly, and (b) the ion optical column 66 Figure 3.7 Schematic diagram for the MIG100 liquid Ga ion gun in (a); (b) indicates the beam focusing unit 67 Figure 3.8 Schematic diagram for the MM12-12S SIMS quadrupole mass analyzer: (a) major components of analyzer; (b) the mass filter 70 Figure 3.9 Schematic diagram for the scintillator-photomultiplier detector: the generation of a SEM image for a sample by scanning the ion beam across the sample surface..72 Figure 4.1 Mo3d spectra from two samples after drying at 100°C: (a) 13 wt. % M0O3 on high surface area alumina, and (b) Mo03/A10x model planar catalyst 76 XV Figure 4.2 Ion-induced SEM and SIMS images from three Mo03/A10x planar catalysts calcined at 350 and 450°C 79 Figure 4.3 Positive ion SSIMS spectra measured from four Mo03/A10x planar samples: one is uncalcined, the other three have been calcined at 200, 350 and 450°C 80 Figure 4.4 (a) A12p spectra and (b) Al(KLL) spectra from the four Mo03/A10x planar samples considered in Figure 4.3 82 Figure 5.1 Reactor used showing the gas manifold for sulfidation and HDN reaction, the cold trap for HDN reaction products and the furnace 87 Figure 5.2 (a) S2p and (b) Mo3d spectra measured from a natural M0S2 (molybdenite) single crystal 89 Figure 5.3 Mo3d spectra from sulfided (a) uncalcined catalyst, (b) 200°C calcined catalyst, (c) 350°C calcined catalyst, and (d) 450°C calcined catalyst 97 Figure 5.4 S2p spectra from sulfided (a) uncalcined catalyst, (b) 200°C calcined catalyst, (c) 350°C calcined catalyst, and (d) 450°C calcined catalyst 98 Figure 6.1 Schematic diagram for the plasma treatment chamber 107 Figure 6.2 Mo3d spectra from (a) the initially-formed M0O3 thin film; (b) the film formed after treating that in (a) with the cold H plasma for 10 min; (c) the film formed after treating that in (b) with the cold N plasma for 10 min; and (d) the film formed after heating that in (b) at 220°C for 10 min 109 xvi Figure 6.3 01s spectra from: (a) the initially-formed M0O3 thin film; (b) the film formed after treating that in (a) with the cold H plasma for 10 min; (c) the film formed after treating that in (b) with the cold N plasma for 10 min; and (d) the film formed after heating that in (b) at 220°C for 10 min 111 Figure 6.4 The overlapping Mo3p3/2 and Nls from the M0O3 fdm after heating in NH 3: (a) from 25 to 350°C; (b) from 350 to 450°C; (c) from 450 to 700°C; and (d) at 700°C for lh. The Nls and Mo3p components are shown by dotted lines and continuous lines respectively...." , 113 Figure 6.5 Ols spectra from (a) the initially-formed Mo0 3 thin film; (b) the film from (a) after heating in NH 3 from 25 to 350°C; (c) the film from (b) after heating in NH 3 from 350 to 450°C; (d) film from (c) after heating from 450 to 700°C; and (e) the film from (d) after heating at 700°C for lh 116 Figure 6.6 Mo3d spectra from (a) the initially-formed M0O3 thin film; (b) the film from (a) after heating in NH 3 from 25 to 350°C; (c) the film from (b) after heating in NH 3 from 350 to 450°C; (d) film from (c) after heating from 450 to 700°C; and (e) the film from (d) after heating at 700°C for lh. 120 Figure 6.7 Low-resolution valence level spectra measured from (a) the initial M0O3 thin film, followed by exposures to the H plasma and the N plasma; (b) the nitrided film after three stages of heating in NH 3 123 Figure 6.8 Schematic indications of: (a) the effects of adding H to Mo0 3 followed by removal of either H2O or H; and (b) the incorporation of N species into the film xvu structure with local vacancies. The Mo=0 may correspond to two O atoms ridging to the same indicated Mo atom 125 Figure 6.9 Relative proportions of the different Mo components after the initial M0O3 film is treated in N H 3 to the different heating stages 128 Figure 7.1 Comparison of Mo3d spectra from MoCVAlOx samples in the uncalcined form continuous line) and after calcination at 450°C (dashed line) 135 Figure 7.2 Comparison of (a) Mo3d spectra and (b) overlapping Mo3p3/2 and N l s spectra from M o 0 3 / A 1 0 x and M o 0 3 / M o samples which have been nitrided at 550°C, and from Mo0 3 /Si02 which has been nitrided at 700°C. The N l s and Mo3p 3/ 2 components for the spectra under (b) are identified by dashed lines and continuous lines respectively 136 Figure 7.3 Mo3d spectra from MoCVSiCk samples: (a) uncalcined, (b) calcined at 450°C, (c) and (d) same sample as in (b) but the spectrum was collected at an exit angle of 30° 140 Figure 7.4 Spectra from the Si0 2 /S i sample after nitriding at 700°C: (a) Si2p including SiC>2, elemental Si and the new formed Si-N species; (b) N l s 142 xvii i Acknowledgments I would like to thank my supervisor, Professor K.A.R. Mitchell, who introduced me to surface science. I am grateful for his sponsoring this work as well as his advice and guidance throughout the period of the research. I thank Professor K.J. Smith for providing guidance in catalysis research and for his comments on this thesis. I am especially grateful to Dr. P.C. Wong, Dr. B.J. Flinn and Professor M.Y. Zhou for helping with the XPS and SIMS measurements, and for having many helpful discussions with me. I have greatly appreciated interactions with other members of the group, including Ms W.F. Heung, Dr. W. Liu, Dr. Y .M Wang, Dr. Y.S. Li , Dr. K.C. Wong, Mr. M. Saidy, Ms. J. Fang, Dr. J.F. Ying, Ms. M. Kono, Ms. D. Susac, and Ms. L. Shi. I also acknowledge the efforts of staff in the Departmental Mechanical and Electronic shops for making equipment and keeping it all in working order for this research. Finally, I would like to thank my family and my friends for their support and encouragement during these years of my studies. xix 1. Introduction 1.1 Introduction Oil refining consists of a complex scheme of physical and chemical conversion processes. Figure 1.1 presents a simplified system of a typical oil refinery [1]. The crude oil is first separated into different portions by fractional distillation. Some distillates, such as butane (liquefied petroleum gas, LPG) and gasoline, can be used directly. Some heavier fractions, on the other hand, need further refining in order to have acceptable performance. The naphtha fraction, for instance, has the right boiling point (80-160°C) for gasoline but the octane number is too low. Thus octane number upgrading is necessary and that is achieved by catalytic reforming. Nevertheless the reforming catalysts can be poisoned by S-containing and N-containing compounds [1,2]. Under this circumstance, the hydrotreatment process must be done prior to the reforming process. During the hydrotreatment, a catalytic hydrogenation is carried out to eliminate N and S as NH3 and H2S. Although this clean-up process is principally for the protection of downstream catalysts, it is also essential to avoid environmental impact problems. Since the liberation into the air of pollutants NO x and SOx, from the combustion of N- and S-containing fuel oil, is a severe hazard to the global environment, effective hydrotreating is obligatory in the oil refining industry. Compared to the other refining processes, hydrotreatment has been considered as relatively primitive. In the early period after 1950's [1], it was discovered that Co, Ni, Mo and W sulfides and their mixtures were active and relatively cheap hydrotreating catalysts. Catalysts such as Co/Mo/AI2O3 and Ni/Mo/Al203 have been widely used in refining for many years [1,3]. Due to the comparatively high S content in conventional petroleum feedstocks (the quantity of N-1 c o t/3 Q c p o 03 Ul liquefied petroleum gas (LPG) and gas straight run gasoline naphtha^ | ^ I I I t I Y D R O | T R E A T I N G | middle distilates CATALYTIC REFORMINC heavy atmos gas oil vacuum gas oil lube base stocks CATALYTIC CRACKING HYDRO TREATING gasoline HYDRO CRACKING gasoline, naphtha, middle distillates I gasoline, naphtha, middle distillates fuel oil I asphalt ' gasoline, naphtha, middle distillates refinery fuel gas > LPG -> regular gasoline ~^ premium gasoline solvents -> aviation fuel -> diesel heating oil lube oil • greases asphalt ~^ industrial fuels > coke Figure 1.1 A scheme of conversion processes in the oil refining industry (modified after J.A. Moulijn et al. [1]). The major refining processes are catalytic reforming (to increase the octane number), catalytic cracking (to reduce the molecular size), and hydrotreatment (to remove undesirable constituents, like S and N). 2 containing compounds (0.1-0.2% [4]) is 5 to 20 times less than that of S-containing compounds [5]), hydrotreating catalysts were basically developed and optimized for hydrodesulfurization (HDS) of sulfur compounds. However, the need to develop more effective HDN catalysts has drawn a good deal of attention in recent years. This is partly set by the consumption of high-quality oil supplies, and the consequent growing need to use alternative sources of lower quality. These latter feedstocks are rich in refractory nitrogen compounds, an average value being 1 to 2% by weight but up to 3.9% [7] in heavy gas oil .from the Athabasca tar sands. The quality control of liquid and lube base stocks will not reduce as pressure increases for more stringent environmental legislation about polluting emissions from automobiles. Even though it was recognized that hydrodenitrogenation (HDN) of N-containing compounds is more difficult than HDS, in most refineries, HDN processes have generally been carried out using transplanted HDS catalyst technology. Owing to the different chemical natures, inevitably HDS catalysts are not ideal for N removal [7]. Severe reaction conditions are needed to achieve acceptable levels of HDN, but since these catalysts lack selectivity for HDN reactions, large amounts of hydrogen are generally needed [4]. Also, with current technology, reaction conditions for HDN can result in over cracking of the gas oil and hence lead to a loss of valuable hydrocarbon products. In any event, from both environmental and economical points of view, it is necessary to develop catalysts which can effectively remove N; or both N and S, and provide better selectivity, low hydrogen consumption and high yields of distillates. Much effort is now aiming to understand the fundamental basis for the activity and selectivity of the hydrotreating catalysts. These researches are leading to new results that hopefully will facilitate the industrial development of improved processes. However many problems have not yet been solved, and they depend on a need for improved catalyst 3 characterization and identification of catalytic sites, as well as the development of knowledge for the main reaction mechanisms. 1.2 HDN processes In current industry, HDN processes are not separated from the other hydrotreatments, which are typically carried out at around 350 to 420°C at pressures from 40 to 100 bar. The severity of the processing conditions depends on the quality of the feedstock [4]. Figure 1.2 shows a simplified scheme for a hydrotreatment reactor [4]. Briefly, the feedstock is mixed with H2, heated, and then introduced to the reactor packed with Co/Mo/AI2O3 or Ni/Mo/A^Os catalysts. The resulting gases and liquids are separated in a high pressure separator, in which H2, H2S, NH 3 and light hydrocarbons are recovered or recycled while hydrotreated liquids are distilled into different fractions. 1.2.1 Nitrogen compounds Both heterocyclic and non-heterocyclic N compounds are found in petroleum feedstocks [8]. Since HDN of the latter compounds occurs rapidly in ordinary hydrotreatment conditions [8], the hydrotreating of the heterocyles has been of most concern to the refiners. The heterocycles are generally 5-membered pyrrolic ring structures or 6-membered pyridinic ring structures; some typical heterocyclic N compounds found in light petroleum feedstocks are listed in Table 1.1. The natures of these two ring structures are different. N in the pyridinic structures is generally basic while that in the pyrrolic structures is non-basic. Based on HDN of model compounds, quinoline and indole being typical examples [9,10], evidence indicates they have different adsorption modes on the catalyst surfaces. These studies suggest that adsorption through N is preferred for 6-membered ring compounds [9] but when the N atom is sterically 4 Hydrogen T Hydrogen recycle t Feedstock Furnace 1 A ammonia e o 1—H *•«—> Q High pressure separator naphtha gas oil -> residue Figure 1.2 Simplified diagram of a hydrotreatment process (modified after M.J. Ledoux [4]). 5 Tab le 1.1 T y p i c a l ni t rogen compounds in l ight petro leum feedstocks. Six-membered Five-membered Pyrrole Indole Indoline Carbazole 6 hindered, adsorption through the aromatic ring can occur [10]. The adsorption of five membered ring structures, on the other hand, appears to favor the side-on interactions through the n electrons of the aromatic ring [9]. The activities of N-containing compounds are believed to be It is generally accepted that HDN reactions proceed via (a) hydrogenation of the aromatic ring structure followed by (b) hydrogenolysis of the strong C-N bond. Reaction (1) gives an example for pyridine HDN under typical industrial conditions: Two different reaction sites are believed to be present on the catalysts (i.e., hydrogenation site and hydrogenolysis site) [14,15]. The catalysts require a balance between the relative activities of these two processes, and this balance will depend on the type of compounds reacting. A large number of intermediate species, with different reactivity and adsorptivity, are believed to be involved even for simple model compounds, but understanding is still very limited. 1.2.2 HDN catalysts Industrial HDN catalysts are mixed metal sulfides on a support, such as alumina (AI2O3). Examples are supported mixed phases of M0S2 and CoS, M0S2 and NiS, and WS2 and NiS, where the first compound in each pair is the major component. The minority component of Ni or Co, corresponds to about one quarter of the total metal [4] present, and these Ni and Co components are considered as promoters. Bhinde et al. [16] found that all catalysts containing Ni are better for HDN than Co catalysts, because they are more effective for hydrogenation, while Ni-Mo is superior to Ni-W (Co catalysts on the other hand are better for HDS). Comparatively, strongly related to their electronic properties [10-13]. C 5 H 1 2 + NH 3 R n . ( l ) H 7 Co-Mo catalysts have been more extensively studied than Ni-Mo catalysts. The generally accepted description for sulfided Co-Mo catalysts is that small M0S2 crystallites lie with their basal planes parallel to the AI2O3 surface, or are edge-bonded to the support surface, and Co ions, as the promoters, are adsorbed on the edges of the M0S2 crystallites [17]. A similar hypothesis has been applied to the sulfided Ni-Mo catalyst systems [8]. As stated earlier, new and more effective HDN catalysts are urgently needed from both the environmental and economical points of view. In this context, Chinaelli et al. [18] have recently discussed the need to develop a new generation of catalysts for the removal of S and N impurities from low quality feedstocks. A potentially important development has been provided by recent demonstrations that Mo nitrides can act as HDN [19-21] and HDS [22,23] catalysts. It has been reported that these catalysts are more active and selective for HDN compared to the sulfided counterpart. However current research in this area is just in the preliminary stage. 1.3 Surface characterization of catalysts All reactions which occur on a catalyst surface consist of adsorption steps, surface reactions and desorption of products. To understand the entire catalytic mechanism, it is necessary to have molecular-level understanding of the adsorption/desorption reactions, the structure of catalytic surfaces, and the bonding and coordination of reaction intermediates. The elemental analysis for a catalyst material is not the only work for scientists. It is valuable to know the exact structure of the catalyst, including defects and the location of promoters, and ultimately there is the challenge to learn what happens to the surface atom-by-atom during a reaction. The first requirement to understand catalysis is to have detailed knowledge of surface composition. Many surface analytical techniques are available, and a partial list is presented in Table 1.2. Among these tools, XPS is the most widely used for identification of surface 8 Table 1.2 Some surface and bulk analytical techniques for catalyst characterization. Technique Acronym Physical mcasurcniL'ni Information cained Refs. Surface Analysis Auger electron spectroscopy AES kinetic energies of electrons emitted by two-electron de-excitation of core hole elemental composition [24] Infrared spectroscopy IRS absorption of IR radiation as function of energy perpendicular vibration modes [25] Ion scattering spectroscopy ISS measure directions and energies of scattered ions topmost layer and elemental composition 126] Secondary ion mass spectrometry SIMS ions sputtered from surface measured in mass spectrometer elemental composition 127] Thermal desorption spectroscopy TDS desorbed species as a function of sample temperature desorption energy, multiple bonding states [1] X-ray photoelectron spectroscopy XPS photoemission of core electrons elemental composition and chemical states [24] Bulk analysis Extended x-ray absorption fine structure spectroscopy EXAFS photoelectron interference effects manifested in x-ray absorption fine structure local structure information on bond lengths and coordination numbers 128] Laser Raman spectroscopy LRS change in frequency of laser radiation after scattering by material vibrational modes [29] Nuclear magnetic resonance spectroscopy NMR absorption of radio-frequency radiation due to excitation of nuclear spin states bonding arrangements [30] X-ray diffraction XRD intensity distributions in diffraction x-ray bulk composition for polycrystals, geometrical arrangements of atoms in single crystals [31] composition. But bulk composition analysis also plays an important role in catalysis studies, and therefore some commonly used bulk analytical techniques are additionally given in Table 1.2. 1.3.1 Sample charging of commercial catalysts A major problem, which can interfere with measurements of surface composition of commercial catalysts, is the presence of sample charging resulting from the photoemission process. With conducting samples grounded through the spectrometer during the measurement, no charge accumulates, but for non-conductive samples, like supported catalysts, charges can build up because of the incomplete electron neutralization [32] (see Section 2.5). In principle, the presence of sample charging in XPS may give an apparent constant shift in binding energy for the photoemission peaks (constant charging is normally relatively easy to correct for) but with variations in, for example, particle size [33] and sample homogeneity [34,35] differential charging may happen. Wagner and Windawi [33] have described catalyst systems where difference with dopants can yield different particle sizes, which in turn experience different degrees of charging on the supports. The inhomogeneity of the particle size therefore produces mixed charging with broadening in the photoemission lines. In some cases, these charging effects may be mistaken for the appearance of new chemical species. In general, the constant charging shifts can be corrected by energy referencing (Section 2.5.1) but the differential charging is difficult to handle. Although the use of the Auger parameter [36] (Section 2.5.2) can be very helpful for catalyst analysis [37], the low intensity of some Auger lines observed from commercial catalysts can also limit its application. 10 1.3.2 Model catalysts For ease of handling, commercial catalysts including the support are often configured or compressed into spheres or extrudates. With this physical morphology, many active sites are located at different places, both inside or outside of the porous regions of the support. A large proportion of the active surface sites are accessible to reactants (under high pressure), but those sites within the pores are not accessible to most surface analytical tools. It is a logical speculation that a detailed analysis of these hidden sites would permit a finer control of selectivity and activity for the reaction. Therefore there has been some interest recently in studying simulated catalysts on planar supports [38-43]. These model catalysts are prepared similarly to the way commercial catalysts are prepared, except planar supports are used instead of porous supports. Real supports are usually oxides, and "planar" means a thin (<104 cm) oxide layer of uniform thickness present on a suitable substrate, such as a thermally oxidized layer on top of Al metal (A10X/A1). Such model systems, are suitable for surface analysis, and they can have uniform dopant size and distribution [44]. As well the sample charging problem can be alleviated. Several catalyst systems have been explored by adsorbing organometallic complexes [45,46] or by evaporating metals such as rhodium, platinum, or molybdenum [38-40] onto the supports. Nevertheless, in order to mimick the chemistry on the real catalyst system, it is more attractive to prepare model catalysts in the same way as the real catalysts are normally made and that is by wet chemical techniques [41-43]. In this case, each preparation step (i.e., drying, calcination and activation) can be studied systematically. 11 1.4 Preparation of HDN catalysts - literature review The preparation of a supported Mo catalyst contains the following steps. Initially the active phases (i.e., Mo and Co (or Ni)) are applied to the pre-shaped supports, and this is followed by drying, calcination and activation to make the working catalyst. In fact each preparation step is related to each other. The behavior of the catalyst at one step is conditioned by all previous steps, and in turn this influences all subsequent steps and the final active form. 1.4.1 Application of Mo In general, a commercial catalyst consists of two or more components: the active phase(s) and the support. The former is principally responsible for the catalytic activity while the latter is the carrier for the active phases. As mentioned before, the support is used for reasons of cost (effective use of metal) and ease of production and handling (pyrophoricity and storage stability). Usually the support is considered to be catalytically inactive. However it can affect the catalytic reaction if there is an interaction with the active phase(s). Y-AI2O3 and silica gel are typical high-area supports which are obtainable with surface areas in the range of 100-800 m2/g. In the case of a Mo catalyst, the content of M0O3 is usually between 0.1 to 20 wt % of the total catalyst. The Mo and any promoters are separately applied onto a support by impregnation via the solution phase of ammonium heptamolybdate ((NH4)6Mo70"24, or AHM) and Co (or Ni) nitrate. Four preparation methods are commonly used to prepare the catalysts. A. "Dry" impregnation [47,48] The required amount of AHM is dissolved in a sufficient volume of water to fill the total pore volume of the support batch; this is referred to as pore volume saturation. After drying the required amount of Co (or Ni) is applied to the Mo-support in the same way. 12 B. Equilibrium adsorption [49-51] By circulating the AHM solution through a bed of supports, AHM diffuses from the solution and is absorbed on the support until equilibrium is reached. The loading, and the nature of the ions in solution, are controlled by varying pH with addition of either nitric acid or ammonium hydroxide. For instance, at high pH (>8), the adsorption of Mo on V-AI2O3 is low and M 0 O 4 2 " (molybdate) is the dominant species, while at low pH (<6), the Mo loading is higher and Mog0264 and M07O24 6 " ions (polymolybdates) are predominantly adsorbed on the support [50]. C. Metal deposition - model catalysts [40,52] This technique has been used for planar model catalyst studies. Alumina (or silica) supports are prepared by heating a pure Al metal plate (or Si wafer) in air while Mo is deposited on the support by a sputter deposition system. Mo oxide is then produced through the oxidation of the Mo metal in air. D. "Wet" impregnation of Mo - model catalysts [43,53] This method has also been applied for studies on planar model catalysts. The support is prepared as in C, but the Mo is applied on the support by the wet chemistry methods of dipping and spin coating of AHM or Mo0 2 Cl 2 [43]. 1.4.2 Drying process For hydrotreating catalysts, the freshly impregnated Mo catalysts are usually dried in air at 100°C for 24 hours. The drying step is considered to be a thermal treatment which aims to eliminate the water [54]. In addition to the removal of water, it has believed that the drying process may help the partially dried or dehydrated salt to disperse on the support [54]. 13 1.4.3 Calcination process During the calcination process, Mo catalysts are heated at ~500°C in air for several hours. In general, calcination of the deposited active phases may lead to several transformations and solid state reactions [54]. For Mo catalyst systems, the main change involves interactions between the Mo oxide and the support. As mentioned earlier, the support is present primarily for practical reasons, but cases are known where interactions between the support and the Mo apparently influence the performance of the resulting catalyst. For example, silica is less preferred in hydrotreating catalysts because it promotes the formation of C0M0O4, which is considered as detrimental [47,55] for hydrotreating. Interactions of the types Mo-support [48,50,52], Co-support [56-58] and Mo-Co [58,59] have been active areas of study. Numerous reports have been published, and here we just focus on reviewing those studies for the M0-AI2O3 and Mo-SiC>2 interactions. XPS has been widely used as the probing technique since 1975, but the limited conductivity for the supports can interfere with the interpretation of the measured data. The XPS spectra of supported M0O3 are broad when compared to that of bulk M0O3 [60,61]. The broadening has been attributed to electron transfer [60], although others believed that the broadening results from the presence of different Mo species that could not be resolved [61]. Nevertheless Zingg et al. [62] determined that most of the broadening arises from sample charging. Miller et al. [60] were the first to report differences in the binding energies for the Mo3d doublet when measured for bulk and alumina-supported M0O3. They interpreted the latter as indicative of a strong M0-AI2O3 interaction caused by partial electron transfer from the Mo-O bond to A l [60], or alternatively some Al atoms in the alumina lattice may be substituted by Mo atoms [63]. Cimino and De Angelis [59] interpreted the shifting to higher binding energy 14 observed for the M0O3/AI2O3 system as due to sample charging rather than electron transfer from Mo to the Al oxide support. Although the chemical shift values vary from article to article [64-68], it is generally accepted that M 0 O 3 strongly interacts with AI2O3 through the dehydroxylation of the alumina surface. The question of whether the Mo-alumina interaction (Mo-O-Al) can be detected by XPS depends on how well the sample charging problem can be handled. Infrared spectroscopy (IRS) is another technique that has been used to investigate interactions between Mo and alumina [69-71]. Observations of the decrease in the absorption due to surface O-H stretching frequency as the Mo content increased appears to result from the interactions between molybdate species and the alumina surface [50,69,70]. Okamoto and Imanaka [69] reported, for Mo concentrations less than 4 wt %, that the M0O4 2 " species react with surface OH groups and these attached Mo species simultaneously consumed the adjacent OH groups in a tetrahedral configurations represented schematically as O O % // H H M 0 O 3 H M o 1 1 - H 2 0 I I H / \ M 0 O 4 2 " + H + + O O = — . O O ^ Q ' N 0 Al Al Al Al | j | , As the Mo concentration increased, the OHVMo ratio greatly reduced with the formation of multilayer molybdate, and for concentrations beyond 12 wt %, Al2(Mo04)3 and bulk M0O3 are formed. Laser Raman spectroscopy (LRS) has- also been widely used for M0/AI2O3 calcination studies [51,62,69]. There seems to be general agreement that three distinct Mo species exist in M0/AI2O3 systems for different Mo loadings [62,69]: these can be classified as the interaction species, with direct Mo-O-Al bonds, and as the bulk A ^ M o O ^ and M0O3 compounds. At lower Mo concentrations (4-12 wt % [69], <20 wt % [62]), the reaction species are tetrahedrally 15 and octahedrally coordinated Mo, possibly resulting from the reaction of molybdic/polymolybdic acid with surface OH groups. Zingg et al. [62] reported, as the M0O3 loading increases up to 20 wt %, the Al2(Mo04)3 starts to form while at-Mo loadings greater than 30 wt %, bulk M0O3 is detected. It was also reported that high temperatures (up to 700°C) and long calcination times favor the formation of Al 2(Mo0 4) 3 [62,69]. The Mo distribution is another important topic and XPS is a useful technique for monitoring the Mo dispersion on a support [51,62,69]. Okamoto and Imanaka [69] summarized work on this topic. They observed that the Mo/Al ratio increases linearly with Mo loading up to 12 wt % [69] and that there is an excellent Mo dispersion in this range of concentration. Beyond the 12 wt % loading, the slope for the Mo/Al ratio decreases, and this suggests the onset of bulk-like and/or subsurface molybdenum species. Kasztelan et al. [51] applied Mo to the support by equilibrium adsorption, and they reported that the Mo3d/A12s intensity ratio increases after the calcination process. Zingg et al. [62] found that Mo species have high dispersion at concentrations up to 20 wt %. Thus it is generally accepted that Mo species are well dispersed up to a Mo surface concentration of at least 15 wt % to 20 wt % (~5 x 1014 Mo/cm2) [72,73]. Taking the number of OH groups on alumina surfaces to be ~1015 OH/cm2 [74], this gives an indication of the approximate "monolayer" capacity of A1203 for the incorporation of molybdate [72]. Since silica-supported catalysts have much lower activity than alumina-supported catalysts for the HDS of thiophene [75,76], less attention has been paid to systems based on the former support. The reason may be due to lower concentrations of hydroxyl groups on the silica surfaces [55], and hence less Mo-silica (Mo-O-Si) interaction, and that may encourage formation of inactive phases such as C0M0O4 [47]. Techniques such as XPS, LRS, and x-ray diffraction 16 (XRD) have been used to study silica-supported M0O3 catalysts [47,48,55,76-79]. With XPS, broad Mo3d spectra were seen [47,48], although the resolution of the Mo3d doublet improved with increase in the Mo loading [48]. Gajardo et al. [47] described the unresolved peaks as resulting from the existence of several species, while Kerkhof et al. [80] believed them to be indicative of differential charging. Based on Mo/Si intensity ratios from XPS, Okamoto et al. [48] reported that a monolayer coverage corresponds to less than 2 wt % M0O3 (that is 10 times less than for the corresponding alumina-supported system), and that multilayered Mo species and microcrystalline M0O3 developed as the loading increased to about 9 wt %. Above 13 wt %, crystalline M0O3 was detected by XRD. From observed binding energy shiftings for this system, Okamoto et al. [48] concluded that charging effects can be very important for the interpretation of XPS results in this context. They observed that the binding energy of Mo3d increases slightly with increase in Mo loading up to 5 wt %, whilst a shift to lower energy at higher Mo loading is due to differential charging. Plyuto et al. [81] suggested that a new Mo species, resulting from a strong interaction with the silica support, has a higher binding energy than bulk M0O3. Gajardo et al. [47], on the other hand, reported the binding energy of Si2p shifts to higher energy with increase of Mo loading, while other studies [82-84] indicated the presence of Mo species with lower binding energies which are attributable to the presence of Mo ions in lower oxidation states. 1.4.4 Sulfidation process The sulfidation of C0-M0-AI2O3 or Ni-Mo-Al 20 3 prior to use in a hydrotreating reaction, is normally carried out by reaction with H2S, CS 2 or organosulphur compounds in H 2 at ~350°C. Most workers agree that the resulting catalyst is almost completely converted into MoS2 during the sulfidation. Therefore the key point is to understand the role of MoS2 in the hydrotreating 17 reaction. Many studies [85-89] have been done on MoS2 single crystals to help gain insight into the catalytic activity. The basal plane of MoS2 is generally considered to be chemically inert, unless defects are induced in the surface [90,91], and so most work has been focused on ion bombarded MoS2 surfaces, for which surface analyses and catalytic reactivities [91,92] have been studied. It has been generally thought that MoS2 surfaces sputtered by noble gas ions yield the presence of Mo(+3) species which is believed to form active sites [88] for the hydrotreating reaction. The preparation of highly reactive Mo sulfide catalysts therefore appears to the question of how to create a high concentration of Mo(+3) species. Okamoto et al. [69] employed IRS to examine reactions on a M0-AI2O3 system, and reported that some hydroxyls groups are regenerated on sulfidation. That would require Mo-O-Al bonds to cleave between the support and polymolybdate species [69,70,93]. The bound tetrahedral molybdate however is less reducible and forms S-deficient MoS2 species. This species, as well as regular MoS2, shows activity for hydrotreating reactions but the former is more active, apparently because of its higher concentration of anion vacancies [94]. Spevack and Mclntyre [95] used XPS to investigate model-Mo/Al203 catalyst systems. They reported a new Mo species with Mo3d binding energy of 228.4 eV, which they assigned to a Mo(+3) species formed after the sulfidation reaction. The sulfidation mechanism is another topic of interest. Aronoldy et al. [96,97] used the temperature-programmed sulfiding (TPS) technique to compare the relative abilities of bulk M0O3 and Al203-supported M0O3 to be sulfided. They concluded that H2S readily reduces bulk M0O3 to MoS2 at 325°C. This process starts with O-S exchange, the intermediate Mo02S gives Mo0 2 and elemental S, and the Mo0 2 converts easily to MoS2. Although pure bulk Mo0 2 is generally more resistant to sulfidation, the authors believed that the Mo0 2 formed in situ in this 18 process is sufficiently porous to facilitate the sulfidation. By contrast, the sulfiding of M0O3/AI2O3 occurs at lower temperature compared with the bulk compounds ( M 0 O 3 and M0O2), and at 223°C a monolayer formation of MoS2 is reported. The authors believed the sulfiding mechanism is dominated by O-S exchange reactions; the intermediate MoOS2 is transformed to Mo(+4) oxysulfide through rupture of Mo-S bonds. In another work, Scheffer et al. [96] reported that the sulfidability is determined by the level of Mo dispersion and its interaction with alumina. They concluded there are two kinds of active species after the sulfidation [97]. The first originates from highly dispersed Mo oxide, which has a strong interaction with alumina, and the second is from species which are well dispersed on the support but are more Mo03-like. The sulfidation of the former, which is more difficult than the latter, resulted in a partially sulfided monolayer species, while sulfidation of the second Mo species results in the formation of MoS2-like microcrystals. Studies on HDS reactivity showed that the catalytic activity is mainly due to microcrystalline M0S2. Analysis with the extended x-ray absorption fine structure (EXAFS) method indicates that small or irregular M0S2 crystals are present on the support during sulfidation [98], although there is disagreement on whether the Mo monolayer is preserved after sulfiding. Some studies [99,100] have reported that a high coverage of Mo is maintained after the sulfidation, but results from electron microscopy [101,102] of non-porous model systems indicate that the morphology and orientation of the Mo, relative to the support, may change during sulfiding. On the basis of bond energy considerations, Voorhoeve [103] assumed that anions in the basal planes of M0S2 are more strongly bonded to the Mo cations than are the anions at edges or corners. Based on results from IRS, Tops0e and Tops0e [70] suggested there is a significant lateral agglomeration 19 of the Mo phase, and that the MoS2 phase formed is present mainly as a single-slab structure oriented flat on the alumina surface. 1.4.5 Nitridation process As noted above, C0-M0-AI2O3 catalysts are used basically for HDS reactions, although those of the Ni-Mo-AI2O3 type are more effective for HDN [16]. Economically and environmentally it is necessary to develop HDN catalysts with better activity and selectivity. Recently, molybdenum nitrides have drawn considerable attention because of their promising activity and selectivity for HDN reactions [19-20,104]. There are many methods to synthesize Mo nitrides and historically they have been prepared at very high temperature in the form of sintered low surface area products [105]. Since high surface area materials are needed for effective catalytic reactivity, an alternative way for nitride preparation is necessary. In the 1980s, Vople and Boudart considered this problem and established a moderate temperature (700-900°C) preparation of high surface area M02N by linearly heating M0O3 with NH3 (the temperature-programmed reaction, TPR) [106,107]. Based upon observations of the products with XRD [106], they suggested a series of topotatic transformation processes occurred [107]. Jaggers et al. [108] used observations from XRD and thermogravimetric analysis (TGA) to propose that the nitridation of M0O3 with NH3 proceeds through two parallel reaction pathways: Mo0 3 -» MoO xNi. x -> Y-Mo2N and M0O3 - » M0O2 -» y-Mo2N + 8-M0N. The first reaction pathway is considered as topotatic since both MoO x Ni. x and Y-M02N have crystallographic orientation relationships to M0O3, and this results in nitride with high surface area. In the second reaction pathway a low surface area nitride is formed, apparently because the non-topotatic M0O2 is involved. Choi et al. [21] considered how the preparation conditions 20 determine the properties of the final nitrides. They found higher Mo 2N surface areas result with higher flow rates of NH 3 and lower heating rates; Demczyk et al. [109] also examined this catalyst system and found that the near-surface composition and crystal structure are different from the those of bulk M02N. Further recent work reports that Y-AhCVsupported Mo nitride shows superior HDN activity to those of the commercial sulfided Ni-MoCVA^Cb catalysts [110-112]. However most interest so far has been on the non-supported rather than supported Mo nitrides. Colling and Thompson [112] used XRD analyses supplemented with O2 chemisorption to characterize the compositions and active sites of Y-AI2O3 supported Mo nitrides. They reported that the nitriding of the catalysts depends on the structure of the oxide catalysts and the nitriding conditions. At lower Mo loadings (e.g., 4 and 8 wt %), the resulting nitride is highly dispersed on the support while at higher Mo loading (e.g., 16 wt % Mo) y-Mo2N is formed. They believed that the most active sites are located at the perimeters of two dimensional, raft-like domains of the product. Regions near the perimeter appear to be nitrogen deficient and more active for HDN reactions, while the regular Y-M02N crystallite surfaces are associated with lower HDN activity. 1.5 Objectives for this research Studies on real catalyst systems discussed above suggest that Mo oxide interacts strongly with supports (alumina or silica) during the calcination process, and that interaction species are formed through Mo-O-Al or Mo-O-Si linkages. Although characterizations of the catalysts have been performed extensively by IRS, LRS and XRD, the existence of these surface species has not been verified. XPS has the potential to clarify this point but there are problems arising from sample charging when dealing with real catalysts. The same statement applies to studies on activation processes, such as sulfidation and nitridation. 21 The work in this thesis uses the philosophy of the model catalyst. This research aims to characterize surface and interaction species during the preparation stages of calcination, sulfidation and nitridation for catalyst systems related to those discussed above. The work uses thin, oxide planar supports (A10x and SiOi) which are formed on elemental substrates, and designed to limit any charging problems. The model catalysts are prepared by wet chemical methods such as dipping and spin coating, and although these models are not identical to real catalysts, it is hoped that observations on them can help add insight into the performance of real catalysts. In return, this further knowledge may help contribute to the development of new catalysts in the future. Nevertheless it must be realized that model catalysts are not identical to real catalysts; results or observations obtained from the modeling studies still need analysis in relation to real catalyst systems. Work described in this thesis have been carried out in four main directions: (1) to investigate Mo-alumina interactions during calcination; (2) to interpret the sulfidability of Mo/A10x catalysts which were calcined at different temperatures; (3) to study the nitriding process for non-supported M0O3; (4) to interpret the different nitriding results of A10x-supported and SiCVsupported Mo catalysts. The objectives of this thesis are therefore to provide new insight into these points. The characterization of the catalyst surfaces has been carried out with XPS, but supplementary work with secondary ion mass spectrometry (SIMS) and Auger electron spectroscopy (AES) has also been done. 22 2. X-ray Photoelectron Spectroscopy 2.1 Introduction The history of x-ray photoelectron spectroscopy (XPS) can be traced back to Hertz [113] who discovered photoelectron emission in 1887 by illuminating matter with ultraviolet light. An understanding of the phenomenon was provided by Einstein in 1905 [114]. Experiments with higher energy sources were carried out by Moseley, Robinson and Rawlinson [115] before the first World War and continued by Robinson afterwards, but the beginnings for modern developments were made in the 1950's by Siegbahn and his coworkers [116]. Commercial equipment became available in the late 1960's, and since then there have been wide ranging applications to study chemical states and chemical composition at surface regions of materials of interest both technologically and from the point of view of pure science. Many applications have been made with relevance to catalysis. Siegbahn coined the term ESCA (electron spectroscopy for chemical analysis) [117] to emphasize the fact that both photo and Auger electron emission peaks appear together in an XPS spectrum, and therefore the basic principles of both techniques are introduced here. The XPS experiment is accomplished by irradiating a sample with soft x-rays (MgKa or AlKa are commonly used). The photons interact with atoms in the material and transfer energy especially to core level electrons (e.g., the Is orbital (K level), in Figure 2.1 (a)), which are emitted as photoelectrons. To a first approximation, the kinetic energy (Ek) of a photoelectron is E k = h v - E b Eq. (2.1) 23 Photoemission • • • • • • 2p photoelectron (a) X-ray fluorescence Auger emission Auger electron -2,3 J 2 ,3 hv <<vwv V (b) (c) Figure 2.1 Schematics for (a) the photoelectric effect: the emission of photoelectron and the formation of a atomic core hole; the de-excitation of the ion by (b) x-ray fluorescence and (c) Auger emission. Note that the ion is doubly charged after the Auger de-excitation. 24 where hv is the energy of the photon and Eb is the binding energy of the electron in the sample. As different atoms have different sets of electronic binding energies, elemental identification is possible through the measurement of photoelectron kinetic energies. As illustrated in Figure 2.1 (a), the resulting ion is in an excited state after the photoemission process. The excess energy can be dissipated through two possible de-excitation mechanisms for the atomic core hole. One is x-ray fluorescence emission (Figure 2.1 (b)) and the other is Auger electron emission (Figure 2.1 (c)). These two relaxation processes compete together, and their relative probabilities vary with atomic number (Z) and binding energy associated with the core vacancy. For initial vacancies in levels with binding energy of 2 keV or less, the probability of Auger emission is dominant [118]. Figure 2.1 (c) displays the production of a KLiL2I3 Auger electron whose kinetic energy, in a first approximation, is EKLIL2,3 = E K - E L l - E L 2 3 Eq. (2.2) where E K , E L l , E L 2 3 are the binding energies for the levels involved. hi contrast to photoelectrons, Auger electrons are independent of the excitation source. Similarly each element has its own set of Auger electron kinetic energies and that allows for direct elemental identification from an observed spectrum. 2.2 XPS spectral features The low-resolution spectrum from a silver sample [119] in Figure 2.2 (a) shows a series of peaks on a step-like background. These peaks can be grouped into three types: peaks due to photoemission from core-levels and from valence-levels, as well as peaks due to x-ray induced Auger emission (Auger series). Throughout, each prominent structural peak is accompanied by a higher energy background, which corresponds to photoelectrons which have undergone inelastic scattering on their outward path and so emerge with lower kinetic energy. Additionally, 25 b CO C 0> 376 374 372 370 368 366 364 362 Binding Energy (eV) (b) Figure 2.2 (a) A low-resolution XPS spectrum of Ag, and (b) a high-resolution Ag3d spectrum, excited by MgKcc. 26 considerable background is present in the low binding energy region, as a result of photoemission by bremsstrahlung radiation (see below). The core-level peaks directly reveal electron orbital structure of silver, for example the 3s, 3p, 3d, 4s, 4p and 4d levels are seen in Figure 2.2 (a). At higher resolution, doublet structure is observed for photoelectrons from states with non-zero orbital angular momentum (e.g. p and d core levels) due to spin-orbit coupling. Figure 2.2 (b) shows a high-resolution spectrum for the Ag3d level. The doublets are characterized by j values (i.e. 1/2, 3/2 for p orbitals, 3/2, 5/2 for d orbitals, and 5/2, 7/2 for f orbitals) and the corresponding areas are determined by the respective (2j + 1) number of states. Accordingly, Figure 2.2 (b) shows the 3d3/2 and 3d5 / 2 peaks with the peak area ratio of 2:3; the energy separation between the peaks is 6 eV. Figure 2.3 (a) shows a low-resolution spectrum from a research system, in this case a thin film of M0O3 deposited on a thermally oxidized Si(100) wafer (Mo03/Si02). Now features of the spectrum involve different elements and in general peaks can be assigned by comparing with standards provided from handbooks [119]. In addition to elemental analysis, the presence of the same type of atom, but in different chemical states, can often be distinguished from interpretations of chemical shifts. That is due to the fact that the binding energies of the inner core electrons are affected by changes in the valence electron environment. An example is illustrated in Figure 2.3 (b) for a high-resolution Si2p spectrum from the same Mo0 3/Si0 2 sample. The peak at 99.4 eV is identified as Si2p from the elemental form, whereas that at 103.8 eV is from Si2p from the oxide form. A Si atom in Si0 2 has a net positive charge from electronegativity considerations, and consequently a higher binding energy, compared with elemental Si. Often component peaks from different chemical states overlap, and then a curve synthesis process is necessary to identify the different contributions from a measured spectrum; that is discussed in Section 2.4.5. 27 < Ols C(KLL) O(KVV) Mo3d Cls Mo3p Si2s l l l l l l l l l 1 l l l l I I I l I I- I I i i i II i i i i I i i 1200 1000 800 600 400 200 Binding Energy (eV) (a) I CD Si2p \ metallic Si S i 0 2 / I I I M I N I i i i i 1 i i i i 1 i 110 108 106 • 104 102 100 98 96 Binding Energy (eV) (b) Figure 2.3 X P S spectra from M o C y S i 0 2 : (a) low-resolution and (b) Si2p at high-resolution. 28 As mentioned earlier, the Auger peaks are also useful for elemental identification. In fact, a change in chemical state for an atom may give rise to changes in the Auger spectrum. For example, the Auger chemical shift between elemental Si and Si02 is 8.8 eV [24]. Moreover, when Auger spectra are differentiated, characteristic variations in peak shape may be apparent from one chemical environment to another [24]. Nevertheless, the chemical shifts and peak shape changes observed in Auger spectra are often less easy to interpret than is the case with XPS, and this follows the fact that the Auger electrons are emitted through transitions involving three energy levels. Consequently in this work, Auger spectra are just used to supplement information on chemical state analysis obtained from photoemission spectra. 2.3 Quantitative analysis 2.3.1 Surface sensitivity and sampling depth During a XPS measurement, the incident photons penetrate up to 10 urn into the solid but the photoelectrons excited from this region lose energy via inelastic scattering with the matrix. The sampling depth in a XPS measurement is determined by the inelastic mean free path (IMFP, X), which is defined as the mean distance traveled in the solid by the photoelectron before it undergoes some inelastic scattering. Empirical values of the IMFP have been provided by Seah and Dench [120] and Figure 2.4 shows the dependence of IMFP as a function of energy for different materials [121]. Since the measured kinetic energies of photoelectrons or Auger electrons are usually in the 100 to 1000 eV range, IMFPs of 6 to 20 A are typical in this energy range and this makes XPS and AES surface sensitive techniques. Conventionally, the sampling depth for XPS is taken to be 3^ from which 95% of the signal is contributed. 29 I I I 1111 1—i M I N I ~i—i—i i i 1111 50 100 Electron energy (eV) 500 1000 2000 Figure 2.4 Inelastic mean free path of electrons as a function of kinetic energy inside a solid [121]. 30 2.3.2 Atomic concentration determination The intensity of a photoemission peak in a measured spectrum, for the purpose of surface analysis, can be defined as the area of the peak left after removing the background contributions associated with inelastic scattering. The number of photoelectrons collected by the analyzer per unit time is determined by the x-ray flux (f), the photoelectron cross section (a), the number of atoms per unit volume (n), the area of the sample from which the photoelectrons are collected (A), the instrumental transmission function (T) and the inelastic mean free path (X). The intensity contribution from an incremental thickness dx in the sample (Figure 2.5) is: dl = f o n A T exp (- x/X) dx Eq. (2.3) Simple integration from x=0 to x=°o for a semi-finite homogeneous sample gives: I = f a n A T A. Eq. (2.4) For comparing relative surface compositions of two elements in a sample, a commonly used approach is to group f, A, T and rj into a sensitivity factor S, values of which are determined by the particular instrumental settings and the particular peak under study. These factors, derived relatively to the Fls peak, are available for the MAX 200 spectrometer where the transmission function is corrected for the particular instrumental conditions used for each measurement. The composition ratio for two elements in a sample, can then be expressed as: m, n 2 = [(Ii/S,)/(l2/S2)] [(X2)/ (A.,)] Eq. (2.5) In principle, the atomic ratio (nj/n2) can be determined by using tabulated values of Xi and X2 for the appropriate photoelectrons and sample [119], but for semi-quantitative work, the X2/X\ ratio is commonly taken as constant and equal to unity. The Ii/I2 ratio, from the measured peak intensities and the ratio of sensitivity factors, are used to estimate the ni/n2 ratio within the depth probed. 31 e" Figure 2.5 Notation for photoelectron intensity from semi-infinite homogeneous sample. 32 2.3.3 Angular dependent measurements As illustrated in Figure 2.6 (a), surface sensitivity can be enhanced by varying the take-off angle (0, angle between the axis of the energy analyzer and the plane of the sample surface). For normal take-off angle (0=90°), the sampling depth is 3A, while at grazing angles of emergence, the sampling depth is reduced to 3X,sin0. Figure 2.6 (b) compares Si2p spectra taken from 0=90° and 30° for an thermally oxidized Si(100) wafer. At 0=30°, the signal contributed from uppermost SiC>2 layer is more pronounced. Consider a sample with an uniform overlayer (thickness, t) on top of a substrate material. In relation to measurements associated with different take-off angle (0), Eq. (2.3) integrates after correcting for angular effects, to give: I0 = fa 0 n 0 ATX 0 ( l -e - t / X o S i n e ) Eq. (2.6) Is = f a s ns A T ?Ls e-^051"9 Eq. (2.7) for the intensities of the overlayer (I0) and the substrate (Is); X0 and Xs are inelastic mean free paths for photoelectrons originating in, and traveling in, the overlayer and substrate respectively, whereas X,so is the inelastic mean free path of photoelectrons from the substrate traveling in the overlayer. The ratio lo/ls increases strongly as 0 decreases, and Eqs. (2.6) and (2.7) are useful to interpret the relative orientation or composition gradients for different species in the sample [122]. In addition, Eq. (2.7) can be modified to compare Is values at various 0 against that at 0=90°: In [IS(0)/IS(9O0)] = (-tAso)(l/sin0) + (tAS0) Eq. (2.8) The overlayer thickness (t), as a result, can be found from the slope of In [IS(0)/IS(9OO)] versus l/sin0, supposing information is available for A,so. 33 To detector A To detector 3X Binding Energy (eV) (b) Figure 2.6 Angle dependent measurements: (a) definition of take-off angle 9, the angle between the sample surface and the collection direction; (b) measurements at 0=30° and 90° from an oxidized Si sample. In (b) the substrate layer of Si0 2 are emphasized relative to the elemental form at 0=30°. 34 2.4 Instrumentation A Leybold MAX 200 spectrometer is operated in our laboratory and a plan of this facility (a "bird-eye-view") is shown in Figure 2.7. The spectrometer is composed of four interlinking chambers: the sample transfer chamber, the analysis chamber, and two sample preparation chambers (#1 and #2). The transfer chamber houses an automatic transfer rod and a sample magazine. The analysis chamber accommodates the main parts of the spectrometer, including the x-ray sources, the manipulator (PTM 60), the energy analyzer (EA 200) with its input lens and detector. In addition, the analysis chamber is equipped with an ion gun for sample sputtering and for ion scattering spectroscopy (ISS), an electron gun for measuring Auger spectra, an electron flood gun to help compensate for any sample charging, and an x-ray monochromator. In-situ sample preparations are performed in the preparation chambers. Preparation chamber #1 is used for low-pressure surface treatments, like metal deposition or gas dosing, where background pressure during the treatments is around 10"6 to 10"9 torr. An evaporation source, a variable leak valve (for gas introduction) and a quadruple mass spectrometer (RGA) (for residual gas detection) are included in this chamber. Higher-pressure surface treatments (e.g. up to -10 torr), such as for plasma modification, are performed in preparation chamber #2. Two manual transfer rods (#1 and #2) are available for transferring treated samples from these preparation chambers to the analysis chamber. 2.4.1 Ultrahigh vacuum (UHV) Modern surface characterization studies are necessarily performed under ultrahigh vacuum (UHV) conditions, in which the system pressure at time of measurement is reduced to around 10"9 torr. This prevents a build up of contamination from the residual gas in the vacuum system. According to the kinetic energy theory of gases [123], at room temperature and a 35 J L Preparation Analysis chamber #1 chamber Figure 2.7 A Schematic diagram of the MAX 200 system viewed from top. pressure of 10"9 torr, a surface can be covered by an adsorbed monolayer, of for example CO, in about one hour assuming a sticking probability of unity. In practice, sticking probabilities are usually considerably lower but background pressures around 10"9 torr or less are necessary to keep a surface in a constant state (free from contamination) during the time to make a measurement. A schematic indication of the pumping system used in the MAX 200 facility is shown in Figure 2.8. A combination of rotary pumps and turbomolecular pumps are used for the main chambers, while ion pumps are used for the x-ray sources and the gas manifolds. Additionally titanium sublimation pumps are attached to the analysis chamber and preparation chamber #1 for supplementary pumping. To achieve the UHV condition, the chambers are initially pumped down to around 10" torr by rotary pumps, and subsequently to 10" torr (or below) by the turbomolecular pumps. In order to reach the 10"9 torr region (or below), the chamber is baked at ~120°C for 12 h (or more). Typically the base pressures of the analysis chamber and preparation chamber #1 are around 5xl0"10 torr, while those of the transfer chamber and preparation chamber #2 are about 2x10"8 torr. 2.4.2 Sample handling Figure 2.9 (a) shows the sample transferring systems possible in the MAX 200 facility. Samples that are to be studied in the "as received" condition are mounted on a standard sample holder and locked on a sample magazine (can hold up to seven sample holders) which is then introduced to the transfer chamber prior to the initial pumpdown. Once the transfer chamber pressure is reduced to 10"8 torr, one sample holder is transferred to the analysis chamber and locked on the manipulator (PTM 60) dock. This manipulator gives five degrees of movement to the sample (Figure 2.9(b)), three for linear motions (X, Y, and Z) and two for rotational motions 37 gure 2.8 The pumping system for the MAX 200 system (modified from ref. [24]). 38 Preparation chamber #2 Pedestal Transfer rod #1 Analytical chamber Manipulator Preparation chamber #1 T Transfer rod #2 • Transfer chamber Transfer rod Sample magazine Hindge door (a) Figure 2.9 (a) Transfer devices in the M A X 200 facility (modified from ref. [24]). (b) Shows the five degrees of movement possible for sample on the manipulator; there are three linear motions (X, Y, Z) and two rotations (T and R). 39 (T and R). This enables proper position settings and the ability to perform angular dependent measurements. Similar procedures are carried out for air sensitive samples except they are mounted and transferred in a N 2 glove-box to prevent air exposure. When a sample is to be given a plasma treatment, it is mounted on a special sample holder associated with a heating device. The sample is introduced to the preparation chamber #2 and the plasma treatment is made there. Then the preparation chamber #2 is pumped down to around 10"8 torr and the sample is transferred to the pedestal in preparation chamber #1, through which the sample can be heated or exposed to dosing gases. Transfer to the analysis chamber is made after the pressure in preparation chamber #1 has been pumped down to 1CT8 torr. 2.4.3 X-ray sources Figure 2.10 (a) shows a schematic diagram for the dual anode x-ray source equipped in the MAX 200. The anode has two faces, with separate deposited films of Mg and Al (-10 pm thick) and there are two filaments located on each side of the anode (Figure 2.10 (b)). The external circuit can switch from one filament to the other; the operational one is at earth potential while a positive potential of up to 15 kV can be applied to the anode. Accelerated electrons emitted from the filament hit only the nearest anode face, and the radiation generated passes out through an aperture, covered with a thin Al (~2 urn) window. This window prevents stray electrons, radiation and contamination reaching the sample from the source, which accordingly provides either MgKa (1253.6 eV) or AlKa (1486.6 eV) radiation. The x-ray source is generally operated at 10 kV and 20 mA, but a range of currents can be selected with the maximum power dissipation.being up to 600 W for the Mg source and 800 W for the Al source. It follows that the heat generated during the measurement must be removed from the anode quickly and efficiently. For good thermal conductivity, the body of the anode is 40 Figure 2.10 Schematic diagrams of (a) the Mg/Al dual anode x-ray source and (b) the filaments in the x-ray source, (c) Energy distribution of A l x-rays, showing the Kcc and K(i lines and the high-energy bremsstrahlung radiation [124]. 41 made of copper, while the cooling process is accomplished by internal water circulation (see Figure 2.10 (a)). Figure 2.10 (c) shows an energy distribution of the radiation produced by the Al source; there are the characteristic K a (Kai,2 and Ka3,4) and K(3 lines on a broad background (the bremsstrahlung radiation) [124]. Although the emission is dominated by Koti,2, the x-ray source is not monoenergetic. To remove the K0C3,4 and K p lines, an x-ray monochromator is needed. The monochromator consists of a bent quartz crystal (on the surface of a Rowland circle) through which the raw radiation is monochromatized by diffraction [24]. The resulting beam has a narrower energy spread and a lower background, because the bremsstrahlung radiation is removed as well; the overall flux is reduced by several orders of magnitude compared with that from an unmonochromatized source. 2.4.4 Energy analyzer A schematic diagram for the EA 200 energy analyzer of the XPS facility is shown in Figure 2.11. It consists of three main components: (a) the input lens system, (b) the concentric hemispherical analyzer (CHA) and (c) the multichannel detector (MCP). The input lens system transfers an electron image of the analyzed area on the sample to the analyzer. In the EA 200, the electron image experiences two transferring stages. The first lens stage, including a variable angular aperture A l and image aperture A2, controls the analysis area (spot size) and acceptance angle (£2) for the input electron image. After passing through A2, the selected electron image is focused onto the slit S1 by the second lens stage, which also acts to retard the electron energy to a particular pass energy (see below). In addition the second lens stage controls the entrance angle (a) for the electrons to enter the analyzer by A3. 42 Figure 2.11 Schematic diagram for the concentric hemispherical analyzer and input lens system in the M A X 200 system'. 43 The CHA (Figure 2.11) is constructed from two concentric hemispherical electrodes (inner radius Ri and outer radius R2) and to which a deflecting potential AV is applied [24]. Electrons travel on the central circular trajectory and reach the detector planar S2 at the nominal radial position Ro, where Ro = (Ri+R2)/2, provided their kinetic energy En, inside the analyzer (the pass energy) satisfies eAV = E 0 (R2/Ri - R1/R2) Eq. (2.9) The scanning of kinetic energies, therefore, can be accomplished by continuously varying AV. With entrance angle of a, the analyzer resolution of the analyzer, AEanaiyzer [24] is given by AEanaiyzer/Eo = w/2R0 + a2/4 Eq. (2.10) where w is the slit width (equal at the entrance and exit). Since w, Ro, and a are limited by the spectrometer construction, the resolution of the analyzer is therefore not constant but varies with En. In fact, a measured peak width, AEpeak (defined as the full width at half-maximum, FWHM), also has contributions from the inherent line width of the atomic level involved (AEiine) and from the natural line width of the x-ray source (AE s o u r c e). The observed peak width satisfies 2 2 2 1/2 AEpeak — (AE source AE analyzer "t" AE [jne) Eq. (2.11) provided all contributions have the Gaussian form [24]. The EA 200 is normally operated in the constant resolution mode to ensure a constant analyzer resolution at all energies in a spectrum. The entering electrons are first retarded by the lens system to a fixed pass energy, which is set and remains unchanged during an entire measurement. The deflecting potentials on the analyzer are therefore pre-set, according to Eq. (2.9), for that particular pass energy before the acquisition. The actual scanning of the kinetic energies is done by ramping the retarding field voltage during the pre-retardation process. According to Eq. (2.10), the lower the pass energy, EQ, the better the analyzer resolution, but the 44 resulting signal intensity drops as well. Therefore an optimal balance is required between resolution and intensity, and accordingly an appropriate pass energy is chosen for each measurement. In this work, a maximum pass energy of 192 eV was used for measuring wide scan spectra, while spectra at higher resolution were measured with pass energies of either 48 or 96 eV. Measurements on a standard Au sample gave values of AEpeak for the Au4f7/2 peak equal to 1.84, 1.28 and 1.06 eV for the pass energies 192, 96 and 48 eV, respectively. Now let us consider those electrons which enter SI at an angle 8a to the tangential direction, with an energy slightly different from the pass energy. In spite of shifting off the Ro radius, these electrons still roughly travel along circular trajectories, but they reach further inside (or outside) on S2. This offers the possibility of using multichannel plate (MCP) detection for the simultaneous recording of an energy band around E 0 . The detector in the EA 200 is constructed from two MCPs assembled in a back-to-back configuration to form a chevron array. Each MCP is an array of 18 capillary-type microchannels, each of which acts as an individual electron multiplier. The plates are oriented so that the channel angles of the two plates are in opposition; this suppresses feedback by trapping ions at the interface between the two plates. The voltage across the plate is set to allow count rates at above 107 s"1. As illustrated in Figure 2.12, the kinetic energy of a photoelectron measured in the spectrometer (E'k) is referenced to the spectrometer's vacuum level, while the binding energy of the electron inside the sample (Eb) is referenced to the Fermi energy of the sample. For a conducting sample in electrical contact with the spectrometer, so the Fermi energies are equal, the energy balance requires E ' k = hv - E b - W s p Eq. (2.12) and this represents a modification of Eq. (2.1). The spectrometer work function (Wsp) remains constant while the analyzer is held under UHV; its value is routinely checked by calibrating with a standard gold sample. Therefore Eq. (2.13) enables measured kinetic energies of 45 Vacuum level Fermi level Sample Spectrometer Ek Itt) Ws Eb v Sample T Ek1 Wsp Vacuum level Fermi level Spectrometer h-u = energy of the photon Wsp = work function of spectrometer (i.e. energy difference between the Fermi level and vacuum level) Ws work function of sample Ek = kinetic energy of photoelectron with respect to vacuum level of sample Ek' = kinetic energy of photoelectron measured by the spectrometer E b = binding energy of electron in solid with respect to the Fermi level Figure 2.12 Schematic of the relevant energy levels for binding energy measurements. 46 photoelectrons to be converted to binding energies for the electrons in the solid. The spectra shown in this thesis have been calibrated against the gold 4fja peak, whose binding energy is at 84.0 eV. 2.4.5 Data processing Some processing of raw XPS measurements is needed before doing a spectral interpretation, and this especially involves subtracting background in order to emphasize contributions from the surface region. Several methods have been proposed for making this correction [126,127]. In the DS 100 data processing system, the Shirley non-linear background subtraction [126] is used, and that is illustrated in Figure 2.13. This method assumes that the dominant contributions to the background come from inelastically scattered photoelectrons, such that at any point in some spectral structure, the background signal is proportional to the number of electrons elastically scattered at higher kinetic energies. This correction for inelastic contributions over an energy range E i to E 2 (lower and upper energies chosen by the operator) is determined by the iterative algorithm: N'k+i (E) = N(E) - N(E2) - C 2 N'k (E) dE Eq. (2.13) where N(E) is the measured count rate and the N'k(E) identify count rates after subtraction of background contributions (the index k indicates the kth iteration). The reference background level is provided by N(E2) and C is fixed by the requirement that N'k(E|) =0. The process starts with N'i(E)=0 and continues until N'k+i - N'k' usually it converges after three or four iterations. After the background subtraction, the peak area associated with the elastically scattered photoelectrons can be determined by integration. As mentioned earlier, high resolution spectra often show structure arising from overlapping chemical species. A curve synthesis procedure is needed to identify the individual components with regard to peak position, intensity and peak width (FWHM). To accomplish 47 Figure 2.13 Shirley non-linear background subtraction applied to a spectrum with overlapping Mo3p and Nls peaks. 48 this, it is necessary to choose a functional form for each component profile. In the DS 100 data processing program, these functions have the mixed Gaussian/Lorentzian form [24]: f(E) = peak height / [1 + M (E-E0)2 / p 2 ] exp{(l-M) [ln2(E-E0)2] p 2} Eq. (2.14) where Eo is the maximum of the individual component peak, p fixes the FWHM, and M is a mixing ratio (1 for pure Lorentzian; 0 for pure Gaussian). In this work spectra were fitted with a 80% Gaussian/20% Lorentzian peak shape as optimized in curve fittings to M o 3 d and S2p spectra measured for single crystal M0S2. An example of curve synthesis is illustrated in Figure 2.14 for a M o 3 d spectrum from a thermally oxidized Mo metal sample. Two major peaks, at 235.8 and 232.6 eV respectively, correspond to the 3d3/2 and 3 d 5 / 2 doublet of the Mo(+6) oxide. However, the doublet is not symmetrical, and it appears that a small peak at around 228.0 eV must be present. Therefore, it is concluded that the sample has more than one kind of Mo species. Further, within the doublet the area ratio and the splitting should have appropriate values. For each component in M o 3 d , the 3 d 3 / 2 to 3 d 5 / 2 area ratio is kept at 2:3 while the splitting is fixed at 3.2 eV. Three overlapping components are then seen to be required with 3 d 5 / 2 at: 232.6, 230.4 and 228.0 eV. The curve synthesis is done iteratively by optimizing the fit between the measured curve (after background removal) and the sum of the optimized component functions. The quality of fit is evaluated by the least-squares function: X = 1 ^ 0^me&,\ Yfit.i) N Y 1 free i=l *• mea.i Eq. (2.15) where the sum is over the N data points in the spectral region to be fitted, Ymea,i is the measured count rate at the ith data point, Yfltji is the corresponding value of the fitting function which involves a sum of functions of the type in Eq. (2.15), and N f r e e equals N - N f i, where N f l, is the number of parameters to be fitted. If the curve synthesis is repeated with a different set of starting parameters, the new % value will indicate whether the change has improved the fit or not. 49 Mo3d Mo(+6) 244 239 234 229 224 Binding Energy (eV) Figure 2.14 Three Mo.species identified from a Mo3d spectrum measured from an oxidized metal sample. Each species has two (3d3/2, 3d5/2) spin orbit coupling components with a splitting of magnitude 3.2 eV. 50 Generally the smaller the % value, the better is the fit. Nevertheless visual comparisons are still very useful to make sure that satisfactory agreement has been reached between the simulated and measured spectra. Also, the optimal application of curve fitting methods requires knowledge about the chemistry of the system to ensure that a proposed fitting is reasonable on chemical as well as on mathematical grounds. The work in this thesis has many examples of curve fittings using the principles just outlined. Figure 2.15 shows two examples for Mo3d spectra measured from M 0 O 3 thin films after nitriding with NH 3 at different temperatures. Each example has been presented in duplicate from independent measurements and independent curve fittings. The tables included show the relevant binding energies and the compositions (on a percentage basis) indicated for each component in each case. This thesis emphasizes trends through sequential operations, and although duplicate measurements have generally been made, results are tabulated for particular sets of measurements. The comparisons shown in Figure 2.15 are typical for individual measurements at a particular stage; the binding energies quoted later for particular components at one measurement stage can be considered reliable to the ±0.1 eV level, while the associated percentage compositions are reliable to about the ±5% level. 2.5 Sample charging 2.5.1 Energy referencing With conducting and semiconducting samples, no charge is accumulated during photoemission because there is sufficient electrical contact with the spectrometer ground circuit to provide replacing electrons. For insulating samples, on the other hand, the electrical contact with the spectrometer is lost and positive charges can build up on the surface. The exiting 51 Figure 2.15 M o 3 d spectra from M0O3 thin films after nitriding at (a) 450°C and (b) 750°C. The 3d5/2 binding energies (eV) and proportions for different components (see chapter 6) from each individual measurement are listed in the tables. In each case independent preparations and measurements are made (trial 1 and 2). 52 photoelectrons then require extra energy to overcome this potential barrier. Consequently, when a positive sample charge builds up, the photoelectrons emerge from the sample with a lower kinetic energy, and they appear in an XPS spectrum as at a higher binding energy. When non-monochromatized x-ray sources are used, there is some self-control in this sample charging. This arises because of the presence of the high-energy bremsstrahlung radiation, which provides an electron atmosphere in the vacuum (e.g., photoelectrons emitted from chamber walls) through which the charges on the sample can be partially neutralized. Then a steady-state charging of several eV may be reached, and this can result in an uniform binding energy shift throughout the spectrum. In this situation of uniform sample charging, the spectrum can be corrected by a calibration to an energy reference of known binding energy. Since most samples contain some adventitious carbon from the atmosphere, the associated Cls photoemission peak is commonly used as an internal standard. When comparing data from different groups some care may be needed if results are referenced to slightly different binding energy values; for example, the Cls peak just referred to has generally been referenced in the range 284.6 to 285.2 eV in the literature [128]. In this work 284.7 eV is chosen because this value was measured on a Mo substrate which is believed to be free from charging (see Chapter 6). For non-conducting samples, especially with inhomogeneous components, different areas of a sample may have different photoemission rates and electrical characteristics, and hence may have different steady-state potentials. That means the sample will have developed a non-even surface charging, which is generally known as differential charging. The consequent varying shiftings of structure in spectra may be mistaken for chemical shifts associated with different chemical states, and therefore it is very important that this phenomenon of differential charging be recognized. The approach used in this laboratory to identify this problem is to use the bias potential technique, which is described in the next section. 53 2.5.2 Differential Charging Empirically it is believed that the presence of differential sample charging can often be identified by applying a large negative bias potential (e.g. -93 V) to the sample holder during XPS measurements [119,129,130], and these measurements are compared with the case where no bias potential is applied. The spectrum measured with the -93 V bias potential is then mathematically shifted back by +93 eV on the energy scale to compare with the spectrum measured with no bias potential. Areas on the sample with different degrees of electrical contact with the sample holder are likely to develop their own steady state potentials, but these will also vary with the applied bias potential. The comparison of spectra measured, on the same energy scale, for the biased and unbiased cases should therefore identify contributions to the spectra which are especially associated with the charging problem. In particular, contributions to spectra which are indicated to occur at the same energy for both types of measurement can generally be taken to be in proper electrical contact with the spectrometer so that the associated binding energies can be given a chemical significance. Another very useful approach for handling XPS spectra from insulating regions of materials is provided by the Auger parameter, a', introduced by Wagner [36]. He defined this parameter for a particular element as the sum of the kinetic energy of its sharpest Auger peak and the binding energy of its most intense photoelectron peak; that is a'=KE(Auger) + BE(photoelectron) Eq. (2.16) The importance of this quantity is that it is independent of the energy zero, so that if one of these energies is too small by Ax then the other is too large by the same amount. This inevitably makes a' independent of charging effects, and in general the Auger parameter has provided a convenient approach for interpreting XPS measurements from poorly conducting samples [36,37]. Examples of the use of this parameter are given in Chapters 4 and 5. 54 3. Secondary Ion Mass Spectrometry 3.1 Introduction Secondary ion mass spectrometry (SIMS) is a very sensitive UHV technique for surface analysis. Over the past twenty years SIMS has been developed into a powerful tool for studying chemical composition and structure of solid surfaces [131,132]. It is also a complementary technique to other methods like XPS and AES. SIMS involves bombarding a surface with primary ions of known charge and energy, and detecting secondary positive (or negative) ions after they have passed through a mass analyzer (Figure 3.1 (a)). Figure 3.1 (b) schematically outlines the SIMS sputtering process. An incident ion beam (energy 0.5-20 keV [133]) impacts with the sample surface. These particles transfer their energy to atoms on the surface; neutral secondary species and ions are emitted. The process of sputtering has been studied extensively [134-136] and the theory that best describes the quantitative aspects of sputtering is the collision cascade model of Sigmund [135]. The theory relates energy loss in the primary ions to a cascade of collisions with surface and near surface atoms. During the bombardment, some energy will be dissipated into the bulk by the displacement cascades as subsurface atoms get moved out of their regular lattice sites. With the random motions in this process, some cascades will return to the surface, causing the emission of secondary particles. The emitted secondary particles can be electrons, neutral species (i.e., atoms and molecules) and ions (i.e., atomic and cluster ions); most are neutral but a few percent carry positive or negative charges. These ions are conveniently detected in a mass spectrometer, which measures the mass/electric charge (m/e) ratio. Since the emission process occurs within 1 to 2 nm of the surface, SIMS has high surface sensitivity [132]. SIMS can analyze all elements in the periodic table including isotopes (from H to U). With the most advanced equipment, detection limits are in the parts-per-million (ppm) range for 55 c 3 c o Sample Mass analyzer Ion detector (a) 11 lull i. ll Mass spectrum Depth profile Chemical image Primary ions Secondary particles err o o:o;q o o o o _ o „ o o^Jq\J ^ o \ o 'O OOODO Implanted primary ion (b) Figure 3.1 Secondary ion mass spectrometry: (a) schematic diagram of the major components of a SIMS spectrometer; (b) illustration of phenomena in the SIMS process. 56 most elements and parts-per-billion (ppb) for a few [132]. In addition, it can indicate local bonding arrangements in the solid from the form of molecular fragments detected among the secondary ions [132]. SIMS can give its information as a function of depth, in sputter profiles, or as a function of position, in chemical maps. However the proportion of secondary ions generated during the sputtering process varies greatly between different elements and compounds, and matrix effects can be strong [131,137]. 3.2 SIMS analyses SIMS necessarily disturbs a surface under study, and after a SIMS characterization strictly the sample cannot be recovered in its original form. SIMS analyses may be conducted within three regimes, namely static SIMS, dynamic SIMS and imaging SIMS. 3.2.1 Static SIMS Static SIMS (SSIMS) uses a low energy and low flux inert ion beam (e.g. Ar+). Typically, the current density used is of the order of a few nAcm"2 [138] whereas the beam energy is between 0.5 and 5 keV. This gives a low sputter rate and so limits the disruption of the sample surface. For example, the lifetime of a monolayer on the material can be extended to several hours, which is in excess of the time required for a measurement [132]. In general, no one point of the surface is hit more than once in SSIMS, and therefore this type of analysis gives information on the original surface of the sample, which can essentially be taken to be non-destructive. During SSIMS measurements, secondary ions are detected within a particular chosen ion mass range (e.g., 1-250 amu) and a mass spectrum is obtained with secondary ion intensity given as a function of m/e. SSIMS spectra are useful for studying elemental composition (with sensitivities slightly less than those applying in XPS or AES [24]), but they 57 can also indicate chemical structure of the topmost layers by detecting particular groups of bonded atoms in the secondary clusters. Figure 3.2 compares positive and negative ion SSIMS spectra measured from a planar A10x support and from a sample with a thin film of M 0 O 3 deposited on A10x. The MoCVAlOx sample shows extra secondary ion clusters of Mo+, MoO+ and MoC»2+ types in the positive ion SSIMS spectrum (Figure 3.2 (c)), and of M 0 O 3 " and M 0 O 4 " cluster ions in the negative ion spectrum (Figure 3.2 (d)). The surface appears heterogeneous, perhaps as a result of M 0 O 3 segregation, since A l + is detected, but compared with the spectrum from the A10x support, the M0O3 influences the ion yields of O" and OH" as well. 3.2.2 Dynamic SIMS Dynamic SIMS uses much higher current densities (typically p:Acm"2 to mAcm"2) for the primary ions (e.g. Cs+ or O2"1" [132]), and this ensures much higher yields of the secondary ions. Then the surface can be eroded rapidly to yield high-sensitivity concentration-depth profiles of the sample [139]. Nevertheless the primary ion doses are large enough to cause significant disruption of the near-surface region of the sample, so information on molecular bonding arrangements is lost. Consequently dynamic SIMS is used almost exclusively for elemental analysis [132]. To obtain a depth profile, the mass analyzer is set to the masses of the interested elements and signals are collected from each element in rapid succession. In each cycle, these signals come from elements at the same depth. Then spectra measured as a function of sputtering time provide information about the change in elemental concentrations as a function of depth (this is depth profiling). Dynamic SIMS measurements are also useful to study changes in a sample as a function of treatment, for example to assess interface sharpness and diffusion broadening after carrying out an annealing process [131-133,137]. 58 Positive ion SSIMS Negative ion SSIMS 3000 2500 o 2000 u >• 1500 '(A c ID 1000 c 500 0 " A1+ — K + , O H 3 + — | A10+ Q H 7 + T | Cu+ 1 1 v J i IL ..... M 111 1 I 1 1 11 1 11 1 ii 1 I I I II I I M I I I I II I I n i II11111111111 II 111 0 20 40 60 80 100 120 140 160 180 Mass (a.m.u) (a) 3000. 2500 o u WD 2000 o 1500 'So c 1000 c 500 0 200 150 100 50 0 Mo" U1DJ3U MoO + M0O2' 1. i l l i l . A i J l , , 85 95 105 115 125 135 I I I!! iI til 1 I 1!It jIM !!I I! I 1 i I i I I i j I 1 [ 1 H I I I I I I I I 1 1 1t I i [I II Itl I I M I I It I I II I I I II ItI I M \I I J.HI 0 20 40 60 80 100 120 140 160 180 Mass (a.m.u) (c) 10000 8000 6000 4000 2000 0 — 0-y OH" / o r - 1 Q H" Cl" F l l i ' I 1 A-ld^^n. , . . V. i i i n i i 1 1 II1ll1 II II 1 III III II II11111| 111 II 1 1 1 1 1 II11 1 1 II 1 1 1 1 II I • 1 :1 1 • I 1 - 1 0 20 40 60 80 100 120 140 160 180 Mass (a.m.u.) (b) 10000 8000 6000 "S 4000 a 2000 0 130 140 150 160 170 180 r 1 III 1 tit a i Il I I I III II II I I II I I I l l i l l II I l l l i l l l l l l l l l l l l l l l ! I II II! II II I l l l l I I i l l III I I I III II II I II I I 0 20 40 60 80 100 120 140 160 180 Mass (a.m.u.) (d) Figure 3.2 Static SIMS spectra: (a) positive ion and (b) negative ion spectra from an oxidized Al support; (c) positive ion and (d) negative ion spectra from a MoOVAlOx sample. 3.2.3 Imaging SIMS Imaging SIMS uses a microfocused liquid metal ion beam (e.g. Ga+, width typically 0.05-0.5 pm [132]) in which metal ions are field ionized from a sharp tungsten tip [140]. SIMS images can then be obtained by raster scanning such a beam over a defined region of a sample. For each incident beam position, the controlling computer system records intensities of chosen secondary ions at the corresponding point, and images of the lateral distribution of elemental components across the sample can be built up as the incident beam moves through its different positions. Images from ion-induced secondary electron emission can also be recorded (Section, and Figure 3.3 compares a 9 5 Mo + image from a MoCVAlOx sample, which had been calcined at 200°C, with the corresponding secondary electron micrograph (SEM) from the same area. The SEM image shows the roughness of the surface while the SIMS image indicates the Mo distribution over the same region. 3.3 Matrix effects To quantify SIMS data, the measured secondary ion intensity from the analyte must be converted to a concentration. However one major challenge in SIMS is to assign concentrations to secondary ion signals. This difficulty arises in part from the large variation in the proportion of ions formed during the sputtering of different elements and materials. For the same type of material, the sputter yield and ionization probability can also depend on such features as atomic composition, chemical state [141] and crystallographic orientation for the target [142]. For instance, the positive ion yield generally increases as a metal is oxidized [141] (e.g., the ion yield for Al increases by a factor of 10 after oxidation, and this increase can be greater for other elements [143] Thus the dependence of ion yield on details of the matrix limits the quantitative nature of SIMS unless very careful calibrating corrections are made [131,137], and this is 60 2000X SEM UNG29.SEI1 50 ^ Figure 3.3 Ion-induced SEM and SIMS images from a Mo03/A10x sample which had been calcined at 200°C. 61 particularly challenging for non-homogeneous samples. For more routine studies, as done in work with SIMS in this thesis, this technique should just be regarded as being semiquantitative. 3.4 Instrumentation SIMS measurements carried out in this work were performed in a VG Scientific SIMS spectrometer, which is schematically shown in Figure 3.4. The main chamber is separated from the transfer chamber by a gate valve. After samples have been loaded, the transfer chamber is pumped down to 10~8 torr. Then the sample is transferred to the analysis chamber by the wobble stick while the analysis position can be adjusted through the manipulator. The main chamber pumping is maintained by two diffusion pumps with one shared backing pump (Figure 3.5). An additional supplementary pump on the analysis chamber is provided by the Ti sublimation pump. The base pressure of the analysis chamber is around 2x10"10 torr while that of the transfer chamber is at 5x10~9 torr. The duoplasmatron ion source has its own combination of diffusion and backing pumps through which its base pressure can be maintained below lxlO"7 torr. Three ions sources are present in this facility. The argon ion source (AG61) is used primarily for static SIMS measurements, while the gallium liquid metal ion source (MIG100) and the duoplasmatron ion source (DP5B) are used for SIMS imaging and depth profiling respectively. The quadrupole mass filter (MM12-12S) functions as the mass analyzer for the SIMS measurements and also as a residual gas analyzer (RGA). Its mass detection range is 1 to 800 amu. The SIMS spectrometer additionally is equipped with a scintillator for ion-induced secondary electron imaging, and with a charge neutralizing electron source (LEG31) to aid the neutralization of surface charging which can occur during SIMS measurements on insulating samples. 62 Quardupole mass filter Precision stage manuipulator Photomultiplier & scintillator Analysis chamber Duoplasmatron ion gun Ga gun Transfer chamber Transfer rod Wobber stick Ti sublimation pump Figure 3.4 Schematic diagram showing the main components of the VG Scientific SIMS spectrometer. 63 ^^J) Rotatory pump Difussion pump Ti sublimation pump (^) Valve Figure 3.5 Pumping system for the VG SIMS spectrometer. 64 3.4.1 Ion sources Electron impact ion source Figure 3.6 shows a schematic diagram for the AG61 electron impact ion gun in the VG Scientific SIMS system. The gun consists of two major components: the electron impact source and the ion optical column. The electron impact source has two filaments, one on each side, and is suitable for using high-purity noble gases, such as Ar or Xe. Gas is supplied to the source region via the leak valve from the high pressure (~2 atm) gas line. Current (up to 5 A) is passed through the filament, for electron emission by the thermionic effect, and these electrons are accelerated to ionize the gas. A voltage (0.1-5 kV) is applied on the extractor to attract and accelerate ions produced by the electron impact. The ion optical column consists of lenses to control the spot size and give beam focusing, while the four electrostatic rods, at the end of the gun, are used to raster the ion beam across the sample [144]. The AG61 gun gives a broad beam (diameter -200 urn), but this is satisfactory for SSIMS measurements; those done in this work used 5 keV Xe+ primary ions with a sample current of 0.2 nA. Gallium liquid metal ion source The gallium liquid metal ion source (MIG100) is schematically shown in Figure 3.7 (a). The major components of the gun are the Ga source (Ga reservoir and needle), the heater, the extractor and the optical lenses. The use of Ga metal depends on its low melting point (38°C). On heating the reservoir with a current of 1 A for 15 minutes, the metal melts and wets the needle whose tip radius is 10 |im. A potential difference is applied between the source and extractor. For a potential difference of 4 kV, the electrical field-stress on the surface of the liquid 65 Source gas inlet Ion source assembly Ion optical column Differential pumping port Scan rods Figure 3.6 Schematic diagram for the AG61 electron impact ion source showing: (a) the ion source assembly, and (b) the ion optical column (modified from ref. [144]). 66 Extractor Ceramic FT A F > <7T\ < < W W W A A A A A A A / V 1 Needle Reservoir — Ga source Heater bundles (a) Extractor Optical Column Heater Liquid film (b) Figure 3.7 Schematic diagram for the MIG100 liquid Ga ion gun in (a); (b) indicates the beam focusing unit (modified from ref. [144]). 67 metal overcomes the surface tension; a small cone of liquid Ga forms at the tip of the needle (Figure 3.7 (b)) and ion emission occurs from the cone. The outgoing Ga ion beam is then focused and given fine adjustments by the optical lenses [144]. The ion beam energy is controlled by the voltage (typically 1-10 kV) on the needle and reservoir. The MIG100 Ga gun is used for imaging SIMS because its beam diameter of about 0.5 u\m allows for rastering over the sample surface. The imaging SIMS measurements done in this work were performed with a beam energy of 6 keV and a current of 0.5 nA. Duoplasmatron ion source The VG Scientific DP5B duoplasmatron ion source in the SIMS facility normally uses a mixture of argon and oxygen (e.g. 90% Ar), and it produces primary ions by plasma arc discharge (300-600 V) between the cathode and anode of the source. High density plasma is achieved by the dual constricting action of a small bore passageway between cathode and anode and the effect of a magnetic field applied in the vicinity of the constriction [131,145]. An opening in the anode allows ions to be extracted from the plasma. Although the ion beam from the duoplasmatron source is broader (diameter -1-100 urn) than that from the Ga gun, the former provides higher current densities, which in turn give faster sputter rates and more rapid data acquisition. The DP5B duoplasmastron ion gun is convenient for measurements with dynamic SIMS, but that was not used in the work reported in this thesis. 3.4.2 Detectors Quadrupole mass filter Figure 3.8 (a) shows a schematic diagram for the MM12-12S quadrupole mass analyzer which is based on three major components: (a) the ion energy filter, (b) the mass filter and (c) the 68 ion detector. The ion energy filter acts to prevent high energy ions from entering the mass filter, which is composed of four molybdenum rods (diameter 12 mm) aligned by ceramic discs. Opposite pairs of the rods are connected electrically; one pair has a sinusoidal potential as in Eq. (3.1), while the other pair has an equivalent but reversed-sign potential (Eq. 3.2): <Kt)= U + Vcos (2TC ft) Eq. (3.1) <Kt) = - U - Vcos (2K ft) Eq. (3.2) Parameters U and V are in the ranges to 100 V and 1500 V respectively, while f is the radio frequency range (e.g., 2 MHz). Ions entering the region between the rods along the z-axis experience one of two motions as a result of fields applied perpendicular to the z-axis. These motions are indicated schematically in Figure 3.8 (b). One, shown by the full line, gives an oscillation about the z-axis while still allowing emergence from the far end of the filter. The second motion (indicated by the dashed line) is unstable with respect to motion on the z-axis, and ultimately leads to the ion striking a rod and being neutralized. Possiblities for transmission through the filter, for ions with a specific value of m/e, are determined by the instrumental parameters U, V, and f. Mass scanning is usually achieved by varying U and V while keeping the V/U ratio and f both constant [146]. Ions exiting the mass filter are detected by a channeltron multiplier which can give gain of up to 108 over the input signal. Scintillator-photomultiplier Collisions between primary ion beams (e.g., Ar+, Ga+) and a sample also generate secondary electrons, and the number released depends especially on the surface morphology and variations in composition. Ion-induced secondary electrons can be conveniently detected by a 69 (b) Figure 3.8 Schematic diagram for the MM12-12S SIMS quadrupole mass analyzer: (a) major components of analyzer (modified from ref. [144]); (b) the mass filter. 70 scintillator-photomultiplier (Figure 3.9) whose phosphor emits light on electron impact. The secondary electrons are emitted in all directions and have low energies (<50 eV), but they can be collected by application of a positive potential (e.g. 100 V) on the metal grid at the front of the scintillator [147,148]. The latter is biased with a voltage of -10 kV to accelerate the secondary electrons sufficiently to excite the scintillator, which can then emit a proportional number of photons. The photons travel along the light pipe to a photomultiplier and generate an amplified photocurrent signal which is displayed on the cathode ray tube. For each incident beam position, the controlling system records the secondary electron intensity at the corresponding point, and a SEM micrograph (physical image) of the sample can be built up by scanning the probing beam across the sample [144]. 71 Figure 3.9 Schematic diagram for the scintillator-photomultiplier detector: the generation of a SEM image for a sample is accomplished by scanning the ion beam across the sample surface (modified from ref. [144]). 72 4. Study of Calcinated Molybdenum-Aluminum Oxide Model Catalysts 4.1 Introduction For all supported catalysts, high dispersion is required for the active reagent, and there are indications that this can vary substantially during the preparation process [54]. For M0/AI2O3 catalysts, it is the calcination step which determines the nature and dispersion of Mo in the final working catalysts (sulfided or nitrided form) [54,73]. Calcination at around 500°C appears favored on empirical grounds, although understandings of the dispersion in terms of fundamental bonding involved are still limited. A number of XPS studies of this catalyst system (with or without sulfur) have been published since 1975 [33,62,69,97]. XPS has helped probe the Mo dispersion [51,62,69] and studies covering a range of Mo loadings have led to the belief that well-dispersed, supported Mo of approximately monolayer thickness is obtained after the regular calcination treatment [33,69,73]. However that approach has been less effective at probing the nature of the Mo-support interaction, although laser Raman spectroscopy (LRS) and infrared spectroscopy (IRS) have indicated the presence at lower coverages of both tetrahedrally and octahedrally coordinated Mo, resulting apparently from the reaction of polymolybdic acid with surface OH group [33,62,69]. For loadings corresponding to more that 20% M 0 O 3 on alumina, both Al 2(Mo0 4) 3 and bulk M 0 O 3 are reported [62,73,147,148], with their formations being favored by both increased calcination temperature and increased calcination time [62]. As mentioned in Section 1.3.1, one difficulty for XPS studies on real supported catalysts is that the photoemission peaks are often broadened [33,61,62], and it is not always clear whether this broadening arises from a summation over components which are genuinely shifted in energy, or whether sample charging has a role. The present work, which follows the model catalyst approach, uses a planar support of alumina on metallic aluminum, and studies the dispersion of 73 added Mo as a function of calcination pretreatment. Model catalyst samples have been studied before [149,150], but, rather than evaporating the metal under vacuum, our emphasis is to use equilibrium adsorption [51,151] from solution following procedures used for forming real catalysts. The objective here is especially to probe the relationship between surface bonding and dispersion in model samples. XPS is the main technique used, but supplementary characterizations are made with Auger electron spectroscopy (AES), secondary ion mass spectrometry (SIMS) and secondary electron microscopy (SEM). 4.2 Experimental Planar alumina supports were prepared by heating pure aluminum panels (0.5 cm x 1 cm) in compressed air at 200°C for 30 min. Molybdenum oxide was applied by immersing the panels in a 0.009 M aqueous solution of ammonium heptamolybdenate (AHM), (NH4)6(Mo7024).6H20 (pH=5.3), for 1 min, and then evaporating the-samples to dryness at room temperature, followed by drying at 100°C for 30 min. The dried samples were calcined to specific temperatures (200, 350 and 450°C) for 3 h in an air flow. A large-area sample (13wt% M 0 O 3 on alumina) for comparison was prepared by pore volume impregnation of activated alumina (diameter 3.2 mm, surface area 355 m2g"', pore volume 0.5 cm3g"') by an aqueous solution of AHM (0.26 M), followed by drying at 100°C for 12 h. Photoelectrons were collected from a 2x4 mm2 area and measured in MAX 200 using MgKoc radiation from a dual anode source operated at 15 kV and 20 mA. XPS spectra for elemental identification were measured with the analyzer pass energy set at 192 eV. The same setting was used for observations of the Al(KLL) Auger peak excited by bremsstrahlung radiation, again using the MgKoc source (this allowed systematic study without interference by Al plasmon loss peaks). A pass energy of 48 eV was used for measuring Cls, Mo3d and A12p 74 spectra at higher resolution. Guides to the relative elemental percentages probed by the photoelectrons were estimated by integrating peak areas after standard background subtraction and correction for the instrumental sensitivity factor. Binding energies were referenced to the Au4f7/2 peak at 84.0 eV, although on thicker oxides (e.g. where no metallic Al peak could be detected by XPS), the Cls peak at 284.7 eV was used as a secondary reference (this ensured consistency with the value measured on samples for which metallic Al could be detected). SIMS was performed with a VG MM12-12S quadrupole mass spectrometer using 5 keV Xe+ primary ions (current density 3 nAcm"2) for measurements in the static SIMS (SSIMS) mode, while imaging SIMS used a VG MIG 100 Ga ion gun (current density 6.5 nAcm" ). The latter gun was also used to induce the SEM images which were measured with a scintillation/photomultiplier detector. 4.3 Results The A12p spectrum of the planar A10x support consists of a peak at 72.9 eV, characteristic of the Al metal bulk, and a broader peak at 75.5 eV which is expected for a amorphous oxide layer [151,152]. From the attenuation of the metallic peak, the oxide thickness was estimated at about 48A; this uses the approach indicated with Eqs. (2.6) and (2.7) while assuming a mean free path (A,) of 23 A for the A12p photoelectrons [120]. 4.3.1 Planar and real catalysts Figure 4.1 compares Mo3d spectra of a planar model catalyst and the 13% Mo0 3 large-area sample, both uncalcined. In each case the Mo3ds/2 binding energy of 232.7 eV is consistent with the indications of Mo(+6) (e.g. 232.7 eV by Zingg et al. [62] and 232.5 eV by Spevack and Mclntyre [52] after referencing to Cls at 284.7 eV), but the peaks are different when examined in detail. The spectral peak from the large-area sample is less symmetrical and broader (FWHM 75 _ l I I I I I I I I I I I I I 1 I I 1 I 1 I I I I 240 238 236 234 232 230 Bind ing Energy (eV) Figure 4.1 Mo3d spectra from two samples after drying at 100°C: (a) 13 wt. % M 0 O 3 on high surface area alumina, arid (b) Mo03/A10x model planar catalyst. 76 values are 1.6 and 2.8 eV for the planar and large-area samples respectively), and the Mo signal is reduced by a factor of seven compared with the planar sample. Even though the Mo loading is believed to be comparable, the greater diffusion of Mo into the pore structure of the large-area support inevitably results in the reduced XPS signal. It is evidenced that the planar catalyst experiences less charging than the large-area sample. 4.3.2 Dispersion of Mo at varied calcination temperatures Table 4.1 includes Mo/Al ratios from XPS for model catalysts treated at varied calcination temperatures. Since the Mo loading is the same for all samples, it is clear that the higher calcination temperature improves the Mo dispersion. The improved dispersion is particularly significant after heating at 450°C. The lateral Mo distribution on the planar support was assessed more directly by SIMS imaging. Figure 4.2 shows SIMS images of 9 5 Mo + after the catalyst samples were calcined to 350 and 450°C, along with the corresponding ion-induced SEM images on the same areas. SEM shows that the surface roughness is similar in each case, but the SIMS images indicate that the Mo dispersion is more uniform for the sample calcined at 450°C. Non uniform 9 5 Mo + images were obtained for calcination at 200°C, but SEM also indicated that this sample had comparable surface roughness. 4.3.3 Interfacial analysis of Mo03/A10x The positive ion SSIMS spectra (Figure 4.3) exhibit very similar secondary ion clusters for all samples. They contain Mo + (mass range 92 to 100), MoO+ (108 to 116) and Mo0 2 + (124 to 132) in addition to A l + and A10+ ions originating from the alumina support. For the uncalcined sample, and those calcined at the lower temperatures, the corresponding Mo + /Al + ratios (values of 0.24, 0.29 and 0.23 indicated for uncalcining and calcining at 200 and 350°C) 77 Table 4.1 Surface science characterizations for different samples studied in this work Sample XPS SSIMS Auger atomic intensity parameter ratio ratio (oO Mo/Al Mo + /Al + Initial AlOx/Al planar sample (A) 1461.4 eV A after calcination (450°C, 3h) - 1461.6 eV A after treatment with 1 % AHM and dried (B) 0.2 0.24 1461.4 eV B after calcination at 200°C 0.6 0.29 1461.5 eV B after calcination at 350°C 0.4 0.23 1461.4 eV B after calcination at 450°C 1.6 0.14 1462.4 eV 78 SEM Images SIMS Images Ion-induced SEM and SIMS images from three MoGVAlOx planar catalysts calcined at 350 and 450°C. 79 Figure 4.3 O GO o • rH A1 + 4 5 0 ° C A1CT Mo M o O " MoO + A) lk. l ]L . . . JLi J . . . . .^ 3 5 0 ° C 1 Ik 2 0 0 ° C A. uncalcined Ik. J J J L . JI*.. 1 1 1 1 _LLLL J_L JXLL I I I I 0 20 40 60 80 100 120 140 160 180 200 M a s s ( a . m . u . ) Positive ion SSIMS spectra measured from four Mo03/A10x planar samples: one is uncalcined, the other three have been calcined at 200, 350 and 450°C. 80 are roughly constant. However, calcination at 450°C leads to a relative enhancement in the Al + intensity, and hence to a reduction in the M'o+/Al+ ratio to 0.14. Intuitively this may appear surprising, given the other evidence above that points to a uniform layer of Mo over the high-temperature calcined sample, but it is well-known that ion yields can vary sensitively with bonding arrangement (and contribute to the so-called matrix effects [142], see Section 3.3). Heating at 450°C takes beyond the transition point for converting thin amorphous oxide to y-alumina [153]. At this time we cannot be sure to what degree this contributes mechanistically to the change in Mo + /Al + ratio, but that change is tentatively attributed here to a change in bonding between the molybdate species and the alumina support. Each sample shows the Mo3d5/2 binding energy at 232.7 eV, consistently with the presence of Mo(+6). Nevertheless, the A12p spectra show that a change occurs for the sample that has been calcined at 450°C; this peak shifts to lower binding energy by 0.5 eV compared with the other samples (Figure 4.4 (a)). That change shows the advantage of studying the model catalyst; such effects would normally be hard to detect in spectra from real catalysts. Further evidence for changes in the model catalyst sample calcined at 450°C are seen with reference to the bremsstrahlung-excited Al(KLL) Auger spectrum. For example, Figure 4.4 (b) shows structure at higher kinetic energy from the original hydroxylated Al oxide. Table 4.1 quotes values of the Auger parameter, a' for each planar sample studied. The untreated A10X/A1 sample, that heated to 450°C for 3 h (no Mo), the uncalcined Mo sample and the Mo samples calcined at 200 and 350°C, all show the same value of a', to within 0.1 eV; by contrast the Mo-treated sample calcined at 450°C has an a' value which is greater by about 1.0 eV. Again this is taken to indicate the presence of a new bonding situation for Al. 81 Figure 4.4 (a) A12p spectra and (b) Al(KLL) spectra from the four Mo03/A10x planar samples considered in Figure 4.3. 82 4.4 Discussion The conversion from an initial A10X/A1 support sample to a model catalyst after calcination is inevitably complex, and the spectroscopic observation made here can only take a broad view of the changes occurring during this chemical evolution. The surface of amorphous alumina grown by thermal oxidation is generally believed to be terminated by hydroxyl (-OH) groups [44]. These protonate at pHs below the isoelectric point (6-8) and in turn encourage the adsorption of M07O24 " (plus a minor amount of M 0 O 4 ") [50] after treatment with the AHM solution. The drying initiates dehydration, a process that continues with the calcination. This leads to coordinative unsaturation for some surface Al [154], and such sites will be available for further bonding. Neither XPS nor SSIMS provides evidence for significant differences in chemical bonding in the uncalcined sample from those calcined at 200°C and 350°C. Nevertheless the Mo dispersion is different for these two calcination temperatures; indeed the coating is less uniform at 350°C according to the SIMS images. At 200°C, the surface dehydroxylation presumably encourages the Mo dispersion, but by 350°C both the SIMS imaging and the Mo/Al ratio from XPS detect a lower presence of Mo, although with no substantive change in chemical state. This is consistent with increased Mo diffusion below the surface. The behavior of the sample calcined at 450°C is morphologically and chemically different from those calcined at the lower temperatures.. The SIMS images and the Mo/Al ratio from XPS confirm that the Mo is better dispersed on the alumina at 450°C, and similar results have been obtained in calcination studies on M0/Y-AI2O3 systems. For example, Kasztelan et al. [51] reported that the calcination treatment at 500°C modifies the Mo dispersion on real catalysts, and Edmonds et al. [155] suggested that Mo migrates from multilayer dispersion at the dried step to 83 monolayer dispersion at the calcination step. The present work supplements previous observations on real catalysts [51,155] that calcination at below 350°C can slightly enhance the Mo dispersion, and that calcination at 450°C facilitates the Mo-O-Al bonding (which in turn can greatly enhance the Mo dispersion). The further observations here for the Mo + /Al + ratios in SSIMS, as well as the shifting of the A12p and Al(KLL) structure (and hence the Auger parameter, a'), are taken to indicate the formation of a new surface species. This has been described previously as surface molybdate or as an 'interaction' species [33,51,52,62,69,73] and differs from Al2(Mo04)3 which is believed to be a sub-surface species [52,62,69,156]. For example, Spevack and Mclntyre report [52] that Mo thicknesses greater than 20 nm are required to form Al2(Mo04)3 during the calcination step, but the thicknesses on our samples appear to be less since Al signals are still detected in XPS. Also observations with static SIMS here provide direct evidence that Mo is present at the surface of the model catalyst after calcination. Using Wang and Hall's data for y-alumina for a pH of 5.3 [50], the Mo loading for the planar sample is estimated to be about 1.4xl014 atom/cm2; this value is in the medium range for which surface molybdate (tetrahedrally or octahedrally coordinated) has generally been believed to form [62,69]. In summary, observations made here on model catalysts, for changes in the surface chemistry and Mo dispersion with calcination temperature, should have relevance to effects operating at surfaces of real supported Mo-Al 20 3 catalysts. Specifically, this work provides evidence to support the hypothesis that Mo-O-Al surface linkages are formed, and that this in turn provides the driving force for the enhanced Mo dispersion during the 450°C calcination process. 84 5. Sulfidation of Thin Films of M 0 O 3 on a A ! O x Planar Support 5.1 Introduction In the previous chapter, it is hypothesized that direct Mo-O-Al surface linkages are formed during the 450°C calcination process for a Mo03/A10x system, and that this provides the driving force for the enhanced Mo dispersion. The next question that comes up is to ask how this surface linkage affects the subsequent sulfidation step. Many publications have reported on the sulfidation of M0/AI2O3 systems (with and without promoters), including catalytic activity studies [91,92], surface composition analyses [94,157], and mechanistic studies [97,158]. Models have been proposed to interpret the interactions between Mo, S and the support [159-161]. There is some broad acceptance that tetrahedral molybdate is only partially sulfided, whereas octahedral molybdate is transformed to MoS2, in microslab form [94,97], with defect centers that have been postulated as involving Mo(+3) species [88], although there is little direct supporting evidence. Understandings of the processes involved during the sulfiding of Mo0 3/Al 203 seem to have been even less clear than those for calcination. For example, there are conflicting ideas for the preservation of the monolayer Mo coverage and the Mo-O-Al linkages after sulfidation [70,69,97]. Okamoto et al. [69] as well as Tops0e and Tops0e [70] reported breaking of Mo-O-Al bonds and aggregation of Mo atoms, while Arnoldy et al. [97] believed the Mo monolayer coverage and Mo-O-Al linkages are preserved after sulfidation. IRS has been especially used in this area [69,70,94], while the application of XPS has been limited, except for analyzing the final sulfidation products [94,162]. Additionally, most earlier works simply aimed to identify any associations between the sulfidability and the Mo loadings of the catalyst samples [94,97], and interest in the relationships between calcination and sulfidation has been limited. 85 In this chapter, we used model catalysts developed in the study described in Chapter 4 to assess their sulfidability and to relate to changes in Mo dispersion and Mo-A10x interaction. High-area catalysts, which were treated as similar as possible to the model catalysts, were used in initial tests on the HDN reaction for pyridine. With the assumption that the local structures on the high-area and model catalysts are similar, we tried to relate changes in activity for the high-area catalysts to XPS observations on the model catalyst samples. 5.2 Experimental 5.2.1 Sulfidation of planar catalysts The model Mo03/A10x catalysts (uncalcined, calcined at 200, 350 and 450°C) were prepared as described in Chapter 4. These samples were sulfided in a stream of 2% H2S/H2 (50 mL/min) for 1 h while held at 350°C. This was done in a U-shaped stainless reactor (internal diameter of 1/8 inch) as shown in Figure 5.1. Immediately after the sulfidation, the catalyst samples were cooled to room temperature in flowing N 2 . The reactor was then introduced to a N2-filled glovebox and the samples transferred to the XPS facility without air exposure. 5.2.2 Catalytic activity studies Large-area catalyst samples were used in the HDN activity studies. 13% M 0 O 3 / A I 2 O 3 catalysts were prepared through pore volume impregnation of spherical Y-AI2O3 (diameter 3.2 mm, surface area 355 m2/g, pore volume 0.5 cm3/g). The impregnated catalysts were dried in air and then heated at 100°C for 24 hours. Calcination treatments were performed at 200, 350 and 450°C respectively. Catalytic activity tests were performed in the same reactor as used for 86 Pyridine injection H2 S / H2 ^ NH 3 > H 2 > o 2 > He —>-N 2 > Gas manifold ] • a sample collecting tube cold trap Tubular furnace Figure 5.1 Reactor used showing the gas manifold for sulfidation and HDN reaction, the cold trap for HDN reaction products and the furnace (temperature controlled by a variable transformer). 87 sulfidation studies. The calcined catalysts (5 g) were introduced in the reactor and treated at 350°C with a flowing 2% H 2S/H 2 mixture (50 mL/min) for 3 h. The H 2 S/H 2 gas was then stopped and replaced by a H 2 flow. Subsequently 10 (iL of pyridine was injected into the gas line and carried into the reactor by the H 2 flow. The products and non-reacted pyridine were collected in a solid C 0 2 cold trap (see Figure 5.1). After the reaction (30 min), the collected liquid was sent for analysis by gas chromatography and mass spectrometry (GC/MS) in the Department of Chemistry Mass Spectrometry Laboratory. Blank tests could not detect any conversion of pyridine by the empty stainless steel reactor. 5.2.3 XPS measurements The XPS measurements were made with the same settings as those detailed in Chapter 4. The binding energies in the spectra from model catalysts considered in this work were referenced to the Au4f7/2 peak at 84.0 eV, provided the metallic Al component could still be detected from the substrate. For other samples, the effects of any charging were compensated by using the secondary reference, namely the Cls peak from adventitious carbon, which was set at 284.7 eV. For referencing purposes, a natural molybdenite (MoS2) single crystal was analyzed by XPS, and the Mo3d and S2p spectra are shown in Figure 5.2. The measured binding energies of Mo3d5/2 (FWHM=1.0 eV) and S2p (FWHM=0.9 eV) are at 228.7 and 161.5 eV respectively, while the S/Mo ratio is indicated to be about 1.8. Measured Mo3d spectra from the uncalcined and calcined Mo03/A10x samples, prior to sulfidation, indicate that the only Mo species on the support corresponds to the (+6) oxidation state [62,95] and for this the Mo3d5/2 binding energy is measured at 232.7 eV. After sulfiding, various Mo and S species were detected, and the Mo3d and S2p spectra were fitted with 80% Gaussian/20% Lorentzian peak shapes with spin-orbit splittings of 3.2 and 1.2 eV, respectively. 88 Figure 5.2 (a) S2p and (b) M o 3 d spectra measured from a natural M0S2 (molybdenite) single crystal. 89 Details of the components used were guided in the first instance by previous reports of relevant binding energies for Mo3d5/2 and S2p species that have been identified in reduced and sulfided M0O3 samples. Heating M0O3 in the presence of hydrogen inevitably gives some reduction. For the conditions used in the sulfidation with H2S in this project, the oxidation states Mo(+6), Mo(+5) and Mo(+4) state are particularly involved, and values of Mo3d5/2 binding energies used for the interpretations (both now and in other work) are listed in Table 5.1. Specific results are quoted for the components M0O3, M0O2 and M0S2, but other chemical states are possible even with the particular oxidation states noted. For example, others have reported the existence of a Mo3d5/2 component whose binding energy is ~1 eV greater than that for M0O2, but is still apparently associated with the (+4) oxidation state of Mo. This chemical state probably involves bonded OH groups, for which an increase in binding energy is commonly observed even without a change in formal oxidation state [24]. Information is also given in Table 5.1 for different S2p components in relation to interpretation of the sulfidation process. The S2p spectra for the sulfided thin films often contain one or two other S components in addition to the sulfide (S2~) identified from M0S2 at 161.5 eV. Peaks at -162 and -163 eV are taken to originate from disulfide and polysulfide respectively [43,95]. Two other species are noted for comparison: elemental S has the S2p binding energy at 164.1 eV [119], while a lower binding energy form has been reported for non-stoichiometric MoS2.xat 161.0 eV [95]. 5.3 Results 5.3.1 Sulfidation of Mo03/A10x samples The A12p spectrum from a planar A10x support has been described in Section 4.3.1. In summary, two chemical states are present, namely metallic Al at 72.9 eV and amorphous Al 90 Table 5.1 Reported Mo3d5 / 2 and S2p binding energies relevant to reduced and sulfided M0O3. Example M o 3 d 5 / 2 * S2p* Ref. Mo(0) 227.8 [95] ' 227.7 [119] Mo(+4) M0S2 228.9 161.7 [95] 229.0 161.8 [59] 228.4 161.2 [43] 228.7 161.5 this work Mo0 2 229.0 [95] 229.3 [59] 229.4 [43] • 229.9 [95] 229.8-230.6 [163] Mo(+5) 231.0 [95] 230.9-231.4 [59] 230.6-231.9 [163] Mo(+6) M0O3 232.5 [95] 232.6 [62] 232.7 this work M0S2-X 161.0 [95] disulfide: S22~ 162.1-162.5 [43] 162.4 [95] polysulfide: Sn2" (n>2) 163.2 [95] elemental S 164.1 [119] * Binding energies are referenced to CIs peak at 284.7 eV (except Mo(0)). 91 o oxide at 75.4 eV. The thickness of the oxide layer is estimated to be about 48 A; this used the approach indicated with Eqs. (2.6) and (2.7) (in Section 2.3.3) as well as a mean free path of 23 A for the A12p photoelectrons [120]. After treating with H 2S/H 2 at 350°C for 1 h, no S could be detected by XPS on the support and hence this support appears basically inert to H 2S/H 2 for these reaction conditions. Table 5.2 lists values of the Auger parameter (a') for the model M o C V A l O x catalyst samples before and after sulfiding. For the samples which are uncalcined or calcined at 350°C or less, a' remains essentially unchanged after the sulfidation, and is always close to 1461.4 eV (uncertainties in a' values are at the 0.1 eV level). For the sample calcined at 450°C, a' decreases from 1462.4 to 1461.4 eV after sulfiding by H 2S/H 2. This may suggest a rupture of the Mo-O-Al linkages during the sulfidation, given the earlier conclusion (Section 4.4) that higher a' appears to correlate with the presence of direct Mo-O-Al linkages. This conclusion is consistent with reports by Okamoto et al. [69] and by Tops0e and Tops0e [70]; these sets of authors used IRS to examine the sulfidation of high-area Mo03/Al 2 0 3 catalyst samples. They observed the regeneration of OH groups, which had previously been consumed by interaction with molybdate during the calcination at 500°C. Table 5.2 also compares the Mo/Al atomic ratio according to XPS for the various Mo03/A10x samples before and after the sulfidation As mentioned in Chapter 4, better Mo dispersion is found on the sample calcined at 450°C, and that has been attributed to the formation of the Mo-O-Al linkages during the calcination process. The subsequent sulfidation, however, decreases the Mo/Al ratio in each case, and that is most pronounced for the 450°C calcined sample (decreases from 1.62 to 0.23). This observation is consistent with Portela et a/.'s work [162]. They used XPS to study sulfided high-area Mo/y-Al203 and Co/Mo/y-Al203 catalysts and reported that the Mo/Al ratio decreases on sulfidation. From extended x-ray absorption fine 92 Table 5.2 Characterizations for model catalysts before and after sulfidation at 350°C (see text). Sample Auger Parameter (a') Mo/Al Before After Before After A10X/A1 support 1461.3 1461.4 - -uncalcined Mo03/A10x 1461.4 1461.4 0.16 0.12 200 °C calcined Mo03/A10x 1461.6 1461.4 0.58 0.21 350 °C calcined Mo03/A10x 1461.4 1461.3 0.41 0.23 450 °C calcined Mo03/A10x 1462.4 1461.4 1.62 0.23 93 structure spectroscopy (EXAFS), Clausen et al. [98] concluded that during sulfidation Mo atoms in the calcined catalyst are converted from a patchy overlayer form to MoS2-like structures, which are ordered in very small domains. Here, observations of the Auger parameter are consistent with rupture of the Mo-O-Al bonds after sulfidation, and that can subsequently lead to agglomeration of Mo on the support. Table 5.3 lists binding energies used as starting points for the Mo3d and S2p curve fitting analyses done in this work. Table 5.4 collects new XPS results for sulfiding the four Mo03/A10x samples considered, namely the uncalcined form plus those calcined at 200, 350 and 450°C prior to the sulfidation. The Mo3d and S2p spectra from each of these samples are shown in Figures 5.3 and 5.4, and the corresponding binding energies for the components after the curve fitting analyses are listed in Table 5.4. After sulfidation, all samples show a major component with a Mo3d5/ 2 binding energy of 228.7±0.3 eV. This is interpreted as arising from a Mo(+4) sulfide, but all samples show additional Mo(+4) or Mo(+5) species corresponding to binding energies at 230.1±0.1 eV and 231.3±0.1 eV, respectively. The major component in the S2p spectra from most of the sulfided films is the S " peak at 161.6±0.1 eV. Additional components are seen at the higher binding energies of 162.2±0.2 eV and 163.3 eV, and these are interpreted as corresponding to S22~ and Sn2", respectively. No attempt has been made to give a closer interpretation. It is known that quite complex sulfide structures can occur with Mo [164], and other techniques including vibrational spectroscopy would be needed to identify these structures further. Results for the sulfidation of the four Mo03/A10x samples, to see the effects of the precalcination, can be summarized as follows: 94 Table 5.3 Binding energies used as starting points for XPS curve fitting analyses for MoCVAlOx samples after sulfidation. Chemical state Binding energy (eV) Mo3d5/2: Mo(+4) sulfide, disulfide, polysulfide 228.7 Mo(+4) with bonding to O/OH 230.0 Mo(+5) 231.0 Mo(+6) as in M0O3 232.7 S2p3/2: . MoS2-x 161.0 S2" 161.5 S22" 162.0 S„2"(n>2) 163.0 95 Table 5.4 Binding energies (in eV) and relative proportions of different components in Mo3d and S2p spectra from MoOVAlOx samples after sulfidation. These samples have been calcined at different temperatures. Calcination temperature Mo3d5/2 S2p3/2 S/Mo None (uncalcined) Mo(+4) sulfide 228.5 (82%) Mo(+4) with OH 230.1 (11%) Mo(+5) 231.3 ( 7%) S2" 161.5(80%) S22~ 162.4(20%) 2.2 200°C Mo(+4) sulfide 228.6 (89%) Mo(+4) with OH 230.2 ( 6%) Mo(+5) 231.4 ( 5%) S2" 161.5(100%) 1.8 350°C Mo(+4) sulfide 229.0 (78%) Mo(+4) with OH 230.1 (13%) Mo(+5) 231.4 ( 9%) S2" 161.5(23%) S22" 162.0(64%) Sn2' 163.3 (13%) 1.3 450°C Mo(+4) sulfide 228.8 (86%) Mo(+4) with OH 230.1 ( 9%) Mo(+5) 231.3 ( 5%) MoS2-x 161.1 (18%) S2' 161.7(62%) S22" 162.4 (20%) 1.8 96 238 234 230 226 222 Binding Energy (eV) Figure 5.3 Mo3d spectra from sulfided (a) uncalcined catalyst, (b) 200°C calcined catalyst, (c) 350°C calcined catalyst, and (d) 450°C calcined catalyst. 97 168 166 164 162 160 158 Binding Energy (eV) Figure 5.4 S2p spectra from sulfided (a) uncalcined catalyst, (b) 200°C calcined catalyst, (c) 350°C calcined catalyst, and (d) 450°C calcined catalyst. 98 (1) The simplest result appears for the sample calcined at 200°C insofar as all S is indicated to be in the S2" form. The S/Mo ratio of 1.8 is consistent with observations for MoS2 single crystal, although the Mo spectrum does indicate that other Mo chemical states are involved. (2) Formally the uncalcined sample appears to show a larger S/Mo ratio (2.2) than the other samples. The sample calcined at 450°C gives a value of 1.8 for this ratio (just as for the sample calcined at 200°C), but that calcined at 350°C shows a relatively low ratio (1.3). It is not clear whether this latter value represents a total lower amount of S present, or whether this ratio appears low as a result of the S being distributed differently in the sample. (3) Larger amounts of the non-sulfided Mo(+4) and Mo(+5) chemical states appear to be present on the uncalcined (18%) and the 350°C-calcined (22%) forms, while these amounts for the samples calcined at 200°C (11%) and 450°C (14%) are indicated to be less. Compared with the first pair of samples, the latter pair therefore appear to give relatively more complete reductions and sulfidations in the approach to M0S2 from the initial M0O3. (4) The component attributed to Sn2~ was only detected for the 350°C-calcined sample (and in the relatively small amount of 13%), while that for the S species at 161.1 eV binding energy was just seen for the 450°C-calcined sample. The latter species was observed by Spevack and Mclntyre [95] in their work on the sulfidation of planar supported M0O3 thin films, and it has also been reported on the basal plane of M0S2 after ion bombardment [85,87]; this species is described as corresponding to non-stoichiometeric M0S2-X. 99 5.3.2 Catalytic activity studies High-area catalysts, prepared as in Section 5.2.2, were used to test the effects of treatment on activity for HDN reaction of pyridine (C5H5N). The hope was to find evidence for pentane (C5H12) in the products, although this was not seen in the tests to date. Blank reactions were run on the Y-AI2O3 support with no added Mo and on the 450°C-calcined Mo03/Al203 sample which had not been sulfided; trace amounts of C4H5N were detected in each of these experiments but there was no evidence for a HDN reaction. Results of the GC/MS analyses for the sulfided catalysts are in Table 5.5. HDN reaction occurs by dimerization, insofar as some pyridine is converted to C10 compounds, although only the sulfided 450°C-calcined catalyst gave a small amount of cyclopentane (C5H10) as well. Sonnemans et al. [165] studied the denitrogenation of piperidine on H2-reduced CoO/Mo03/Al203 and reported that the product composition had a strong dependence on H 2 pressure. At low pressure (e.g. 1 atm), significant amounts of C10 compounds result, while at higher pressure of H2 (e.g. 60 atm) pentane is the major product. Kherbeche et al. [166] criticized the conclusions from the work of Sonnemans et al. as not having relevance to practical HDN conditions because the M0O3/AI2O3 catalyst was only H 2 reduced. Kherbeche et al. believed that the production of C10 compounds is due to incomplete sulfidation of the catalysts. We have no further evidence at this time, but the current work does suggest that the sulfided 450°C-calcined catalyst, as formed here, has the highest activity for the HDN reaction of pyridine because it is the only sample studied which gives some cyclopentane as a product. 5.4 Discussion At the outset it must be emphasized that the work in this chapter can only be seen as a preliminary study, and that it has considerable scope for refinement. The ultimate objective is to 100 Table 5.5 Products obtained from sulfided high-area catalysts during the HDN of pyridine. Catalyst Products Sulfided Y-A1203 trace amount of Cio compounds Sulfided Mo03/A10x: uncalcined C10H14 and C10H16 calcined at 200°C C10H14 and C10H16 calcined at 350°C C10H14 and C10H16 calcined at 450°C cyclopentane, C10H14, C i 0 Hi 6 and C10H22 101 establish relations between surface activity for HDN reaction on high-area catalyst samples and results from surface characterization of model catalysts prepared under comparable conditions to the high-area forms. Although to date studies of HDN reaction of pyridine did not produce the hoped-for product (i.e. pentane C5H12), nevertheless this work still indicates that the sample that had been calcined at 450°C prior to sulfidation provided the best activity. Two special features are apparent on the corresponding model catalyst. First, this sample appears to give the best Mo dispersion, as well as the most direct Mo-O-Al bonding, prior to sulfidation (Chapter 4). Second, after sulfidation, this is the only sample that seems to require a component species designated as M0S2-X in the curve fitting for the Mo3d spectra. The nature of the MoS2.x species is still not completely clear, but it has been reported on ion-bombarded M0S2 single crystal samples [85,87]. Lince et al. [85] believed it has an amorphous structure with considerable S depletion compared to M0S2. From our HDN activity studies, this species may be associated with active sites on the sulfided catalyst and it possibly provides unsaturated Mo for the reaction. Although such unsaturated Mo atoms should have a lower oxidation state than +4 [88,95], no such Mo species was detected in this work. However its identification may be restricted by low concentration and by an overlapping S2s peak. Indeed, the use of the monochromatized x-ray source may help to distinguish this species from the major Mo(+4) sulfide peak (228.8 eV). In addition to M0S2-X, some S22" was detected on the sulfided 450°C-calcined sample. Lince et al. [85] also reported S22" is formed during the Ne+ bombardment of single-crystal M0S2. When a Mo atom is removed from the M0S2 lattice, it leaves three neighboring S atoms with dangling bonds. Since these S atoms are now bonded to two Mo atoms rather than three, it may appear natural for two of the S atoms to rearrange and bond to each other with formation of the (S-S)2" species. Although the formation mechanism is unclear, S22" is apparently involved in the 102 disordered MoS2 lattice, which may also provide sites for the HDN reaction. Polz et al. [167] also detected S22" in sulfided Mo0 3 /Al 2 0 3 high-area catalysts by Raman spectroscopy, and these authors believed that this species plays a role in hydrogen splitting on the sulfided catalysts. In fact, the present work indicates a high amount of S22" on the sulfided 350°C-calcined sample, and follow up studies are needed to assess the importance of this species to HDN activity. Differences in sulfidations of the four Mo03/A10x samples, including the formation of MoS2_x and S22~ on the 450°C-calcined sample, may relate to the different Mo dispersion and Mo-A10x interaction prior to sulfidation. From results established in Chapter 4, our calcined samples can be divided into two groups: the low Mo dispersion samples (i.e. the uncalcined and 350°C-calcined samples) and the higher Mo dispersion samples (i.e. the 200°C- and 350°C-calcined samples). For the former group, Mo0 3 can be considered scattered in cluster form on planar A10x, and this may impede the H 2 and H2S diffusion and reduction compared with the situation on the higher dispersed samples (i.e. those calcined at 200 and 450°C). The samples with poorer dispersion are indicated (Table 5.2) to have less Mo(+4) sulfide after the H 2/H 2S treatment than those with the better dispersion. It was concluded in Chapter 4, that the 200°C calcination process does not develop any direct Mo-O-Al bonding between Mo0 3 and A10x, but it does improve the Mo dispersion on the support. The higher Mo dispersion does not limit the interaction with H2S and H 2 , and O-S displacement occurs readily on this sample. Hence most Mo0 3 on the 200°C-calcined sample can readily convert to MoS2. Other observations in Chapter 4 suggests that the Mo is dispersed most evenly on the support after calcining at 450°C. Nevertheless, compared with the sulfided 200°C-calcined sample, slightly less Mo(+4) sulfide is formed. Therefore, it appears that the sulfidability of the 450°C-calcined catalyst is not only governed by the Mo dispersion, but 103 possibly the Mo-A10x interaction, developed during the 450°C heating process, may hinder the Mo03-to-MoS2 conversion at the sulfidation stage. Plausibly this effect could influence the formation of M0S2-X and possibly S22" are the special species on the sulfided 450°C-calcined sample. Scheffer et al. [96] studied the sulfidation of M0O3/AI2O3 high-area catalysts with the temperature-programmed sulfiding (TPS) technique, and they reported that the sulfidability of high-area M0O3/AI2O3 catalysts depends both on the Mo dispersion and the Mo-Al203 interaction. These authors concluded that the sulfidation for catalysts with low Mo dispersion is mass-transfer limited, whereas on the high Mo dispersion catalysts, the sulfiding rate is mainly influenced by the M0-AI2O3 interaction. Observations reported here seem broadly consistent with this statement. 104 6. Study of the Reduction of M0O3 Thin Films by NH 3 6.1 Introduction As mentioned in Section 1.4.5, attention has recently been drawn to the molybdenum nitrides because of their promising activity and selectivity for HDN reactions [19,21,168], but there has been considerable discussion about optimal methods to synthesize high surface area catalysts [106-108]. Starting from 1980s, Volpe and Boudart [106,107] proposed preparing suitable Mo 2N samples by linearly heating M0O3 with NH3 in a temperature-programmed reaction (TPR) to the 1000-1200 K range. But later work [21,108] suggested that the particular preparation conditions influence the catalytic properties of the nitrides; for example, Choi et al. [21] reported that larger M02N surface areas are formed when the catalyst is prepared with both higher flow rates of NH3 and lower heating rates. Efforts to elucidate the associated surface chemistry have used AES [169] and XPS [168,109], but information on the relationships between catalytic reactivity and surface structure is still very limited. Indeed a popular characterization technique for Mo nitrides in this context has been x-ray diffraction (XRD) [21,106-108], although its identifications inevitably restrict to bulk crystalline phases. The objective of the research described in this chapter is to use XPS to study the reaction pathways of M0O3 and NH3 by monitoring surface Mo species at different stages in the nitridation process. We choose to study a thin film of M0O3 on Mo metal plate (Mo03/Mo) feeling that a planar M0O3/M0 sample provides better morphological conditions for surface characterization, and secondly it eliminates the importance of the NH 3 flow rate that is a factor with powdered M0O3 samples. In situ H and N plasma reactions of M0O3/M0 were also performed to compare with the NH3 nitridation process. 105 6.2 Experimental The starting samples of M0O3 on a planar metallic substrate were prepared by heating Mo metal strips (99.95% purity, 1x0.25cm2) at 250°C in compressed air for 4h. The nitridation reaction was carried out in the U-shaped stainless reactor described in Chapter 5 (see Figure 5.1) with anhydrous NH 3 flowing over the various M0O3/M0 samples. The reactor was heated in a tube furnace with a variable transformer to control heating rate, and the stages of heating are summarized as follows: In the first stage of treatment the M0O3/M0 sample was heated in NH3 with the temperature raised from room temperature to 350°C in 30 min. The second stage involved increasing the temperature to 450°C over a 1 h period (this corresponds to the "high heating rate" used by Choi et al. [2]; their "low heating rate" was done at 40°/h). In the third stage, the temperature was increased from 450 to 700°C at a heating rate of 200°/h, and the final stage involving holding the sample at 700°C in NH3 for an additional 1 h. Nitrided samples after each stage of treatment were removed from the reactor, cooled rapidly to room temperature in flowing NH3, and stored in a nitrogen box before transferring to the XPS facility. Two of the initially prepared M0O3/M0 samples were subjected to in situ plasma treatments carried out in a preparation chamber (Figure 6.1) with base pressure 5xl0"8 torr. For these treatment flowing H2 (or N2) gas was maintained at 1.5 torr, and the H (or N) plasma was generated in a quartz tube located at the top of the chamber. The sample was held approximately 25 cm from the discharge region, which was excited by a microwave generator operated at 2.45 GHz with a total power of 60 W. Both M0O3/M0 samples were treated in the H plasma for 10 min: subsequently one was heated at 220°C for 10 min, and the other was treated with the N plasma for 10 min. After the plasma treatments, these samples were individually transferred by rods through the vacuum system to the XPS analyzing chamber. 106 To preparation chamber #1 r^ri =L^=<«— S A S I N (H2> N 2) Quartz tube V To pump Sample Transfer rod Figure 6.1 Schematic diagram for the plasma treatment chamber. 107 The XPS measurements were made with the same settings as those detailed in Chapter 4 but a pass energy of 96 eV was used for measuring Mo3p, Mo3d, Ols, Nls and Cls spectra at higher resolution in this work. Guides to the relative elemental percentages probed by the photoelectrons were estimated by integrating peak areas after correction for the instrumental sensitivity factor. Curve fittings assumed the standard non-linear background subtraction, and used component peaks based on 80% Gaussian, 20% Lorentzian shape. The component peaks were chosen with full regard to the chemistry involved at each step in the study. All binding energies were fixed with respect to the Au4f7/2 peak at 84.0 eV, which was measured from a small reference Au sample on the main holder; there was no evidence for any charging problems in the M0O3/M0 samples studied here. 6.3 Results 6.3.1 Plasma treated M0O3 thin film Figure 6.2 (a) shows the Mo3d spectrum measured from the initially-prepared M0O3/M0 film. The main Mo3ds/2 component with a binding energy at 232.7 eV is consistent with Mo(+6), but a small amount of Mo(0) is still detected at 228.0 eV; the intensity ratio indicated an oxide layer thickness of about 50A (assuming a mean free path of 25 A [120]). Since the accepted binding energy for pure Mo metal is 227.7 eV [119], it is assumed that this Mo(0) component corresponds to metal with a small amount of dissolved oxygen. The corresponding spectrum, in Figure 6.2 (b), after exposure to the H plasma shows substantial reduction to lower oxidation states, and the main Mo3ds/2 peak shifts to about 230 eV. The subsequent treatment with the N plasma gave some regeneration of Mo(+6), but interestingly no N was detected at this stage. Heating of the sample to 220°C, after the H plasma exposure, yields only small changes to the Mo3d spectrum (Figure 6.2 (d)), although the change is consistent with some oxidation. 108 (c) N plasma treated Figure 6.2 Mo3d spectra from (a) the initially-formed M0O3 thin film; (b) the film formed after treating that in (a) with the cold H plasma for 10 min; (c) the film formed after treating that in (b) with the cold N plasma for 10 min; and (d) the film formed after heating that in (b) at 220°C for 10 min. 109 Compared with the spectrum in Figure 6.2 (b), that in Figure 6.2 (d) shows a slight reduction in the peak at -230 eV, and a corresponding increase in structure related to Mo(+6). The reduction/oxidation processes for the Mo species, seen in the Mo3d spectra, can also be observed in Ols as well (Figure 6.3). In a basic interpretation, components in Ols at -532 eV are generally attributed to OH (i.e. metal hydroxide), while those at -530 eV are related to contributions from O2" [170]. Below it is noted that the latter value may vary a little with the oxidation number of the metal to which it is bonded. At this point we just recognize, for the initially prepared M0O3/M0 thin film (Figure 6.3 (a)), that the Ols component at 530.5 eV, similar to the situation reported by Spevack and Mclntyre [171], corresponds to O2" bonding to Mo(+6). The tail in this spectrum at the higher binding energy is believed to associated with C bonded O contamination (C-O-C or G=0). A small amount of hydroxide may be present, but this increases substantially with the H plasmas treatment (component at -532.2 eV in Figure 6.3 (b)), which as noted above also gives an effective reduction for the metal. The subsequent exposure to the N plasma results in a reduction in the Ols component at -532 eV, compared with that at -530 eV, and this is consistent with N removing H by in effect converting some OH" to O2"; correspondingly there is some regeneration in the Mo(+6) state. Heating to 220°C after the H plasmas treatment shows a small change in the Ols spectrum consistent with loss in OH (Figure 6.3 (d)). The initial M0O3/M0 film shows an O to Mo ratio, estimated from peak areas after background correction, equal to 3.2; this increases to 3.9 after the H plasma treatment and then to 4.3 after the N plasma treatment. These increases probably result from small amounts of H 2 0 present in the plasma gases, as well as contributions desorbed from the walls of the preparation chamber. Heating after the H plasma treatment decreases the O/Mo ratio from 3.9 to 3.4, and 110 Figure 6.3 Ols spectra from: (a) the initially-formed M0O3 thin film; (b) the film formed after treating that in (a) with the cold H plasma for 10 min; (c) the film formed after treating that in (b) with the cold N plasma for 10 min; and (d) the film formed after heating that in (b) at 220°C for 10 min. Ill this is consistent with some 2 OH" -> O2" + H 2 0 conversion, followed by the water being pumped away. 6.3.2 Reaction of M0O3/M0 with NH 3 Nls spectra Spectra in Figure 6.4 have overlapping Mo 3p3/2 and N Is components from samples at different stages of nitridation; the N Is spectra were obtained by a curve fitting process that assumed the relative percentage of each Mo species can be fixed with reference to the corresponding Mo 3d spectrum. After heating to 350°C, three kinds of N were detected, and they have binding energies at 400.0, 398.9 and 397.2 eV (the relative proportions of components in the analysis are included in Table 6.1). The first of these is believed to originate from adsorbed NH 3 since this appears consistent with observations for NH 3 on other transition metals [172,173]. Lower values (390 to 398 eV) have been indicated for NH 2 (or NH) species, although CH 3 NH 2 adsorbed on oxidized Mo(100) has been reported to have a N Is binding energy at 400.0 eV (referenced to Mo3d5/2 at 227.7 eV [174]). The N component observed here at 397.2 eV appears characteristic of nitride, and in this we follow Choi et al.'s conclusion for a Mo nitride at 397.4±0.2 eV (referenced to C Is at 284.7 eV) [175]. Other transition metal nitrides in the literature include that for surface W nitride (at 397.4 eV referenced to Au 4f7/2 at 83.8 eV) as reported by Egawa et al. [173]. At this time we cannot be sure to what degree the N Is component at 397.2 eV is indicative of pure nitride, as opposed to having contributions from oxynitride, but we note the observation by Lyutaya [176] that at 600°C Mo0 3 reacts with NH 3 to transform into Mo0 2NH 2 , and then into oxynitride at higher temperature. Our observations appear to fit such a trend, and therefore we feel that the N Is component at 398.9 eV is likely due 112 (d)700°C-l h (c) 450-700° C Figure 6.4 The overlapping Mo3p3/2 and Nls from the M0O3 film after heating in NH 3: (a) from 25 to 350°C; (b) from 350 to 450°C; (c) from 450 to 700°C; and (d) at 700°C for lh. The Nls and Mo3p components are shown by dotted lines and continuous lines respectively. 113 Table 6.1 Binding energies (in eV) and relative proportions for the different components identified in the Nls, Ols and Mo3d spectra measured from the M0O3 thin film after initial preparation and after subjecting to various stages of treatment in NH 3. Nls Ols Mo3d5 /2 M0O3/M0 (initial) - 530.6 (76%) 228.1 (12%) 532.1 (19%) 230.1 ( 6%) 533.7 ( 5%) 232.7 (82%) 1st stage (25 - 350°C) 397.2 (54%) 530.5 (63%) 228.1 (16%) 398.9 (26%) 532.2 (25%) 229.5 (11%) 400.0 (20%) 533.6 (12%) 230.3 (29%) 231.0 (10%) 232.3 (34%) 2nd stage (350-450°C) 397.1 (91%) 530.1 (65%) 228.9 (55%) 398.9 ( 9%) 532.0 (27%) 230.0 (21%) 533.6 ( 8%) 231.3 ( 9%) 232.7 (15%) 3rd stage (450-700°C) 397.3 (100%) 530.5 (46%) 228.4 (37%) 532.0 (33%) 229.0 (26%) 533.5 (11%) 230.0 (21%) 534.7 (10%) 231.1 ( 5%) 232.7 (11%) final stage (700°C, lh) 397.3 (100%) 530.5 (45%) 228.5 (42%) 532.0 (46%) 229.1 (25%) 533.6 ( 5%) 230.0 (18%) 534.8 ( 4%) 231.1 ( 5%) 232.7 (10%) 114 to the formation of MoOx(NH2)y (or MoOx(NH)y). As the nitridation reaction procedes, the 397.2 eV structure dominates in the N Is spectrum, but at the second nitridation stage, for instance, only -9% of NH 2 (or NH) species was detected, and none was found afterwards. These observations suggest the MoOx(NH2)y and/or MoOx(NH)y species may be precursors for the final nitride (or oxynitride) product. Ols spectra Figure 6.5 compares O Is spectra for the oxidized Mo samples at various stages of nitridation. For the initially oxidized sample, the main peak at 530.6 eV is consistent with Mo in the +6 oxidation state [119,171] and hence the surface is concluded to be dominantly M0O3. Upon nitridation there are changes, although the components with higher binding energies, around 533.6 and 534.8 eV, appear due to contamination (e.g. from C-0 species). The main peak at 530.6 eV remains unchanged on treating in NH 3 at 350°C (Figure 6.5 (b)), but it shifts to 530.1 eV after the treatment at 450°C (Figure 6.5 (c)). Previously Griinert et al. [163] reported that heating M0O3 in flowing Ar, over a range in temperature, yielded a mixture of reduced oxidation states with the O Is peak shifting to 0.6 eV lower than that of the starting material, and others indicated that the O Is binding energy for Mo0 2 is 0.5 eV less than for M0O3 [171,177]. Correspondingly we conclude that the shifting in O Is after the second stage of nitridation is consistent with some reduction occurring for the molybdenum. In the final nitridation stage, the main peak shifts from 530.1 back to 530.5 eV and another component, apparently associated with the presence of OH groups, becomes more strongly established at 532.2 eV. The amount of OH is reasonably constant in the early stages (Figures 6.5 (a) to (c)), but its formation is more pronounced at the final stage (Table 6.1) where it is accompanied by some reduction. 115 Figure 6.5 Ols spectra from (a) the initially-formed M0O3 thin film; (b) the film from (a) after heating in NH 3 from 25 to 350°C; (c) the film from (b) after heating in NH 3 from 350 to 450°C; (d) film from (c) after heating from 450 to 700°C; and (e) the film from (d) after heating at 700°C for 1 h. 116 N/Mo and O/Mo ratios Table 6.2 lists the O/Mo and N/Mo atomic ratios for samples after the various stages of heating in NH3. The atomic ratios were obtained from the relevant peak areas following standard corrections for cross sections and background. All C detected was associated with weakly bound contamination; there was never any evidence for metal carbides (C Is binding energy 283.0 eV or below [178]) being formed during this whole study. But, as noted above, since some O signal was believed to originate from C-0 contamination, this contribution was estimated from C Is measurements for each sample, and subtracted prior to calculating the O/Mo ratio. Analogous corrections were made for the N/Mo ratio by excluding any contributions from weakly adsorbed NH3. For the initial M0O3/M0 sample, the O/Mo ratio (3.2) appears reasonably close to expectation, assuming all Mo detected is in a constant oxidized form. In any event, the initial nitridation reaction results in a significant drop in the O/Mo ratio to 1.4. However the N/Mo ratio is now 1.8, and therefore the combined amount of O and N appears broadly constant. That may suggest this process involves some substitutional replacement of O by N, although the evidence from the Mo 3d spectra (next section) indicates that this occurs in conjunction with a net reduction in the metal. In the nitridation step at 450°C, the O/Mo ratio decreases to 0.5, although there is an increase to 0.8 after heating at 700°C. As well, the N/Mo ratio gradually decreases with the further nitridation, and reduces to 1.2 at the firtal stage. Heating at 450°C reduces the O content, and the summed (N+0)/Mo ratio (-2.0) suggests that memory of the M0O3 structure must now be lost. 117 Table 6.2 Atomic ratios indicated by XPS for the initial M0O3 film and after the different stages of heating in NH 3 O/Mo N/Mo (N+0)/Mo M0O3/M0 (initial) 3.2 - 3.2 1st stage (25 -350°C) 1.4 1.8 3.2 2nd stage (350 - 450°C) 0.5 1.5 2.0 3rd stage (450 - 700°C) 0.7 1.3 2.0 final stage (700°C, lh) 0.8 1.2 2.0 After correcting for effects of C-O contamination and NH 3 adsorption (see text). 118 Mo3d spectra Figure 6.6 shows the Mo 3d spectra observed during this work, and binding energies for the various components are listed in Table 6.1. Three components are apparent for the starting M0O3/M0 system, namely Mo(0) (binding energy 228.1 eV, discussed above), Mo(+4) (230.1 eV) and Mo(+6) (232.7 eV). Treatment with NH 3 at 350°C for 30 min yields a spectrum which can be well fitted by five components with binding energies at 228.1, 229.5, 230.3, 231.0 and 232.3 eV. It is clear from the relative proportions that a substantial reallocation of the Mo chemical states has occurred as a result of heating to 350°C. The proportion of the metallic form appears to have increased slightly from the starting situation, even though there has been an observation that heating to at least 470°C is required for NH 3 to reduce M0O3 to Mo(0) [179]. The largest component at 232.3 eV appears at a slightly lower binding energy than expected for Mo(+6), but greater than that for the Mo(+5) oxide, namely 231.0 eV [171]. The component at 232.3 eV is tentatively associated with some replacement of O in the initial M0O3 lattice by NH or NH2, either of which may be expected to lower the binding energy. But any H released from the NH3 that contributes to OH bond formation will tend to increase the binding energy [24]. Likewise the component at 231.0 eV is viewed as the Mo(+5) oxide, but with incorporation of some N and OH. The components at 230.3 and 229.5 eV are close to expectation for Mo(+4) [171]. We interpret these values as representatives of a range of compositions for Mo(+4) oxynitride, where the first is O-rich and the second N-rich. The situation after heating to 350°C appears quite complex, and the binding energies reported here (as well as those below) should only be used to provide guides to trends in behavior. Indeed all individual values may be best seen as representatives to cover small changes in composition through somewhat heterogeneous material. 119 Figure 6.6 Mo3d spectra from (a) the initially-formed M0O3 thin film; (b) the film from (a) after heating in NH3 from 25 to 350°C; (c) the film from (b) after heating in NH 3 from 350 to 450°C; (d) film from (c) after heating from 450 to 700°C; and (e) the film from (d) after heating at 700°C for lh. 120 Heating at 450°C in the second stage of nitridation resulted in no metallic Mo being detected by XPS (Figure 6.6 (c)). Other work has shown that Mo reacts with NH 3 to produce MoN over the temperature range 400 to 700°C [180], although Maoujoud et al. [169] reported that an approximately 100 nm film of Mo needed more that 15 h to be nitrided at 425°C. We repeated our NH 3 treatment at 450°C on another Mo03/Mo system, whose initial oxide was sufficiently thick that no Mo metal could be detected by XPS. The resulting Mo 3d spectrum was essentially unchanged from that in Figure 6.6 (c), and therefore we conclude that the reaction of the metallic component is not significant to the interpretation of the spectra after heating at 450°C. There are three minority components at 232.7 eV (interpreted as Mo(+6)), 231.3 eV (Mo(+5)) and 230.0 eV (Mo(+4) "O-rich" oxynitride), but the major component at 228.9 eV is interpreted as the "N-rich" Mo(+4) oxynitride. The binding energy is reduced by about 0.5 eV over that from the first stage of heating, but that may result from the presence of (i) more N, and (ii) less OH. The relatively low amount of O detected at this stage suggests that M0O2 cannot have a prominent presence. On the other hand, the relatively high involvement by N in the nitride form (Table 6.2) appears to eliminate M02N as a majority product. In other work, Volpe and Boudart [106] reported that, after extensive exposure to NH 3 at temperatures above 350°C, Mo oxynitride spread rather uniformly on Mo0 3. In summary: we believe that the dominant species from our heating at 450°C is a "N-rich" Mo(+4) oxynitride, although hydrogen may also be involved to some degree as noted above. Heating at 700°C does not have a dramatic effect on the Mo 3d spectrum compared with that after the treatment at 450°C. For the four components identified after heating at the lower temperature, all are close to corresponding components present after heating at 700°C. However the latter also shows a significant component at 228.4 eV. This binding energy is close to values 121 reported for a lithographically textured MoS2 crystal (-228.2 eV) [88] and for Mo 2 0 3 (228.4 eV) [181]. Choi et al. [21,175], in their work on the TPR reaction between NH 3 and Mo0 3 powder, reported the presence of a surface species with binding energy at 228.5 eV (after referencing to C Is at 284.7 eV) which they concluded may be due to Mo(+3). We tentatively follow this assignment for the component observed at 228.4 eV in this work after heating to 700°C. 6.3.3 Valence band spectra Figure 6.7 shows low resolution spectra measured for the valence band after each of the various treatments included in this study. The starting spectrum for Mo03/Mo, with a band maximum at -7.5 eV, is similar to that reported by Katribe et al. [182] for M0O3 in powdered form. The effect of the H plasma is substantial; new structure appears at -1.7 eV and the initial maximum shifts to lower binding energy by about 1.5 eV. Consistently with observations of Katribe et al, we take these changes as indicative of reduction. In turn, the N plasma treatment results in a reduction in the intensity of the peak at -1.7 eV, and the spectrum becomes more similar to that of Mo03/Mo. These latter changes appear to indicate reoxidization with recovery of some Mo(+6). The effects of the H and N plasmas appear informative in relation to changes occurring when the initial sample is heated in NH 3. The nitridation reaction at 350°C results in a general broadening of the valence band, compared with the initial situation, but there is also a shift to lower binding energy (Figure 6.7 (b)). Heating at 450°C results in two new bands becoming established at close to 6 eV and 2 eV. Comparing with the situation after treating in the H plasma, it appears that structure in the valence band is indicative of substantial reduction occurring at 450°C. After the 700°C treatment, the structure near 6 eV diminishes while that at around 2 eV increases significantly. Katrib et al. [182] in their work on powdered Mo0 3 122 Figure 6.7 Low-resolution valence level spectra measured from (a) the initial M0O3 thin film, followed by exposures to the Ff plasma and the N plasma; (b) the nitrided film after three stages of heating in NH3. 123 attributed a band at ~2 eV to Mo 4d-5s states and the broad band with maximum at -7.2 eV to O 2p states. The new features in this work include incorporation of N, which may be expected to lower the energy of the 2p states, and the direct effect of H on the valence band. The latter appears to be an appreciable one when a substantial population of H is added (Figure 6.7 (a)). 6.4 Discussion Studies of surface composition and chemical state during the early stages of the nitridation of M0O3 by NH3 have been quite limited, but it is believed that the current work, including the H plasma, N plasma, and direct NH 3 reactions, gives some suggestive clues to the processes involved. The initial M0O3 structure is built up of (distorted) octahedra, each being joined to six others through the corner O atoms [179]. The H plasma treatment is consistent with H atom addition to break some Mo-O-Mo bonding units; in principle, for each H added, one OH is formed bonded to one Mo atom, while the other Mo experiences a reduced oxidation state. This is illustrated schematically in Figure 6.8 (a). The H-plasma treatment was carried out basically at room temperature, and this appears to prevent H2O being eliminated from the oxide structure. The subsequent heating results in a decrease in the O/Mo ratio, and this plausibly corresponds to the elimination of water. The effect of the N-plasma, as carried out in this work, is consistent with the recapture of some H atoms incorporated from the H-plasma, and this enables a regeneration of Mo(+6), as indicated in Figure 6.8 (a). Although Mo reduces during the H-plasma treatment, and there is an increase in the involvement of OH, no nitride is detected during the N-plasma treatment. This suggests that nitride formation, effectively involving an exchange of O by N, is an activated process which does not occur significantly during the low-temperature treatment. 124 Plasma Treatments: (a) Mo (+6) add H , M a = 0 (+6) Mo ' + (+6) M ° Mo OH (+4) (+5) H ,0 OH " ^ O H (+4) Mo (+5) OH • 2H M o Mo O H (+6) (+6) , Mo = O (+6) Early stage of NH 3 TPR: (a) + (b) (b) Mo ( - )x ( • )5-x-OH + NH 3 -> Mo ( - )x ( • )5.x-NH2 + H 2 0 ( - ) • -OH - N H 2 indicates Mo-0 single bond vacancy in 6 coordinate shell around Mo Mo-OH bond to be reacted with NH 3 Mo-NH 2 bond formed by H 2 0 elimination reaction Figure 6.8 Schematic indications of: (a) the effects of adding H to Mo0 3 followed by removal of either H 2 0 or H; and (b) the incorporation of N species into the film structure with local vacancies: The Mo=0 may correspond to two O atoms bridging to the same indicated Mo atom. 125 The interpretations made through the present studies for the M0O3/NH3 reaction emphasize five main oxidized components for the Mo as detailed in Table 6.3. The heating at 350°C is indicated to result in a product where the proportions of the Mo(+6) and "O-rich" Mo(+4) states are comparable, while the Mo(+5) and "N-rich" Mo(+4) states provide minority components. A view of this process has H being added to the oxide, with both formation of OH bonds and reduction of some Mo atoms to the +4 and +5 oxidation states (Figure 6.8 (a)). The temperature now appears sufficient to allow the elimination of water and the formation of O vacancies, which in turn encourages the occupation by N species. Initially the incorporated N is likely to be bonded to H (Figure 6.8 (b)), but also the N content should increase as H and O desorb as water. This gives a suggestive overview to the formation of the oxynitrides, and it seems broadly consistent with observations made by others for this temperature range. Earlier Lyutaya [176] reported that nitridation of powdered M0O3 with NH3 begins at temperatures less than 400°C, and Volpe and Boudart [106] indicated for 350°C or below that gradual structural change occurs at the M0O3 surface with some reduction. Inevitably the coordination number at Mo will reduce as the N content gets larger, although direct Mo-Mo bonds may also form [179]. For example, in bulk MoN each Mo atom has eight next-nearest neighbouring metal atoms at 2.85 A [183], which is only about 0.12 A greater than the nearest neighbour distance in bulk metallic Mo. 6.5 Concluding remarks A plot showing the relative proportions of these different Mo components with heating stage is shown in Figure 6.9; the basic data for this is provided in Table 6.1, along with corresponding information for components contributing to the N Is and O Is spectra. After heating at 450°C, the "N-rich" Mo(+4) oxynitride dominates, although we cannot yet provide an independent check on whether this product corresponds to the fee oxynitride reported by Volpe 126 Table 6.3 Five main Mo components emphasized for the M0O3/NH3 reaction. Mo3d5/2 Components Description 232.5±0.2 Mo(+6) basically Mo0 3 but with a little incorporated N and OH 231.1 ±0 .2 Mo(+5) oxynitride 230.1 ±0 .2 "O-rich" Mo(+4) oxynitride 228.9 ±0 .2 "N-rich" Mo(+4) oxynitride 228.4 ±0.1 Mo(+3) oxynitride 127 1 st stage 2nd stage 3rd stage Final stage (350°C) (450°C) (450-700°C) (700°C- lh) o Mo(+6) A Mo(+5) • "0 rich" Mo(+4) oxynitride • "N rich" Mo(+4) oxynitride • Mo(+3) oxynitride Figure 6.9 Relative proportions of the different Mo components after the initial M0O3 film is treated in NH3 to the different heating stages. 128 and Boudart [106]. These latter authors believed that the amount of oxynitride considerably exceeded the amount of M0O2, although this oxide was also reported from their XRD studies. The interpretation of our XPS observations cannot be more than semi-quantitative, but the trend toward reduced oxidation state with increased stage of heating seems unambiguous. From the binary phase diagram constructed by Hagg [184], both M02N and MoN are stable at room temperature, and indeed two parallel reaction pathways have been proposed based on the XRD observations [21,108,185]. For low heating rates (350 to 450°C at 40°/h), the reaction route M0O3 —> MoOxNi_x — > Y-M02N is reported to be dominant, while for high heating rates (350 to 450°C at 1007h) the alternative M0O3 -> Mo0 2 -> Y-MO2N + 8 -M0N has been favoured. The first reaction pathway has been considered to give high surface area product, a possible consequence of the topotatic nature of the process (both MoOxNi_x and Y-M02N have crystallographic orientational relationships to M 0 O 3 [106,108]), and correspondingly the formation of non-topotatic M0O2 in the second pathway has been believed to encourage the formation of low surface area nitride [21,106,108]. Our work fits the high heating rate regime, and we conclude that the main species during the second nitridation stage is a Mo(+4) oxynitride with a high N content. This contrasts with observations by Choi et al. [21,185] who, for the same heating rate, reported M0O2 as the major intermediate species in the second stage of the nitridation. For the final stage of nitridation in this work, the indicated N/Mo ratio makes it appear unlikely that M02N could have a dominant role compared with MoN. But against that it must additionally be recognized that we are probing surface regions, whereas the XRD observations are necessarily assessing bulk structures. The two may not even correspond [186]; indeed in related work Demczyk et al. [109], using high-resolution transmission electron microscopy and XPS, concluded that while a bulk material was 129 Mo 2N (fee) the lattice structure near the surface corresponded to Mo2N 3 . x O x . In principle there could be basic differences between surface and bulk compositions from thermodynamics, but differences may result from other factors. Oxygen (either from 0 2 or H20) in the surroundings could contribute impurity, and Toth [105] has mentioned that this element can be a particularly difficult impurity to eliminate and analyze in metal nitrides. Also kinetic factors may be different for different types of sample. For example, with our very thin oxide layer, the apparent non-formation of M0O2 could be a direct result of the possibility that H and N atoms may be able to diffuse through the whole system much more readily than could happen with a bulk crystalline sample. In a real industrial process, the catalysts would likely be deposited on a support (e.g. alumina or silca), and while the information provided by this work should provide a reference point for the behaviors of nitrided M0O3 samples on oxide supports, there is still a need to assess whether the support may provide additional chemical effects for the materials studied here. A start to consider this topic is covered by follow up described in the next chapter. 130 7. Studies of the Nitridation of M0O3 Thin Films on Alumina and Silica Supports 7.1 Introduction Recent work has suggested that unsupported Mo nitrides are active catalysts for the hydrodenitrogenation (HDN) [19-21] and hydrodesulfurization (HDS) [22,23] reactions. The evidence so far has been on synthesis, and on relating reactivity to the bulk structures involved [19-23,106,108]. Additionally there are reports that Mo nitride supported on Y-AI2O3 can show higher activities for the HDN and HDS reactions than the commercial sulfided M 0 O 3 / A I 2 O 3 catalyst [110,112,187]. Characterizations to date have emphasized x-ray diffraction (XRD) analyses to identify composition, and O2 chemisorption to identify active sites [110,112], but direct investigation of the surface species involved have been very limited. Mo-support interactions are reported to be different with alumina and silica [47,48,69,81]. The active species are believed to disperse less well on silica-supported Mo catalysts, and this apparentally results in a much lower activity for HDS of thiophene compared with the situation on alumina supports [75,76]. No surface characterizations have yet been made to elucidate the Mo-support interactions, and their roles in influencing the activity of nitrided catalysts. The study reported in this chapter is part of a broader program to use XPS to characterize model catalyst systems relevant to HDN. This particular study investigates how alumina and silica as supports affect the nitridation of M0O3; comparisons are particularly made with results from the investigation, reported in Chapter 6, for the same process involving a thin film of M0O3 grown on a planar Mo substrate. The present work aims to grow corresponding M0O3 films on oxidized Al (i.e. A10x) and on oxidized Si (i.e. SiO"2), and to compare the film reactivities on interaction with NH3. Investigations were also made for the effect of heating rate on the M0O3 to nitride conversion; other reports [21,108] have suggested that the heating rate can affect the 131 nitridation mechanism, as well as the surface area of the product. Our previous study in Chapter 6 on M0O3/M0 used the so-called "high-heating" rate (i.e. for raising the temperature of the system from 350 to 450°C in 1 h). That is continued through the first part of the work reported in this chapter and applied to the nitridation of M 0 O 3 / A I O X and of Mo03/Si02; in the second part of this study comparable investigations are made for the "low-heating" rate, where temperature is raised from 350 to 450°C in 21/2 h for the Mo03/A10x system, and new results are included for the M0O3/M0 system. 7.2 Experimental As previously, the M0O3/M0 system was prepared by heating Mo metal strips (1x0.25 cm2) at 250°C in compressed air for 4 h. For the model supported catalysts, the supports (1x0.25 cm2) were prepared by heating strips of Al plate and a Si(100) wafer in compresed air at 550°C. The heating to form A10X/A1 was done for 3 h, while that for the Si02/Si system lasted 12 h. The preparation of the Mo03/A10x system started by spin coating the oxidized Al surface with 0.09 M aqueous solution of ammonium heptamolybdate (AHM, for (NH4)6(Mo7024).6H20) (pH = 5.7), and that for the Mo03/Si02 system was initiated by dipping the oxidized Si support into 0.001 M AHM (pH = 4.8) for 1 min. Afterwards the model catalysts were calcined at 450°C for 3 h in flowing air (1 atm). For each system, the nitridation reaction was carried out with flowing anhydrous N H 3 in the U-shaped stainless steel reactor shown in Figure 5.1. The reactor was heated in a tube furnace with a variable transformer to control the heating rates. In summary, for the first stage of heating in N H 3 , the catalyst temperature was raised from 25 to 350°C in about 30 min; for the second stage, the temperature was increased to 450°C at 100°/h ("high" heating rate), or at 40°/h ("low" heating rate), for both the M0O3/M0 and Mo03/A10x systems. In the final nitridation stage, the heating rate was increased to 200°/h, and the final temperatures were 132 limited to 550°C for the Mo03/A10x system, and to 700°C for the M0O3/M0 and Mo0 3/Si0 2 system (each of these latter system was held for a further 1 h at the final temperature 700°C). After these heatings, the model catalysts were rapidly cooled to room temperature in flowing NH3 and stored in a N 2 box before transferring to the XPS facility. XPS measurements were carried by using the non-monochromatized MgKa source (1253.6 eV) operated at 10 kV, 20 mA. Lower resolution spectra, for elemental identification, were measured with the pass energy set at 192 eV. The higher resolution spectra for Mo3p, Mo3d, A12p, Si2p, Ols, Nls and Cls were measured with the pass energy at 96 eV. Unless otherwise stated, spectra were measured for normal take-off angle (0). No charging effects were detected for the M0O3/M0 system, although uniform charging was apparent for the MoCVAlOx and MoC»3/Si02 systems. All binding energies quoted here were referenced to structure for adventitious carbon in the C Is spectra being fixed at 284.7 eV (this value was indicated for the M0O3/M0 system for which Au4f7/2 at 84.0 eV could be used as the primary reference). Guides to the relative elemental percentages probed by the photoelectrons were estimated by integrating peak areas after correction for instrumental sensitivity factors; curve fittings assumed the standard non-linear background subtraction, and used component peaks based on 80% Gaussian, 20% Lorentzian shape. 7.3 Results 7.3.1 Nitridation of Mo03/A10x Mo-A10x interaction The initially prepared A10X/A1 sample shows an oxide layer which is too thick for the metal component to be detected in the A12p spectrum. After heating such a sample to 550°C in NH 3 , XPS could not detect any N; also no changes were seen in the A12p or Ols spectra. The 133 addition of the molybdenum oxide yielded observed binding energies for Mo3d5/2 and Mo3p3/2 at 232.6 eV and 398.4 eV respectively, which are fully consistent with Mo(+6) being the only Mo species present (e.g. 232.7 eV by Zingg et al. [62] and 232.5 eV by Spevak and Mclntyre [52] after referencing to Cls at 284.7 eV). Prior to calcination the Nls spectrum showed a peak at 401.5 eV, apparently related to N H / [188]; however no N species were detected by XPS after calcination, so suggesting that decomposition and desorption occurred in that process. In the study in Chapter 4, we suggested that M0O3 interacts with A10x at 450°C to form direct Al-O-Mo bonding, and this was indicated by shifts in the A12p peak and in the Auger parameter. No comparable shifts were seen in this current work for the pre-calcined and post-calcined states. The difference in observation appears to relate to the pretreatment insofar as the sample used here had been subjected to much more heating (550°C for 3 h) compared with the support in the previous work (250°C for Vi h). It is believed that the current sample after heating is more y-A I 2 O 3 like [44], whereas that used previously had a higher concentration of surface hydroxyl groups. Although changes in Al are not apparent after calcination, following the treatment with AHM, this heating does give small changes in the Mo3d spectrum (Figure 7.1). Specifically, there is a slight shift to higher binding energy (by 0.2 eV) and the FWHM increases from 1.5 to 1.6 eV. Also the apparent Mo/Al ratio increases from 0.48 to 0.59 after the calcination. This trend is in the same direction as that seen previously, although now the enhancement indicated for the Mo distribution appears reduced, presumably because of the smaller number of OH groups on the alumina to help orient the Al-O-Mo bonding interactions. Nitrided Mo03/A10x compared with nitrided M0O3/M0 Mo3d and Mo3p spectra from nitrided M0O3 films on alumina and on metallic Mo are included in Figure 7.2. Based on the results in Chapter 6 (Table 6.3), five Mo species were 134 240 238 236 234 232 230 Bind ing Energy (eV) Figure 7.1 Comparison of Mo3d spectra from M o C V A l O x samples in the uncalcined form continuous line) and after calcination at 450°C (dashed line). 135 Figure 7.2 Comparison of (a) Mo3d spectra and (b) overlapping Mo3p3/2 and Nls spectra from Mo03/A10x and M0O3/M0 samples which have been nitrided at 550°C, and from Mo0 3/Si0 2 which has been nitrided at 700°C. The Nls and Mo3p3/2 components for the spectra under (b) are identified by dashed lines and continuous lines respectively. 136 detected after the nitridation: Mo(+6) at 232.7 eV, Mo(+5) at 231.6 eV, "O-rich" Mo(+4) at 230.3 eV, "N-rich" Mo(+4) at 229.2 eV and Mo(+3) at 228.5 eV. These Mo species are believed to be in oxynitride forms. The binding energies and relative concentrations (Table 7.1), appear similar for nitrided MoCVAlOx as for nitrided M0O3/M0, although the former sample is indicated to have higher concentrations of the lower Mo oxidation states. Two types of N are detected in Nls spectra from these model catalysts; one N component has a binding energy at 397.1 eV and is believed to originate from nitride formation [175], whereas the other at 399.3 eV has been ascribed to Mo-NH x species [174]. Table 7.1 includes the Mo/Al, N/Mo and O(M 0/Mo ratios for the nitrided model catalysts on alumina; 0 (M0) refers to those O atoms bonded to Mo (this was estimated by fitting the Ols spectra to components based on Al-0 bonding at -531.0 eV and on Mo-0 bonding at -530.5 eV). The Mo/Al ratio increases from 0.59 to 0.65 after the nitridation, and this suggests that the Mo dispersion is slightly improved. At the same time the O/Mo ratio decreases from 2.2 to 0.5 after the nitridation while the corresponding change for the M0O3/M0 system is 3.2 to 0.8. The resulting N+O content is similar in both cases, but Mo03/A10x yields more nitride apparently because the replacement of O by N occurs easier than in M0O3/M0. 7.3.2 Nitridation of Mo0 3/Si0 2 Mo-silica interaction The Si2p spectrum confirms the presence of oxide and elemental forms (binding energy difference 4.3 eV) at the Si02/Si support [189]. The O Auger parameter (oc'= 1039.2 eV, defined as the sum of the Ols binding energy and the O(KVV) kinetic energy) is consistent with an amorphous Si0 2 layer [190] formed on top. The ratio of intensities for the oxide and elemental Si components in Si2p suggests an oxide thickness of about 32 A (assuming an electron mean 137 Table 7.1 Atomic ratios and binding energies (eV) measured for MoOVAlOx prior to and after nitridation. That formed after nitridation is compared also with the M0O3/M0 system where both have been heated in NH3 with the high heating rate procedure. Sample Mo/Al 0(M 0)/M0 N/Mo Nls Mo3p3/2 Mo3d5 /2 Mo03/A10x uncalcined 0.48 2.8 - 401.5 398.4 232.6 calcined at 450°C 0.59 2.2 - 398.7 232.8 nitrided to 550°C 0.65 0.5 2.0 399.3 (14%) 397.1 (86%) 398,7 397.6 396.3 395.2 394.5 232.7 (11%) 231.6 ( 6%) 230.3 (17%) 229.2 (19%) 228.5 (48%) M0O3/M0 nitrided to 550°C 0.8 1.5 398.8 ( 5%) 397.2 (95%) 398.6 397.0 395.9 395.0 394.3 232.7 (10%) 231.1 ( 6%) 230.0 (14%) 229.1 (27%) 228.4 (42%) 138 free path of 28 A [120]). No observable change is seen in Si2p after coating with AHM. Compared with results reported for real catalysts of this type [47,48] (e.g. peak widths of 2.8 eV for Mo3d and 3.2 eV for Si2p [47]) much narrower peaks are found in this work (1.6 eV for Mo3d and 2.0 eV for Si2p). Like for the Mo03/A10x system, after treating the silica support with AHM, structure present in Nls can be related to NH 4 + , although again no N components are detected after calcination. Figure 7.3 compares Mo3d spectra for the uncalcined and calcined Mo oxide catalysts on SiC^. The initial Mo3d.v2 and Mo3p3/2 peaks at 232.5 eV and 398.4 eV respectively are consistent with the presence of only Mo(+6) [52,62]. The heating process slightly lowers the binding energy of the Mo3ds/2 peak to 232.3 eV, and this is more clearly seen in measurements made at smaller take-off angles, where indeed a slight increase is apparent for the peak width (from 1.6 to 1.8 eV). XPS has been a commonly used technique for investigating the Mo-silica interaction [47,48,81], but conclusions have been confused somewhat by the frequent presence of sample charging [48]. Although the binding energies for the Mo3d doublet have been reported to have lower values than those of bulk Mo0 3 [48,84], a shifting to higher binding energy has also been indicated [81]. The present work aimed to keep the SiG*2 film relatively thin, and although charging does occur, we feel the reference chosen still allows changes in the Mo species to be followed. No evidence for significant differential charging was apparent from use of the negative potential bias technique [129], and therefore it appears that the increased peak width may be indicative of an additional Mo species. For example, Gardardo et al. [47] reported Mo-silica interaction species, by observing variations in the Si2p binding energy with increased Mo loading (also larger peak width after calcination). Our loading is low and fixed, and therefore we cannot follow that trend, but the observed shift to lower binding energy for Mo3d after calcination is possibly consistent with this interaction. 139 I I I I I I I I I I I I I I I I I I I ! I I I I I I I I I I I I 240 238 236 234 232 230 228 B i n d i n g Energy ( e V ) Figure 7.3 Mo3d spectra from Mo0 3/Si0 2 samples: (a) uncalcined, (b) calcined at 450°C, (c) and (d) same sample as in (b) but the spectrum was collected at an exit angle of • 30°. 140 Calcination reduces the Mo/Si ratio from 0.25 to 0.08, although the latter value from the calcined sample is essentially constant with changing exit angle (0=45° and 30° both give the value of 0.07). This constancy could indicate a uniform distribution of Mo through the silica, but it may also be suggestive of patch formation [122]. De Jong et al. [43] used atomic force microscopy (AFM) to monitor this surface during calcination, and they reported volatilization and sintering for heating above 325°C with subsequent formation of patchy AFM images. Nitrided Mo0 3/Si0 2 compared with M0O3/M0 Figure 7.4 shows the Si2p and Nls spectra resulting after heating the Si02/Si support in NH3. Two kinds of N are detected: one at 397.3 eV (87%) is believed to indicate nitride (i.e. Si-N) formation, while the higher binding energy component at 399.0 eV (13%) is possibly due to the presence of Si-NHX [191,192]. Curve fitting for the Si2p spectrum requires the introduction of a small extra peak at 2.2 eV higher binding.energy than for Si(0), and this probably originates with the Si-N bonding [190]. The relevant peak areas, after correcting for atomic sensitivity factors, give for this species a N/Si ratio of 1.3. Figure 7.2 compares Mo3d, Mo3p and Nls spectra for nitrided M0O3 when supported on Si0 2 and when supported on metallic Mo. The binding energy values listed in Table 7.2 are for the Mo and N species as indicated by the curve fitting results. A major difference between the samples is the presence of an extra Mo species at 227.4 eV on the nitrided Mo0 3/Si0 2. The binding energy difference from the value for Mo(+6) (i.e. 5.0 eV) is consistent with some metallic Mo(0) component [119] being formed on the silica-supported sample. Curve fitting for the overlapping Mo3p and Nls spectra is consistent with all N being present in the nitride (or oxynitride) form. The 0(Mo)/Mo, Mo/Si and N/Mo atomic ratios listed in Table 7.2 have been made after correcting for contributions by metallic Mo and by the direct Si-N bonding. The constant Mo/Si 1 4 1 I i ' I I I I I I I I I I I i : I : I I I I i 1 I I I 404 402 400 398 396 394 392 Bind ing Energy (eV) Figure 7.4 Spectra from the SiGVSi sample after nitriding at 700°C: (a) Si2p including S1O2, elemental Si and the new formed Si-N species; (b) Nls. 142 Table 7.2 Atomic ratios and binding energies (eV) measured for Mo0 3/Si0 2 prior to and after nitridation. That formed after nitridation is compared also with the M0O3/M0 system where both have been heated in NH 3 with the high heating rate procedure. Sample Mo/Si Q(Mo)/Mo -N/Mo Nls Mo3p 3/2 Mo3d5/2 Mo0 3/Si0 2 uncalcined 0.25 401.6 398.4 232.5 calcined at 450°C 0.08 398.2 232.3 nitrided to 700°C 0.08 0 1.9 397.1 398.4 397.2 396.0 395.1 394.3 393.4 232.4 (11%) 231.2 ( 6%) 230.0 (12%) 229.1 (12%) 228.3 (41%) 227.4 (18%) Mo03/Mo nitrided to 700°C 0.8 1.2 397.3 398.6 397.0 395.9 395.0 394.4 232.7 (10%) 231.1 ( 5%) 230.0 (18%) 229.1 (25%) 228.5 (42%) After correcting for the effects of Si-N bonding 143 ratio (0.08) suggests that vigorous migration of Mo may not occur during the thermal nitridation, but the N/Mo and 0 ( M o)/Mo ratios (1.9 and 0.0 respectively) indicate that the silica-supported nitride has more N and much less O than for the system formed on metallic Mo (corresponding ratio values 1.2 and 0.8 respectively). In the supported case, the allocation of O bonding to Mo may be underestimated because of the large O signal from Si02 and perhaps from some overlap with Mo-OH bonding, but in any event it seems clear that the displacement of O by N, that occurs during the nitridation process, is more effective on the SiO"2 support. Also while amounts of Mo(+6) and Mo(+5) species after nitridation are comparable on the two samples, the total amount of Mo in the lower oxidation states is greater with the Si0 2 support. Therefore it appears that the support interactions also aid the reduction during the NH3 reaction. 7.3.3 Effect of heating rate M0O3/M0 Table 7.3 lists results obtained by following the nitridation of the M0O3/M0 and MoOVAlOx systems using different heating rates for the second stage, in which the temperature is raised from 350 to 450°C. In Chapter 6 we reported for M0O3/M0, with high heating rate for this stage, that the major product is "N rich" Mo(+4) oxynitride (55%, binding energy 228.9 eV). The full specification for this product (sample A) is given in Table 7.3. A new result here is for the low heating rate (sample B) when, as well as the Mo(+4) species at 229.0 eV, 20% of the product is composed of Mo(+3). Although the O/Mo ratios are similar (-0.5) from each of these heating rates, the N/Mo ratio for sample B (1.8) is increased slightly over that for sample A (1.5). The low O content in each case suggests that Mo0 2 is not an appreciable component in either product. The higher N content for the sample produced at the lower heating rate suggests 144 Table 7.3 Atomic ratios and binding energies (eV) measured for M0O3/M0 and MoCyAlOx systems after nitriding in NH 3 for different heating rates (high or low) and different final temperatures (450 or 550°C). Sample 0(Mo)/Mo N/MO NJ_s Mo3p3/2 Mo3d5/2 (A) MoCVMo 0.5 1.5 398.9 ( 9%) 398.6 232.7 (15%) high, 450°C 397.1 (91%) 397.3 231.3 ( 9%) 395.9 230.0 (21%) 394.9 228.9 (55%) (B) Mo03/Mo 0.5 1.8 397.1 (100%) 398.7 232.7 (13%) low, 450°C 397.0 231.0 ( 7%) 396.1 230.1 (17%) 395.0 229.0 (43%) 394.4 228.4 (20%) (C) Mo03/Mo 0.8 1.5 398.8 ( 5%) 398.6 232.7 (10%) high, 550°C 397.2 (95%) 397.0 231.1 ( 6%) 395.9 230.0 (14%) 395.0 229.1 (27%) 394.3 228.4 (42%) (D) Mo03/Mo 0.4 1.8 .397.1 (100%) 398.6 232.6 (12%) low, 550°C 396.4 230.4 (12%) 395.2 229.2 (20%) 394.4 228.4 (56%) (E) Mo03/A10x 0.2 2.1 399.5 (16%) 398.7 232.7 (15%) low, 450°C 397.1 (84%) 397.1 231.1 (13%) 395.8 229.8 (20%) 394.9 228.9 (34%) 394.4 228.4 (18%) (F) Mo03/A10x 0 •2.3 399.4 (11%) 398.6 232.6 (15%) low, 550°C 397.0 (89%) 396.5 230.5 (14%) 395.3 229.3 (16%) 394.5 228.5 (55%) (G) Mo03/A10x 0.5 2.0 '399.3 (14%) 398.6 232.7 (11%) high, 550°C 397.0 (86%) 397.5 231.6 ( 6%) 396.2 230.3 (16%) 395.1 229.2 (19%) 394.4 228.5 (48%) 145 that the N displacement is increased by the longer time taken to raise the temperature from 350 to 450°C. For the third stage of nitridation (temperature raised from 450 to 550°C), the O/Mo and N/Mo ratios are not significantly affected by the heating rate (samples C and D). For the high heating rate, the Mo(+3) content is indicated to increase from 0 to 42%, whereas for the low heating rate this increase is from 20 to 56%. According to results from the Mo3d curve fitting, no Mo(+5) species is found on sample D, whereas there is still about 6% in C. In addition, curve fittings for the overlapping Mo3p and Nls structures are consistent with nitride being the only N species after the nitridation at the lower heating rate, whereas two N species, interpreted as nitride and adsorbed NHX, are detected for the sample produced with the higher heating rate. It appears that Mo-NH x is a precursor for the final nitride product, and that the lower heating rate also favours the Mo-NHx to Mo-N conversion. Overall these observations suggest that the heating rate especially affects the degree of reduction, while the total amount of nitridation is affected less. Mo03/A10x As far as Mo species are concerned, the effect of nitriding Mo03/A10x with the high heating rate is very similar to that for M0O3/M0. A comparable relationship applies when Mo03/A10x is nitrided with the low heating rate (sample E). Just as for sample B, Mo(+3) starts to form on nitrided Mo03/A10x at 450°C. Similarly the O/Mo ratio, which reduces to 0.2 in sample E, shows that O removal is more complete when the low heating rate is used. As the temperature increases to 550°C, the Mo(+3) content increases from 18 to 55% and the O-bound-to-Mo component appears to vanish. However, it is indicated that the combined (0+N)/Mo ratio remains equal to that for M0O3/M0 for the comparable nitriding conditions. Also, comparable 146 with the situation noted above, it is recognized that there is some uncertainty in the O-bound-to-Mo content because of possible overlap with contributions from Mo-OH and Al-0 bonding in the Ols spectrum. Nevertheless the trend seems substantiated that both reduction and the O-N replacement occur during the third stage of heating in NH 3. Also by comparing products for sample F (low heating rate) and sample G (higher heating rate) on the A10x support, it seems clear that the lower heating rate favors both the reduction and the O-to-N replacement. 7.4 Discussion Although we could not monitor the absolute Mo loadings for the model catalysts considered in this work, that on A10x is believed to be about twice that on Si0 2 prior to calcination, but after calcination the difference is increased to a factor of about six. According to Wang and Hall [50], AHM solutions in the pH range 4.8-5.7 dominantly contain polymolybdate ions. Since hydroxyl concentrations are less on silica than alumina surfaces [55], a smaller number of Mo ions condense on the former, although some physical adsorption occurs. However the weakly bound species are lost on heating, and the Mo content on the Si0 2 sample is consequently reduced compared with A10x. For the latter, our previous results, and the slight shifting observed here for Mo3d to higher binding energies, encourage the belief that Al-O-Mo linkages are formed. The corresponding situation with Si seems less clear, and although the concept of Si-O-Mo bonding is commonly accepted [193], the degree of involvement by these linkages appears as a function of the M0O3 content [48,194], and it is always less than for Al. The behaviors of the Mo3d peaks suggest that the Mo in Mo0 3/Si0 2 gains electrons from the support, whereas that trend is reversed with Mo03/A10x. The Mo0 3/Si0 2 and Mo03/A10x systems show different behaviors toward the thermal nitridation by NH 3 . Because of our use of aluminum substrates we could not heat these samples 147 much above 550°C, and so we cannot make a full comparison between the nitridation of Mo03/Si02 and MoCVAlOx. Therefore we have compared them individually with M0O3/M0. The amounts of Mo on both the oxide supports have higher N and lower O contents, suggesting the greater effectiveness of the O-N replacement in these cases. Possibly the nitriding process is easier starting with polymolybdate ions than bulk M0O3 with its longer range structure. Ngai et al. [195] also reported that supported Mo nitrides have slightly higher N contents than non-supported samples. The results here suggest that the nitrided Mo03/AIOx samples reduce similarly to the equivalently-treated M0O3/M0 samples. Our observation of the presence of Mo(0) on nitrided Mo03/Si02 suggests that overlayer Mo is more readily reduced by NH 3 than is the case for Mo03/Mo (for this latter system the oxide is sufficiently thick that no Mo(0) is detected by XPS either initially from the substrate or later as a result of the NH 3 reaction). Also the indications here are that the nitrided M0O3/AIOX reduces similarly to nitrided Mo03/Mo, at least up to 550°C. Colling and Thompson [112] studied different loadings of Mo nitride on real alumina-supported catalysts at 700°C and found no evidence for Mo(0) by XRD. It is therefore possible that M0O3/AIOX and M0O3/M0 reduce in similar ways in NH3 at temperatures up to 700°C. The evidence to date appears to indicate that Mo03/Si02 is reduced easier than Mo03/A10x, but further studies (e.g., on sapphire samples) are needed to substantiate this possibility and to assess the involvement of different Mo-support interactions. Our use of model planar catalysts has the advantage of allowing systematic changes in surface composition without significant interferences by the differential charging that can complicate interpretations of direct studies on real catalysts. The observations here that the Mo/Al and Mo/Si ratios change only slightly after nitridation suggest that, for the conditions used, the dispersion of Mo does not change substantially. But other work [69] has shown for sulfidation that the Mo/Al ratio can change markedly in the Mo03/Al203 (and Si02) systems. 148 Future studies are planned to extend the investigations for nitridation to different Mo loadings on these various supports. Colling and Thompson [112] concluded with XRD that a nitrided 16% Mo0 3/g-Al 20 3 catalyst sample is very similar to unsupported Mo nitride. Although different nitridation reaction pathways have been proposed as a result of the particular heating rate used for the second heating stage (between 350 and 450°C), our evidence suggests that changing this heating rate has an incremental, rather than dramatic, effect on the distribution of products. Similar Mo species are seen for the low-heating rate (407h) for both the Mo03/A10x and Mo03/Mo systems. The slower heating rate just seems to result in Mo(+3) appearing earlier. Further this carries through to the third stage of heating (450 to 550°C). That is, with the low heating rate for the second stage, the proportion of Mo(+3) after the further heating to 550°C is greater than when the high rate is used for the second stage of heating. Basically it seems that the reduction reaction is more complete, after reaching 550°C, when the low heating rate is used, instead of the high heating rate, for the second stage. Likewise the higher N content and lower O content suggest that the O-N replacement is more complete when the lower heating rate is used in the second stage. 149 8. Concluding Remarks 8.1 Summary for new results It is believed that the research discussed in this thesis, which emphasized the use of planar (model) Mo catalysts, has provided new information about the chemistry that occurs during their preparations. The hope is that the observations made in this context may extend to help understand behaviors of real, high-area supported catalysts. The samples studied in Chapter 4 and 5 can be designated 1% MoO3/AlOx(200), where the notation AlOx(200) identifies that the A10x planar support was heated at 200°C for Vi h, prior to the application of the M0O3. The effect of the subsequent calcination was then studied, and new evidence was provided to indicate that the molybdate disperses uniformly on the support when the calcination is done at 450°C for 3 h. Further observations, on the variation in the Auger parameter, support the conclusion that direct Mo-O-Al surface bonds are formed during calcination at 450°C. In turn, this surface linkage appears to provide the driving force for the enhanced Mo dispersion with these conditions. Assessments of the effect of dehydroxylation on the original aluminum oxide surface was done (in Chapter 7) by comparing with the support designated AlOx(550), where the Al planar support was heated at 550°C for 3 h prior to the application of the Mo0 3. By contrast, the 10% MoO3/AlOx(550) samples show no comparable shift between the pre-calcined and post-calcined model catalysts, when the calcination is done at 450°C. This appears due to the lower concentration of surface hydroxyl groups on the AlOx(550) support. However, the Mo3d spectrum does have a small shift to higher binding energy with calcination at 450°C, and that gives an indication that some Mo-O-Al linkages are formed during this heating process. 150 Comparisons are made with behavior for M0O3 on AlOx(550) and Si0 2 supports. It is observed after calcination at 450°C that less Mo is found on the Si0 2 surface, while the Mo3d spectral features shift to higher binding energy for MoCyAlOx(550) and to lower binding energy for Mo03/Si02. This is consistent with Mo in the MoCySiCh samples gaining electrons from the support, whereas that trend is reversed with M0O3/AIGV Observations on the sulfidation and nitridation of supported M0O3 catalysts confirm that the details of the Mo dispersion and the Mo03-support interaction (i.e. to form Mo-O-Al or Mo-O-Si linkages) directly influence behavior during these reactions. For the 1% MoO3/AlOx(200) model catalyst systems, the sulfiding of the 450°C-calcined catalyst suggests that the Mo-O-Al linkages are ruptured. The reduced Mo/Al ratio suggests the Mo atoms rearrange from the dispersed state to form clusters of Mo(+4) sulfide and this observation is consistent with the literature [70,98]. Direct comparisons made in this work between model catalysts and comparably prepared high-area catalysts provide a linkage between details of surface characterization and practical catalytic activity, although this area of study is still at an early stage in our research group. In the present work, HDN reactions of pyridine did not produce pentane, but it did appear that the sulfided 450°C-calcined MoO3/AlOx(200) sample provides better HDN activity than the other sulfided samples tested (i.e. the uncalcined, and the 200°C- and 350°C-calcined samples). The activity of the sulfided 450°C-calcined sample is possibly associated with the formation of a species, referred to as MoS2.x, which is only detected in XPS on this particular sample. This S species may provide unsaturated Mo atoms for the reaction, although lower oxidation states than Mo(+4) were not detected in this work. The presence of S22" may also help the HDN process (e.g. by providing sites for hydrogenation), but follow up studies are needed to assess the importance of this possibility. This work also provided indications that the 151 sulfiding process affects the Mo dispersion on the MoO3/AlOx(200) model catalysts. Compared with the uncalcined and 350°C-calcined samples, the higher Mo dispersion on the 200°C-calcined sample gave the highest amount of Mo(+4) sulfide, while the poorer Mo dispersion on the former two samples resulted in smaller amounts of Mo(+4) sulfide being formed. In fact, the Mo dispersion on the 350°C-calcined sample is slightly better than that on the uncalcined sample, and this may somehow lead to the different S species present after sulfidation. This is not understood at the moment, but further analysis in which attempts are made to follow the amount of O bonded to Mo during the sulfidation may be helpful. The studies with XPS reported here provide new insight to the NH3 nitridation of M0O3 thin films in the M0O3/M0, MoO3/AlOx(550) and Mo0 3/Si0 2 systems. The nitridation reaction pathways for M0O3/M0 involve initial hydrogenation, with the subsequent elimination of water, and the effective replacement of some O by N. Heating M0O3/M0 in NH 3 to 350°C gives some conversion of Mo(+6) to Mo(+5) and the "O-rich" Mo(+4) components, while heating to 450°C (lOOK/h, the high-heating rate), and to 700°C, give respectively a "N-rich" Mo(+4) form and a Mo(+3) oxynitride as the dominant components. Comparisons of heating rates for the second nitridation step from 350 to 450°C were made. Differences between the high heating rate (lOOK/h) and the low heating rate (40K/h) are incremental but definite. The lower heating rate appears favorable both for the O-N replacement and for the Mo reduction; the Mo(+3) oxynitride starts to form when the temperature reaches 450°C in the low-heating rate regime. Unlike the situation with sulfidation, the dispersion of Mo on the supports does not change substantially during the nitriding reaction. This work provides new information for the nitridation of MoO3/AlOx(550) and Mo0 3/Si0 2 as compared with that of M0O3/M0. It is found that M0O3 on the oxide supports shows easier O-N replacement than is the case with the 152 M0O3/M0 system. In general, the reduction behavior for MoO3/AlOx(550) is similar to that of M0O3/M0, but Mo species on Si0 2 are indicated to reduce more easily; for example metallic Mo is detected from the nitridation of M0O3 on Si02, whereas the metallic form is not seen when the support is AlOx(550). This may be interpreted by the details of the interaction between M0O3 and the different oxide supports; it appears that Mo in MoCVSiCh gains electrons from the support, whereas that trend is reversed with MoO3/AlOx(550). The surface composition for the nitrided catalysts is complicated. Throughout this work the emphasis has been on trends in Mo oxidation state, as well as on the general degree of involvement by N resulting from the NH 3 reaction. The atomic ratios quoted cannot do more than provide guides to behavior, particularly since the Nls components had to be deduced somewhat indirectly by subtracting Mo3p contributions from the measured overlapping spectra. In any event, the surface regions of the materials formed after each stage of the heating appear to have considerable heterogeneity given the range of oxidation states indicated for the Mo. No studies have yet been made to probe the variation in composition with depth below the surface. Partly for this reason, specific surface compounds have not been emphasized; even with a definite Mo oxidation state, there can be some variety in the possible compounds formed given the propensity for this element to form metal-metal bonds. Observations made here on the evolution of products from bulk M0O3 during the nitriding process contrast with conclusions reached previously from XRD; the discrepancies are not fully understood, but surface characterization with a technique like XPS should be particularly relevant for the development of understanding of the processes involved when surfaces of these materials are used as HDN catalysts. 153 8.2 Future directions 8.2.1 Quantification of Mo loadings To make model catalysts that correspond well to high-area catalysts, it is necessary to use comparable preparation methods. Nevertheless, the kinetics for processes involved in sample preparations can never correspond exactly because of the involvement of capillary interactions in the impregnation of pore structure in high-area supports, an effect that is not applicable to preparations on flat model supports. Throughout the studies reported in this thesis, the Mo has been applied to supports mainly by dipping. This method is similar to the equilibrium adsorption method used for preparing conventional high surface area catalysts. The adsorption of molybdate is determined by the isoelectric point for the supports, while the loading is controlled by the pH of the molybdate solution. However, the drawback of dipping is that some excess liquid may be left on the support which appears to increase the sample loading. Recently spin-coating has been employed for the deposition of thin layers of inorganic salt solutions on flat supports [43,196], and attempts made to quantify the deposits by Rutherford backscattering spectrometry (RBS) [43,196]. In general, RBS is more of a bulk technique compared with XPS or SIMS, but observations with it have helped develop mathematical models [196] for predicting the amount of material deposited in terms of parameters of the solution and the spin process. Future work in this program should extend to study the effects of promoters. The active phases can still be applied by the dipping method because that ensures the selectivity of the active phases by the support. Any stray solution can be eliminated by high speed spinning, and that will also help the macroscopic spreading of the applied material on the supports (at between 1 and several monolayer coverage). The surface regions of the active phases can then be quantified, at least on a relative scale, by XPS as done in work in this thesis. But continued 154 effort is needed to assess standards, and to develop other methods (e.g. RBS), which can give absolute loadings. 8.2.2 Further studies on sulfided Mo03/A10x Studies on MoS2-x, S22" and Sn2" Although a species identified as M0S2-X was recognized by XPS from the sulfided 450°C-calcined sample, and it may have some role in the HDN activity, no Mo species with less than the +4 oxidation state could be detected in the work in Chapter 5. The reason may be due to its relatively low concentration, and perhaps some overlap with the S2s spectrum. By using the monchromatized x-ray source, higher resolution Mo3d spectra may be able to discriminate this species from the entire Mo3d envelope, and thereby add to our understanding for the M0S2-X species. Sn2" and S22" species are commonly identified by XPS from sulfided M0O3 catalysts but knowledge of the importance of these species is still limited. The study in Chapter 5 indicates that the structure for the former species cannot be determined by XPS alone. Vibrational spectroscopic techniques, such as IRS, Raman spectroscopy and high-resolution electron loss spectroscopy may help to determine the detailed structure of S n 2\ and give more insight into its role in the sulfidation mechanism for MoGVAlOx. Although the formation of S22" at ion-bombarded M0S2 single crystals is believed to involve rearrangement of dangling bonds from pairs of S atoms after some Mo removal [85], nevertheless the direct formation of S22~ during sulfidation is not well studied. Although this species has been reported to contribute active sites for H 2 splitting during HDN reaction [167], further studies are needed to define any relationships between S22" and HDN activity. Since S22" appears as the major S species on the sulfided 350°C-155 calcined sample (Chapter 5), more detailed studies for this sample may provide specific information for the formation of this species. Analysis of Ols peak In Chapter 5, the uncalcined and 350°C-calcined Mo03/A10x samples are considered to have low Mo dispersion, and they produce smaller amounts of Mo(+4) sulfide after sulfidation. In fact, the detailed Mo dispersions are different for the uncalcined and 350°C-calcined samples (results in Chapter 4 show a lower Mo/Al ratio for the uncalcined sample) and that may be a factor for reason why different S species are formed during the sulfidation. The meaning of the S/Mo ratio measured in this study is somewhat obscured by the proposed involvement of the Sn2" species. But since Ols spectra have been indicated to show a 0.5 eV difference between the A10x and M0O3 forms [170], measurements of Ols peak at high energy resolution (i.e. with the monochromator) may be able to separate these components. If so, during the sulfidation it may be possible to indicate the amount of Mo that is bonded to O, and thereby follow changes in the relative number of Mo-S bonds. 8.2.3 Further studies on Mo nitride catalysts Promoter effects For high-area supported Mo nitride catalysts, other components, like Ni, Co are added as promoters. Owing to the insulation of the supports, and relatively low concentrations for the promoters (e.g. 3 wt.%), surfaces of high-area catalysts have seldom been characterized in detail. Following the philosophy developed in this thesis, new information should be gained by studying interactions between promoters and Mo, and between promoters and supports, during calcination processes. New indications were made in this work for interaction between Mo and alumina 156 supports by following changes in the Auger parameter. That appears plausible but independent tests should be made, for example using vibrational spectroscopy and perhaps static SIMS for assessing direct Mo-O-Al bonding [27]. Like all supported catalysts, high dispersion is required for the active species (i.e. Mo, Co, Ni etc.). More detailed study can be undertaken to investigate the dispersion of these elements with scanning Auger microscopy (SAM), at least on the model samples, where charging is not a problem. Exploratory study can also be done with atomic force microscopy (AFM). Observations from such techniques on elemental distribution and surface morphology (e.g. with scanning electron microscopy with a field-emission gun) can then complement observations with XPS and SIMS. Although the support is initially present for the practical reason of holding the catalyst, cases are known where interactions between the support and the Mo influence the performance of the resulting catalyst. Much more needs to be known about this phenomenon. For example, silica is not used in hydrotreating catalysts because it is believed that the Mo is less dispersed on this support [55]. However, new information on the role of supports could be gained by investigating variations of dispersion on model catalysts like Ni/Mo03/A10x and Ni/Mo/Si02 for comparable treatments. Further, sequential study should identify how or whether the promoters affect the subsequent nitridation process, as well as the HDN catalytic activity. There has been evidence that the sequence of impregnation of promoters can affect the performance of the final working catalyst [54], but no direct information is available for how this occurs. New modeling studies, following work started in this thesis, could provide information to help explain such effects. 157 Further studies on nitrided Mo03/A10x and Mo0 3/Si0 2 Supported Mo nitride catalysts are still a new topic for HDN reactivity. In this work, the MoO3/AlOx(550) system was only nitrided to a maximum of 550°C because of the relatively low melting point of the support. Planar sapphire could with advantage be used as the support material, then a direct comparison could be made between nitrided Mo03/alumina and Mo03/silica by raising the treatment temperature to 700°C in the former case. Also, in the study described in Chapter 7, surface regions of the materials formed after each stage of the heating have been analyzed, but they appear to have considerable heterogeneity. Information on the distribution of the different components with depth should be obtained by making new measurements with angle dependent XPS (ADXPS) and with SIMS depth profiling. Passivation effects Commonly, high-area Mo nitride catalysts are passivated (e.g. in 1% 0 2 at room temperature) to protect active components prior to use [21]. It is believed that this passivation process forms a slightly oxidized layer which protects the air-sensitive Mo nitrides. Prior to the HDN process, the passivated catalysts are activated in-situ by NH 3 (or H2). More generally it would be useful to directly assess the effects of. these procedures. In particular, such a study could illuminate whether or not the passivation process is effective for the catalyst protection, and whether the NH 3 (or H2) activation step brings the catalyst back to its pre-passivated state. HDN activity studies Although high-area supported catalysts are not ideal for surface analytical techniques, the philosophy of the work reported in this thesis supposes that it is still possible to gain new insight by studying equivalent model planar catalysts. The nitridation of MoQ3 as reported in this thesis 158 has to date only been able to focus on the characterization of surface composition, but two further directions need developing. First, complementary work with x-ray diffraction (XRD) is needed in order to clarify the origin of the discrepancies apparent between my observations and those of Choi et al. [21]. Second, it will be important to make direct catalytic studies on the characterized surface nitrides studied in Chapter 6 and 7. Indeed it has been reported that the catalytic activity of the various Mo nitrides may also be influenced by the presence of oxygen in oxynitride form [186]. From the information in Figure 6.9, it is seen that the Mo chemical state varies substantially with the temperature of treatment for the M0O3/NH3 reaction but knowledge is still needed for the activity in HDN catalysis for the various surface compounds that appear dominant at each stage in the reaction studied. 159 References 1 J.A. Moulijn, R.A. Sheldon, H. van Bekkum, and P.W.N.M. van Leeuwen, in: J.A. Moulijn, P.W.N.M. van Leeuwen, and R.A. van Santen (Eds.), Catalysis, An Integrated Approach to Homogeneous, Heterogeneous and Industrial Catalysis (Elsevier, Amsterdam, 1993). 2 S.E. Voltz, D.M. Naca, S.M. Jacob, and V.W. Weekman, jun., Ind. Eng. Chem., Proc. Des. Dev. 11 (1972) 261. 3 P. Grange, Catal. Rev. Sci. Eng. 21 (1980) 135. 4 M. J. Ledouz, Adv. Catal. 7 (1985) 125. 5 S. Tasztelan, T. des Courieres, and M. Breysse, Catal. Today 10 (1991) 433. 6 J.R. Katzer and R. Sivasubramanian, Catal. Rev. Sci. Eng. 20 (1979) 155. 7 E. Furimsky, R. Ranganathan, and B.I. Parsons, 5th Canadian Symposium on Catalysis, Calagary, Alberta, 1977. 8 T.C. Ho, Catal. Rev. Sci. Eng. 30 (1988) 117. 9 L.C. Gutberlet and R.J. Bertolacini, Ind. Eng. Chem., Proc. Res. Dev., 22 (1983) 246. 10 H. Adkins, L.F. Kuick, M. Farlow, and B. Wojcik, J.Am. Chem. Soc. 56 (1934) 2425. 11 J. Shabtai, L. Veluswamy, and A. G. Oblad, ACS Div. Fuel Chem. Prepr. 23 (1978) 114. 12 C. Moreau, R. Durand, N. Zmimita, and P. Geneste, J. Catal. 112 (1988) 411. 13 O. Weisser and S. Landa, in: Sulfided Catalysts, Their Properties and Application (Pergamon, Oxford, 1973). 14 G. Delvaux, P. Grange and B. Delmon, J. Catal. 56 (1979) 99. 15 R. Prada Silvy, J.L.G. Fierro, P. Grange and B. Delmon, in: B. Delmon, P. Grange, P.A. Jacobs and G. Poncelet (Eds.), Preparation of Catalysts IV (Elsevier, Amsterdam, 1987) p. 605. 16 M.V. Bhinde, S. Shih, R. Xawadski, J.R. Katzer, and H. Kwart, in: H.F. Barry and P.C.M. Mitchell (Eds.), Chemistry and Uses of Molybdenum: Processings of the Third International Conference (Climax Molybdenum, Ann Arbor, 1979) p. 184. 17 R. Pins, V.H.J. De Beer, and G.A. Somorjai, Catal. Rev. Sci. Eng. 31 (1989) 1. 18 R.R. Chianelli, J.E. Lyons and G.A. Mills, Catal. Today 22 (1994) 361. 160 19 J. C. Schlatter, S.T. Oyama, J.E. Metcalfe and J.M. Lambert, Ind. Eng. Chem. Res. 27 (1988) 1648. 20 K.S. Lee, H. Abe, J.A. Reimer and A.T. Bell, J. Catal. 139 (1993) 34. 21 J.-G. Choi, J.R. Brenner, CW. Colling, B.G. Demczyk, J.L. Dunning and L.T. Thompson, Catal. Today 15 (1992) 201. 22 H. Abe and A.T. Bell, Catal. Lett. 18 (1993) 1. 23 E.J. Markel and J.W. Van Zee, J. Catal. 126 (1990) 643. 24 D. Briggs and M.P. Seah (Eds.), Practical Surface Analysis by Auger and X-Ray Photoelectron Spectroscopy Vol. 1 (Wiley, New York, 1990). 25 M.A. Chester, J. Electron Spectrosc. 38 (1986) 123. 26 B.A. Horrell and D.L. Cocke, Catal. Rev. Sci. Eng. 29 (1987) 447. 27 H.J. Borg and J.W. Niemantsverdriet, Catalysis 11 (1994) 1. 28 J.C.J. Bart and G. Vlaic, Adv. Catal. 35 (1987) 1. 29 N.W. Hurst, S.J. Gentry, A. Jones and B.D. McNicol, Catal. Rev. Sci. Eng. 24 (1982) 233. 30 A.T. Bell, NMR Techniques in Catalysis (M. Dekker, New York, 1994). 31 J.B. Cohen and L.H. Schwartz, Diffraction from Materials (Academic Press, New York, 1977). 32 R.T. Lewis and M.A. Kelly, J. Electron Spectrosc. 20 (1980) 105. 33 H. Windawi and C D . Wagner, in H. Windawi and F.F.L. Ho (Eds.), Applied Electron Spectroscopy for Chemical Analysis (John Wiley and Sons, New York, 1982) pp. 191-208. 34 J.S. Brinen, J. Electron Spectrosc. 5 (1974) 377. 35 H. Windawi, J. Electron Spectrosc. 22 (1981) 373. 36 C D . Wagner, L.H. Gale and R.H. Raymond, Anal. Chem. 51 (1979) 466. 37 R.H. West and J.E. Castle, Surf, and Interf. Anal. 4 (1982) 68. 38 L.H. Dubois, P.K. Hansma, and G.A. Somorjai, Appl. Surf. Sci. 6 (1980) 173. 161 39 D.N. Belton, Y.-M. Sun, and J.M. White, J. Phys. Chem. 88 (1984) 5172. 40 N.S. Mclntyre, T.C. Chan, P.A. Spevack and J.R. Brown, Appl. Catal. 63 (1990) 391. 41 L.M. Eshelman, A.M. de Jong and J.W. Niemantsverdriet, Catal. Lett. 10 (1991) 201. 42 H.J. Borg, L.C.A. van den Oetelaar, L.J. van Ijzendoorn, J.W. Niemantsverdriet, J. Vac. Sci. Tehcnol. A 10 (1992) 2737. 43 A.M. de Jong, H.J. Borg, L.J. van IJzendoorn, V.G.F.M. Soudant, V.H.J, de Beer, J.A.R. van Veen, and J.W. Niemantsverdriet, J. Phys. Chem. 97(1993) 6477. 44 D.L. Cocke, E.D. Johnson, and R.P. Merrill, Catal. Rev. Sci. Eng. 26 (1984) 163. 45 B.G. Frederick, G. Apai, and T.N. Rhodin, J. Am. Chem. Soc. 109 (1987) 4797. 46 D.N. Belton and S.J. Schmieg, Surf. Sci. 199 (1988) 518. 47 P. Gajardo, D. Pirotte, C. Defosse, P. Grange, and B. Delmon, J. Electron. Spectrosc. 17 (1979) 121. 48 Y. Okamoto, T. Imanaka, and S. Teranishi, J. Phys. Chem. 85 (1981) 3798. 49 J.A.R. Van Veen and P.A.J.M. Hendriks, Polyhedron 5 (1986) 75. 50 L. Wang and W.K. Hall, J. Catal. 77 (1982) 232. 51 S. Kasztelan, J. Grimblot, J.P. Bonnelle, E. Payen, H. Toulhoat and Y. Jacquin, Appl. Catal. 7 (1983)91. 52 P.A. Spevack and N.S. Mclntyre, J. Phys. Chem. 97 (1993) 11020. 53 B.A. Horrell and D.L. Cocke, 116 (1989) 449. 54 B. Delmon, P. Grange, P. Jacobs and G. Poncelet (Eds.), Preparation of Catalysts (Elsevier, Amsterdam, 1978). 55 N. Yamagata, Y. Owada, S. Okazaki, and K. Tanabe, J. Catal. 47 (1977) 358. 56 J. Grimblot, J.P. Bonnelle, and J.P. Beaufils, J. Electron Spectrosc. 8 (1976) 789. 57 R.L. Chin and D.M. Hercules, J. Phys. Chem. 86 (1982) 360. 58 K. S. Chung and F.E. Massoth, J. Catal. 64 (1980) 3079. 59 A. Cimino and B.A. De Angelis, J. Catal. 36 (1975) 11. 162 60 A.W. Miller, W. Atkinson, M. Barber, and P. Swift, J. Catal. 22 (1971) 140. 61 P. Ratnasamy, J. Catal. 40 (1975) 137. 62 D.S. Zingg, L.E. Makovsky, R.E. Tischer, F.R. Brown, and D.M. Hercules, J. Phys. Chem. 84(1980) 2898. 63 P. Gajardo, R.I. Declerak-Grimee, G. Delvaux, P. Olodo, J.M. Zabala, P. Canesson, P. Grange, and B. Delmon, J. Less Common Metals 54 (1977) 311. 64 A.W. Miller, W. Atkinson, M. Barber, and P. Swift, J. Catal. 22 (1971) 140. 65 Y. Iwasawa, and S. Ogasawara, J. Chem. Soc. Faraday Trans. I 75 (1979) 1465. 66 G. Muralidhar, B.E. Concha, G.L. Bartholomew, and C.H. Bartholomew, J. Catal. 89 (1984) 274. 67 N.K. Nag, J. Phys. Chem. 91 (1987) 2324. 68 H. Shimada, T.Sata, Y. Yoshimura, J. Hiraishi, and A. Nishijima, J. Catal. 110 (1988) 275. 69 Y. Okamoto and T. Imanaka, J. Phy. Chem. 92(1988)7102. 70 N.-Y. Tops0e and H. Tops0e, J. Catal. 139(1993) 631. 71 P. Ratnasamy and H. Knozinger, J. Catal. 54 (1978) 1311. 72 B.C. Gates, J.R. Katzer, and G.C.A. Schuit, Chemistry of Catalytic Processes (McGraw-Hill, New York, 1979) p 390. 73 F.E. Massoth, Adv. Catal. 27 (1978) 265. 74 H. Knozinger and P. Ratnasamy, Catal. Rev. Sci. Eng. 17 (1978) 31. 75 V.H.J. deBeer, M.J.M. Van der Aalst, C.J. Machids, and G.C.A. Schuit, J. Catal. 43 (1976) 78. 76 Y. Okamoto, H. Tomioka, T. Imanaka, and S. Teranishi, Prepr. Int. Congr. Catal. 7th (1980) A43. 77 A. Castellan, J.C.J. Bart, A. Vaghi and N. Giordano, J. Catal. 42 (1976) 162. 78 H. Jeziorowski, H. Knozinger, P. Grange, P. Gajardo, J. Phys. Chem. 84 (1980) 18250. 79 CP. Cheng and G.L. Schrader, J. Catal. 60 (1979) 276. 163 ) 80 F.P.J.M. Kerkhof, J.A. Moulijn and A. Heeres, J. Electron Spectrosc. 14 (1978) 453. 81 Yu.V. Plyuto, I.V. Babich, I.V. Plyuto, A.D. Van Langeveld and J.A. Moulijn, Appl. Surf. Sci. 119(1997) 11. 82 M.A. Banares, J.L.G. Fierro, Catal. Lett. 17 (1993) 205. 83 K. Suzuki, T. Hayakawa, M. Shimizu, and K. Takehira, Catal. Lett. 30 (1995) 159. 84 F. Solymosi, A. Erdohelyi, and A. Szoke, Catal. Lett. 32 (1995) 43. 85 J.R. Lince, T.B. Stewart, M.M. Hills, P.D. Fleischauer, J.A. Yarmoff, and A. Tableb-Ibrahimi, Surf. Sci. 210 (1989) 387. 86 J.R. Lince, D.J. Carre, and P. Fleischauer, Langmuir 2 (1986) 805. 87 N.S. Mclntyre, P.A. Spevack, G. Beamson, and Briggs, Surf. Sci. Lett. 237 (1990) L390. 88 C.B. Roxlo, H.W. Deckman, J. Gland, S.D. Cameron and R.R. Chianelli, Science 235 (1987) 1629. 89 A. Shuxian, W.K. Hall, G. Ertl, and H. Knozinger, J. Catal. 100 (1986) 167. 90 R.H. Williams and A.J. McEvoy, J. Phys. D4 (1971) 456. 91 P. Ratnasamy and S. Sivasanker, Catal. Rev. Sci. Eng. 22 (1980) 401. 92 W.K. Hall, in: R. Vanselow and R. Howe (Eds.), The Chemistry and Physics of Solid Surfaces Vol. VI (Springer-Verlag, New York , 1986). 93 W.K. Hall, in: H.F. Barry and P.C.H. Mitchell (Eds.), Chemistry and Uses of Molybdenumm: Proceedings of the Climax Fourth International Conference (Climax Molybdenum Co., Ann Arbor, 1982) p 224. 94 Y. Okamoto, H. Tomioka, Y. Katoh, T. Imanaka, and S. Teranishi, J. Phys. Chem. 84 (1980)1833. 95 P.A. Spevack and N.S. Mclntyre, J. Phys. Chem. 97 (1993) 11031. 96 B. Scheffer, P. Arnoldy, and J. Moulijn, J. Catal. 112(1988) 516. 97 P. Arnoldy, J.A.M. Van Den Heijkant, G.D. De Bok, and J.A. Moulijn, J. Catal. 92 (1985)35. 98 B.S. Clausen, H. Tops0e, R. Candia, J. Villadsen, B. Lengeler, J. Als-Nielsen, and F. Christensen, J. Phys. Chem. 85 (1981) 3868. 164 99 N.-Y. Tops0e, J. Catal. 64 (1980) 235.' 100 W. Zmierczak, Q. Qader, and F.E. Massoth, J. Catal. 106 (1987) 65. 101 T.F. Hayden and J.A. Dumesic, J. Catal. 103 (1987) 385. 102 T.F. Hayden, J.A. Dumesic, R.D. Sherwood, and T.K. Baker, J. Catal. 105 (1987) 299. 103 R.J.H. Voorhoeve and J.C.M. Stuiver, J. Catal. 23 (1971) 243. 104 D.J. Sajkowski and S.T. Oyama, Appl. Catal. 34(1996) 339. 105 L.E. Toth, Transition Metal Carbides and Nitrides (Academic Press, New York, 1971). 106 L. Volpe and M. Boudart, J. Solid State Chem. 59 (1985) 332. 107 L. Volpe and M. Boudart, Catal. Rev. Sci. Eng. 27 (1985) 515. 108 C.H. Jaggers, J.N. Michaels, and A.M. Stacy, Chem. Mat. 2 (1990) 150. 109 B.G. Demczyk, J.-G. Choi, and L.T. Thompson, Appl. Surf. Sci. 78 (1994) 63. 110 P.A. Aegerter, W.W.C. Quigley, G.J. Simpson, D.D. Ziegler, J.W. Logan, K.R. McCrea, S. Glazier, M.E. Bussell, J. Catal. 164 (1996) 109. 111 N. Nagao, T. Miyao, and T. Tuboi, Catal. Lett. 18 (1993) 9. 112 CW. Colling and L.T. Thompson, J. Catal. 146 (1994) 193. 113 H. Hertz, Annal. Phys. 31 (1887) 982. 114 A. Einstein, Anal. Phys. 17 (1905) 132! 115 H. Robinson and W.F. Rawlinson, Phil. Mag. 28 (1941) 277. 116 K. Siegbahn, Science 217 (1982) 4555. 117 K. Siegbahn, C.N. Nording, A. Fahlman, R. Nordberg, K. Hamrin, J. Hedman, G. Hohansson, T. Bermark, S.E. Karlsson, I. Lindgren and B. Lindberg, ESCA: Atomic Molecular and Solid State Structure Studied by Means of Electron Spectroscopy (Almqvist and Wiksells, Uppsala, 1960). 118 E. Bauer, Vacuum 22 (1972) 539. 119 C D . Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder and G.E. Muilerberg (Eds), Handbook ofX-Ray Photoelectron Spectroscopy (Perkin-Elmer, Minne, 1979). 165 120 M.P. Seah and W.A. Dench, Surf. Interf. Anal. 1 (1979) 2. 121 G.A. Somorjai, Chemistry in Two Dimensions: Surfaces (Cornell University, Ithaca, New York, 1981). 122 B.D. Ratner, T.A. Horbett, D. Shuttleworth and H.R. Thomas, J. Collid Interf. Sci. 83 (1981)630. 123 G. Ertl and J. Kiippers, Low Energy Electron and Surface Chemistry (VCR Verlegs, Weinheim, 1985). 124 MAX200 Manual (Leybold, Koln, Germany). 125 J.D. Andrade, in: J.D. Andrade (Eds.), Surface and Interfacial Aspects of Biomedical Polymer (Plenum, New York, 1985) p.l 14. 126 D.A. Shirley, Phys. Rev. B5 ( 1972) 4709. 127 S. Tougaard, J. Electron Spectrosc. 52 (1990) 243. 128 P. Swift, Surf. Interf. Anal. 4 (1982) 47. 129 A.J. Pertsin and Yu. M. Pashunin, Appl. Surf. Sci. 44 (1990) 171. 130 Y.L. Leung, M.Y. Zhou, P.C. Wong, K.A.R. Mitchell and T. Foster, Appl. Surf. Sci. 59 (1992)13. 131 A. Beninghoven, F.G. Rudenauer and H.W. Werner, Secondary Ion Mass Spectrometry (John Wiley & Sons, New York, 1987). 132 J.C. Vickerman, Chem. in Brit. 74 (1987) 969. 133 S.E. Asher, in: B.G. Yacobi, D.B. Holt, and L.L. Kazmerski (Eds.), Microanalysis of Solid (Plenum, New York, 1994) p. 150. 134 W. Brandt and R. Laubert, Nucl. Instrum. Methods, 47 (1967) 201. 135 P. Sigmund, Phys. Rev., 184 (1969) 383. 136 M. W. Thompson, Philos. Mag., 18 (1968) 377. 137 J.C. Vickerman, A. Brown and N.M. Reed, Secondary Ion Mass Spectrometry, Principles and Applications (Claredon Press, Oxford, 1989). 138 A. Benninghoven, Z. Phys. 230 (1970) 403, 139 E. Zinner, J. Electrochem. Soc. 130 (1983) 199. 166 140 P.D. Prewett and D.K. Jeffries, Inst. Phys. Conf. Ser. 54 (1980) 316. 141 G. Slodzian and J.-F. Hennequin, Compt. Rend. B263 (2966) 1246. 142 H.E. Roosendaal, Sputtering by Particle Bombardment I (Springer, Belrlin, 1981) p.219. 143 A. Benninghoven, Surf. Sci. 53 (1975) 596. 144 VG Scientific SIMS manual. 145 VG Scientific DP5 duoplasmatron ion gun manual. 146 P.H. Dawson, Quadrupole Mass Spectrometry and Its Applications (Elsevier, New York, 1976). 147 P.J. Goodhew and F.J. Humphreys, Electron Microscopy and Analysis (Taylor & Francis, New York, 1988) p. 113. 148 B.L. Grabriel, Biological Secondary Electron Microscopy (Van Nostrand Reinhold, New York, 1982). 149 N.S. Mclntyre, D.D. Johnston, L.L. Coatsworth, R.D. Davidson and J.R. Brown, Surf. Interf. Anal. 9 (1986) 253. 150 N.S. Mclntyre, T.C. Chan, P.A. Spevack and J.R. Brown, in: M.L. Occelli and R.G. Anthony (Eds.) Hydrotreating Catalysts, (Elsevier, Amsterdam, 1989) p. 187. 151 R.Z. Bachrach, S.A. Flodstrom, R.S. Bauer, S.B.M. Hagstrom and D.J. Chadi, J. Vac. Sci. Technol. 15 (1978)488. 152 J.A. Taylor, J. Vac. Sci. Technol. 20 (1982) 751. 153 A. Biaconi, R.Z. Bachrach, S.B.M. Hagstrom and S.A. Flodstrom, Phys. Rev. B 19 (1979) 2847. 154 H. Knozinger, in: B. Imelik, C. Naccache, G. Coudurier, Y. Ben Taarit and J.C. Vedrine (Eds.) Catalysis by Acids and Bases (Elsevier, Amsterdam, 1985) p. 111. 155 T. Edmonds and P.C.H. Mitchell, J. Catal. 64 (1980) 492. 156 J. Medema, C. Van Stam, V.H.J. De Beer, A.J.A. Konings and D.C. Koningsberger, J. Catal. 53 (1978) 386. 157 J. Bachelier, M.J. Tilliette, J.C. Duchet and D. Cornet, J. Catal. 76 (1982) 300. 158 M.J. Ledoux, P.E. Puges, and G. Maire, J. Catal. 76 (1982) 285. 167 159 F.E. Massoth, J. Catal. 36 (1975) 164. 160 N.-Y. Tops0e and H. Tops0e, J. Catal. 139 (1993) 641. 161 H. Tops0e and B.S. Clausen, Catal. Rev. Sci. Eng. 26(1984) 395. 162 L. Portela, P. Grange and B. Delmon, J. Catal. 156 (1995) 243. 163 W. Griinert, A. Yu Stakheev, R. Feldhaus, K. Anders, E.S. Shpiro, and K.M. Minachev, J. Phys. Chem. 95 (1991) 1323. 164 A. Muller and E. Diemann, in: H.J. Emeleus and A.G. Sharpe (Eds), Advances in Inorganic Chemistry, Vol. 31 (Academic Press, New York, 1987) p.89. 165 J. Sonnemans, W.J. Neyens and P. Mars, J. Catal. 34 (1974) 230. 166 A.Kherbeche, R. Hubaut, J.P. Bonnelle, and J. Grimblot, J. Catal. 131 (1991) 204. 167 J. Polz, H. Zellinger, B. Muller, H. Knozinger, J. Catal. 120 (1989) 22. 168 D.J. Sajkowski and S.T. Oyama, Appl. Catal. 134 (1996) 339. 169 M. Maoujoud, M. Jardinier-Offergeld and F. Bouillon, Appl. Surf. Sci. 64 (1993) 81. 170 CD. Wagner, D.A. Zatko and R.H. Raymond, Anal. Chem. 52 (1980) 1445. 171 P.A. Spevack and N.S. Mclntyre, J. Phys. Chem. 96 (1992) 9029. 172 M. Grunze, CR. Brundle and D. Tomanek, Surf. Sci. 119 (1982) 133. 173 C. Egawa, S. Naito and K. Tamaru, Surf. Sci. 131 (1983) 49. 174 R.M. Henry, B.W. Walker and P.C Stair, Surf. Sci. 155 (1985) 732. 175 J.-G. Choi, D. Choi and L.T. Thompson, Appl. Surf. Sci. 108 (1997) 103. 176 M.D. Lyutaya, Poroshk. Metall. 3 (1978)60. 177 T.A. Patterson, J.C. Carve, D.E. Leyden and D.M. Hercules, J. Phys. Chem. 80 (1976) 1700. 178 L. Ramqvist, K. Hamrin, G. Johansson, A. Fahlmann and C. Nordling, J. Phys. Chem. Solids 30(1969) 1835. 179 F.A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry (Wiley, New York, 1988). 180 P. Schwarzkopf and R. Kieffer, Refractory Hard Metals (Macmillan, New York, 1953). 168 181 Y.C. Lu and CR. Clayton, Corros. Sci. 29 (1989) 927. 182 A. Katrib, P. Leflaive, L. Hilaire and G. Maire, Catal. Letters 38 (1996) 95. 183 N. Schonberg, Acta Chem. Scand. 8 (1954) 204. 184 G. Hagg, Z. Phys. Chem., Abt. B7 (1930) 339. 185 J.-G. Choi, R.L. Curl and L.T. Thompson, J. Catal. 146 (1994) 218. 186 R. Ochoa, X.X. Bi, A.M. Rao and P.C Eklund, in: S.T. Oyama (Eds), The Chemistry of Transition Metal Carbides and Nitrides (Blackie Academic, Glasgow, 1996). 187 M. Nagai, T. Miyao and T. Tuboi, Catal. Lett. 18 (1993) 9. 188 K. Burger, F. Tschismarov and H. Ebel, J. Electron Spectrosc. 10 (1977) 461. 189 E. Atanassova and A. Paskaleva, Appl. Surf. Sci. 103 (1996) 359. 190 CD. Wagner, D.E. Passoja, H.F. Hillery, T.G. Kinisky, H.A. Six, W.T. Jansen and J.A. Taylor, J. Vac. Sci. Technol. 21 (1982) 933. 191 G. Dufour, F. Rochet, H. Roulet and H. Sirotti, Surf. Sci. 304 (1994) 33. 192 J.L. Bischoff, F. Lutz, D. Bolmont and L. Kubler, Surf. Sci. 251/252 (1991) 170. 193 A. Castellan, J.C Bart, A. Vaghi and N. Giodano, J. Catal. 42 (1976) 162. 194 W.E. Swartz and D.M. Hercules, Anal. Chem. 43 (1971) 1774. 195 M. Nagai, T. Miyao and T. Kusagawa, Studies Surf. Sci. Catal. 75 (1993) 1939. 196 R.M. van Hardeveld, P.L.J. Gunter, L.J. van Ijzendoorn, W. Wieldraaijer, E.W. Kuipers. and J.W. Niemantsverdriet, Appl. Surf. Sci. 84 (1995) 339. 169 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items