1 GROWTH PROPERTIES AND REACTIVITY OF OXYGEN PHASES ON PLATINUM (111) AND PALLADIUM (111) By SUNIL POONDI DEVARAJAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 2012 Sunil Poondi Devarajan
3 Dedicated t o my dear p arents
4 ACKNOWLEDGMENTS I would like to thank my supervisory committee chair and principal investigator D r. Jason Weaver, for providing encouragement and unending support, without which this work would not have been possible. I thank Dr. Helena Hagelin Weaver for being on my supervisory committee and for the encouragement and support she offered during the co urse of my program. I also want to thank Dr. Kirk Ziegler and Dr. Amlan Biswas for being on the supervisory committee and supporting my research. My sincere thanks go out to Dr. Robert Madix and Dr. Cynthia Friend, who welcomed me to their lab at Harvard U niversity, as a Research Scientist, to learn the nuances of STM operation and also conducting surface science experiments in general. Special thanks to their post doctoral students Weiwei Gao and Ling Zhou for all their help during my training at Harvard. I would like to thank Jose Hinojosa, Heywood Kan, Brad Shumbera and Can Hakanoglu for the delightful experiences in and out of the lab. I am grateful to the Department of Chemical Engineering, University of Florida Graduate Alumni Fellowship and again Dr. Jason Weaver for providing assistantship during my doctoral program. I also thank the National Science Foundation and the Department of Energy for research funding. Lastly, I would like to thank my friends and folks who have stood by me through the ups an d downs over the last several years.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF FIGURES ................................ ................................ ................................ .......... 7 ABSTRACT ................................ ................................ ................................ ..................... 9 CHAPTER 1 INTRODUCTORY REMARKS ................................ ................................ ................ 11 Background and Motivation ................................ ................................ .................... 11 Platinum Study ................................ ................................ ................................ 13 Palladium Study ................................ ................................ ............................... 16 Experimental Techniques Used ................................ ................................ .............. 19 2 SCANNING TUNNELING MICROSCOPY IN UHV ................................ ................. 23 UHV STM System Development ................................ ................................ ............. 23 Basis for Scanning Tunneling Microsco py ................................ .............................. 25 Scanning Tunneling Microscopy: Imaging in Ultra High Vacuum ............................ 27 Tip Sample Approach/Retract ................................ ................................ .......... 27 Imaging Modes ................................ ................................ ................................ 28 Challenges of UHV STM System Design ................................ ......................... 29 Design and Implementation Aspects of UHV STM System ................................ ..... 31 Vibration Isolation ................................ ................................ ............................. 31 Description and Salient Features of UHV STM Apparatus ............................... 35 Common Problems with the UHV STM System and Remedies .............................. 40 Tunneling and Scan Drive Wires ................................ ................................ ...... 40 Sam ple and Tip Condition ................................ ................................ ................ 41 Electronic Noise and Feedback Oscillation ................................ ...................... 43 Special Operating Procedures for UHV STM System ................................ ...... 45 3 STM STUDY OF HIGH COVERAGE STRUCTURES OF ATOMIC OXYGEN ON PLATINUM (111): P(21) AND PLATINUM OXIDE CHAIN STRUCTURES .... 61 Introduction ................................ ................................ ................................ ............. 61 Experimental Details ................................ ................................ ............................... 66 Results and Discussion ................................ ................................ ........................... 68 Thermal Des orption of O Pt(111) ................................ ................................ ..... 68 STM Study of O Pt(111) System ................................ ................................ ...... 69 2 State: p(21) Striped Structure ................................ ............................. 70 1 State: One Dimensional (1D) Pt Oxide Chains ................................ ..... 72 1 State: Chain Br anching and Formation of an Interconnected Network ................................ ................................ ................ 75
6 Implications of Pt Oxide Chain Formation ................................ ........................ 79 Summary ................................ ................................ ................................ ................ 80 4 FACILE C H BOND CLEAVAGE AND DEEP OXIDATION OF PROPANE ON A PALLADIUM OXIDE (101) THIN FILM ................................ ................................ .... 90 Introduction ................................ ................................ ................................ ............. 90 Experimental Methods ................................ ................................ ............................ 93 Results and Discussion ................................ ................................ ........................... 97 Structure of PdO(101) Thin Film on Pd(111) ................................ .................... 97 TPRS from Propane Saturated PdO(101) ................................ ........................ 98 Propane Adsorption on PdO(101) as a Function of Coverage ....................... 102 Desorption and Reaction Yields: Selective Dissociation from the 1 State. ... 106 Kinetic Analysis of Propane Dissociation: Trapping Mediated Mechanism .... 108 Comparison with Other Investigations ................................ ............................ 113 Summary ................................ ................................ ................................ .............. 116 5 CONCLUSIONS ................................ ................................ ................................ ... 125 LIST OF REFERENCES ................................ ................................ ............................. 128 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 135
7 L IST OF FIGURES Figure page 2 1 Schematic representation of the phenomenon of quantum mechanical tunneling.. ................................ ................................ ................................ ........... 48 2 2 Schematic for STM operation using tunneling me chanism with feedback loop 49 2 3 Schematic showing the constant current scanning mode for scanning tunneling microscopy ................................ ................................ ......................... 50 2 4 Schematic showing the constant height scanning mode for scanning tunneling microscopy ................................ ................................ ......................... 50 2 5 Schematic of t unneling stage of the microscope for vibration analysis .............. 51 2 6 Transmission factor for vibration of the tunneling junction caused by forced oscillations of the vibration isolation stage ................................ ......................... 52 2 7 Schematic of fi rst isolation stage of STM for vibration analysis ......................... 53 2 8 Transmission factor for vibration of the isolation stage caused by forced vibration of the sample manipulator ................................ ................................ ... 54 2 9 Overall transmission function for vibration of tunneling junction ........................ 55 2 10 Schematic of UHV STM system ................................ ................................ ........ 56 2 11 Schematic showing the sample holder assembly ................................ .............. 57 2 12 Vertical transmissibility for Newport I 2000 series laminar flow vibration isolation system ................................ ................................ ................................ 58 2 13 Schematic depicting cross sectional view of the STM assembly ....................... 59 2 14 Noise power spectrum indicating large noise around 60 Hz. .............................. 60 2 15 Noise power spectrum indicating reduced noise around 60 Hz. ......................... 60 3 1 O 2 TPD spectra obtained from Pt(111) for various initial a tomic oxygen coverages at 450 K ................................ ................................ ............................ 82 3 2 STM images of Pt(111) covered with 0.30 ML of oxygen atoms prepared by dissociatively adsorbing NO 2 at 450 K ................................ ............................... 83 3 3 STM images of Pt(111) covered with 0.40 ML of oxygen atoms prepared by dissociatively adsorbing NO 2 at 573 K ................................ ............................... 84
8 3 4 STM images of Pt(111) covered with 0.46 ML of oxygen atoms prepared by dissociatively adsorbing NO 2 at 450 K ................................ ............................... 85 3 5 STM image of of Pt(111) covered with 0.58 ML of oxygen atoms prepared by dissociatively adsorbing NO 2 at 450 K ................................ ............................... 86 3 6 Top and side view illustrations of a Pt oxide chain structure that forms on the Pt(111) p(21) O surface as determined from DFT ................................ ........... 87 3 7 STM images of Pt(111) covered with 0.71 ML of oxygen atoms prepared at 450 K ................................ ................................ ................................ ................. 88 3 8 High resolution image of a hexagonally shaped chain superstructure ............... 89 4 1 Top and side vie w schematics of the PdO(101) thin film structure ................... 118 4 2 TPD spectra of masses 2, 18, 28, 29 and 44 amu from PdO(101) thin film after pro pane saturation at 85 K ................................ ................................ ....... 119 4 3 Propane TPD spec tra from PdO(101) and Pd(111) as a function of the initial propane coverage at 85 K ................................ ................................ ................ 120 4 4 TPRS desorption and reaction yields as a function of propane exposure on PdO(101) thin film at 85 K ................................ ................................ ................ 121 4 5 Propyl group coverage as a function of the propane exposure for adsorpti on at 258, 266, 274 and 300 K on the PdO(101) thin film ................................ ...... 122 4 6 Initial dissociation probability of propane on PdO(101) as a function of the surface temperature from 250 to 300 K ................................ ............................ 123 4 7 Arrhenius construction derived from the precursor mediated kinetic model describing the measured dependence of So on surface temperature .............. 12 4
9 Abstract of Dissertatio n Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GROWTH PROPERTIES AND REACTIVITY OF OXYGEN PHASES ON PLATINUM (111) AND PALLADIUM (111) By Su nil Poondi Devarajan May 2012 Chair: Jason F. Weaver Major: Chemical Engineering Oxidation reactions of Pt and Pd under lean burn or oxygen rich conditions are crucial to heterogeneous catalysis systems used in oxidation of hydrocarbons, fabrication of specialty chemicals, power generation through catalytic oxidation, fuel cells and most significantly pollution control through remediation of industrial and automotive exhaust In spite of their tremendous appeal and widespread use in many important applic ations, knowledge used to formulate catalytic systems based on the transition metals has chiefly been derived from empirical data, because of their low reactivity towards molecular oxygen under experimental conditions of Ultra High Vacuum (UHV). Thanks to recent advances in surface science techniques, path breaking research through innovative experimental methods coupled with a renewed vigor towards computational ab initio simulations, have opened avenues for fundamental understanding of this important clas s of reactions. We utilized strong oxidizing agents like nitrogen di oxide and atomic oxygen beams to grow oxygen phases on platinum and palladium single crystals and studied their characteristics using various surface analytic techniques. Our STM work on
10 Pt(111), ends a long standing debate on whether the oxygen atoms continue filling up fcc hollow sites or start filling up hcp hollow sites beyond the well understood 0.25 ML coverage. We also present evidence to demonstrate formation of a Pt oxide chain co mpound which appears as protrusions on the surface and arrange themselves into a well networked superstructure during initial oxidation Our work on Pd(111) using TPRS, reveals for the first time that C H bo nd cleavage of propane occurs on a PdO(101) thin film at temperatures below 200 K under UHV conditions It is also observed that the hydrogen, and propyl fragments resulting from the bond cleavage react with the thin film oxide to undergo complete oxidation releasing H 2 O and CO 2 at higher temperatures. T he C H bond cleavage occurs only because of the formation of a strongly bound molecular state, which in turn is facilitated by the unique local bonding environment of the PdO(101) surface.
11 CHAPTER 1 INTRODUCTORY REMARKS Background and Motivation Fundamen tal knowledge re garding reactions of transition metals (TM) under oxidizing conditions is essential for understanding the working and scientific development of constituent heterogeneous catalysis systems, which are the driving force behind many important i ndustrially relevant applications These include remediation of automotive and industrial exhaust, cat alytic combustion of methane selective oxidation of organic compounds, and oxidation of hydrogen and methanol in fuel cells. Oxidation based TM catalysts have predominantly been developed using empirical methods, which involve extensive testing of TM catalysts under reaction conditions. Diversification in choice of catalytic materials to include combinations of one or more metals and catalyst supports in v arious physical forms and varying operating conditions, translates to endles s permutations and combinations. This has made purely empirical methods extremely cumbersome in terms of time and money. A systematic understanding of the formation and reactivity of various oxidic phases on TM surfaces under oxidizing or lean burn conditions can be a powerful tool for the design and fabrication of more functional and environmentally less impacting catalysts. Experimental s urface science studies are conducted in env ironmentally controlled conditions in ultra high v acuum (UHV), on model single crystal catalytic surfaces Knowledge gained from such studies together with model based ab initio computer simulations can be extrapolated to gain insight into the functioning of the catalysts in real life catalytic environments. Traditionally, experimental surface science studies of TM oxidation has been limit ed to very low oxygen coverages d ue to the
12 limited reactivity of molecular oxygen towards these surfaces under experimen tal co nditions of UHV In contrast, during normal operation of TM based oxidation cataly tic systems the pressure usually equals or exceeds atmospheric pressure, resulting in the formation of various high coverage oxygen phases including intermediates and bulk oxides This raises the issue of the feasibility of performing surface science studies in UHV from a practical standpoint. p cienti fic groups tend to work around this problem by oxidizing the la te transition metal surfaces in UHV through the use of i nnovative methods. For example, high coverage oxygen phases on these surfaces can be achieved in UHV by using stro nger oxidizing agents like NO 2 O 3 and atomic oxygen beams in a controlled and clean manner. The ability to reproduce these high oxygen concentration phases on temperature controlled metallic substrates is critical for carrying out various surface ana lytic/reaction studies. This provides the opportunity to perform exciting and innovative f undamental studies on the oxidation catalysis of late transition metals. The TM or platinum group elements are known for their unique properties and hence find use in widespread heterogeneous catalytic applications, under oxidizing conditions [1, 2] In particul ar, oxidation based catalysis using Pt and Pd is extremely important as illustrated by their ubiquitous use in many important applications. Automotive exhaust remediation where harmful gases like CO, NO x and hydrocarbons are converted to CO 2 N 2 and H 2 O, and fuel cells where H 2 is converted to H 2 O, heat and electricity are some of their very important applications. Formation of various oxidic phases ranging from low concentration chemisorbed phases to bulk oxides, affects the chemical and electronic properties of the substrates, and in turn, results in the reactivity
13 towards reactant species being very different from that of the origina l substrate Hence it is extremely important to try and understand the properties and reactivity of these catalysts under oxidizing environment. The focus of my doctoral study is to investigate the development of oxygen phases on Pt(111) and Pd(111 ) single crystal model catalytic surfaces and to explore their reactive properties. The following sections will focus on specific examples to highlight the importance of such studies on platinum and palladium surfaces respectively. Platinum Study The oscil lations in the rate of CO oxidation on Pt surfaces has been attributed to oxidation reduction cycles of the surface which results in transition of oxygen between two surface states [3, 4] D ramati c increases in CO oxidation activity by several late TM surfaces has been reported, in a narrow range of oxygen coverage near one Mono L ayer (ML), just prior to the onse t of bulk oxidation  A monolayer ( ML ) is defined as equal to the substrate metal surface atom density which is 1.51 10 15 cm 2 for Pt(111) [6 9] In a computational study, it has been report ed that NO oxidation on Pt(111) becomes favorable only at sufficientl y high oxygen coverages due to weakening in the O Pt(111) bond strength with increasing coverage  NO oxidation to NO 2 over Pt based catalysts is a key step in the NO x storage/reduction (NSR) technology pioneered by Toyota for mobile applications  It is proposed that even the traditional continuous selective reduction by hydrocarbons (HC SCR) method could be enhanced by first converting NO x to NO 2 [12 15] It has been evidenced that NO oxidation activity is highest for oxide supported Pt catalysts that expose a large fraction of (111) facets and that the activity diminishes under conditions favoring bulk Pt oxide formation  This again points to a narrow range of oxygen coverage for optimal catalytic activity. Thus,
14 there is a strong motivation to investigate structural and reactive properties of intermediate oxides that form on Pt(111) during oxidation and the mechani sms for oxidation Pt(111) oxidation has been extensively investigated in UHV, using molecular oxygen and it is well established that oxygen atoms bind at fcc hollow sites on Pt(111) up to a coverage of 0.25 ML [17 21] The ordered low density p(22) structure of the 3 state) and the low saturation coverage i s attributed to repulsive interactions between the adsorbed oxygen atoms and kinetic limitations in O 2 bond activation in the presence of chemisorbed atomic oxygen. Higher atomic oxygen coverages have been achieved using stronger oxidants such as NO 2 [18, 22] O 3  and atomic oxygen [19, 24] and therein, temperature programmed desorption (TPD) experiments suggest that oxygen atoms sequentially populate three distinct states 3 2 1 ) as the oxygen coverage increases up to 0.75 ML D ramatic changes in O 2 TPD spectra for atomic oxygen coverages above 1 ML indicate formatio n of three dimensional (3D) PtO 2 domains  Although much has been learnt about the properties of the higher concentration phases through spectroscopic measurements [1 8, 19] LEED studies [18, 19, 23] DFT calculations [25, 26] and TPD of adlayers consisting of isotopic mixtures of 18 O and 16 O atoms  it is still not clear whether the oxygen atoms continue to occupy fcc hollow sites or possibly, start occupying hcp hollow sites 2 phase. Even less 1 phase which predominantly de velops at coverages above 0.50 ML and serves as a precursor to bulk oxide formation.
15 Incorpora tion of a variable temperature scanning tunneling mi croscope (STM), capable of producing in situ, atomically resolved images of adsorbate s on TM surfaces, allows us to further study the structural aspects of the phases that develop during the oxidation process. Details regarding the STM system and addi tion of a plasma source chamber are provided in a separate chapter on UHV STM In our stud y of Pt oxidation using in situ STM, we literally achieved a clearer picture on Pt(111) oxidation which helps us to put an end to long standing debates on the development of high coverage phases of oxygen on Pt(111). Our observations indicate beyond doubt that the oxygen atoms continue to occupy fcc hollow sites and arrange into three rotationally degenerate p(21) domains above a coverage of 0.25 ML corresponding 2 state. A Pt oxide chain compound, which appears as bright protrusions and chains of about 1.7 A in height, starts to develop at a coverage of about 0.40 ML, coinciding with the 1 phase. The Pt oxide chain compound which forms as oxygen atoms occupy sites between the close packed rows of the p(21) structure, has fourfold Pt O coordination, and a local oxygen concentration of 1 ML. With increasing total coverage, the Pt oxide chains develop into an interconnected network of Y s haped branches with regions locally resembling a honeycomb superstructure, as the coverage approaches 0.75 ML. This work demonstrates that formation of a Pt oxide chain compound occurs during the early stages of Pt(111) oxidation and that long range effect s dictate the structural characteristics of the chain network. Polycrystalline Pt exposes large fractions of (111) facets as well as low coordination sites. Pt oxide chains have already been shown to form on the low coordination sites of Pt(110)  and Pt(332)  Since our
16 work has established that Pt oxide chains also form on the close packed (111) surface, it suggests that this compound plays a ge neral role in the catalytic activity of Pt surfaces in oxidizing conditions and that studies on the reactive properties of the Pt oxide chains could provide key insights into understanding the behavior of Pt catalyzed oxidation reactions. The platinum oxid ation work has been well received and has been cited by many groups studying the oxidation reactions of Pt group metals in the recent past. Most notably, the Anders Nilsson group at the Stanford University, SLAC National Accelerator Laboratory, has provide d evidence that the electrochemical oxidation of Pt(111) could proceed via a PtO like surface oxide  by observing fingerprints of surface platinum ox ide using in situ x ray probing. In another study on oxidation of Pt(111) under near ambient conditions using XPS  they have shown that oxide growth commences via a PtO like surface oxide that coexists with chemisorbed oxygen PtO 2 trilayer is identified as the result of bulk oxidation. These findings are in complete agreement with my UHV study of platinum oxidation. Thus, platinum oxidation is found to proceed by the similar proc ess in three very different scenarios and it is really exciting to observe that these results support the findings of my study on platinum oxidation in UHV and validates the significance of the findings in understanding Pt oxidation in general. Palladium Study The oxidation of Pd surface s to bulk PdO is a key aspect in the func tioning of Pd based catalysts under oxidizing conditions. The exceptional activity o f supported Pd catalysts in the lean burn oxidation of CH 4 [32 43] and CO [44 51] is attributed to the formation of PdO. While the PdO(100) facet of bulk PdO may be slightly mo re stable than the (101) facet  the oxidation of Pd(111) preferentially produces PdO(101 ), as
17 discovered by Kan and Weaver  Consequently PdO(101) should be prevalent on the surfaces of supported Pd oxidation catalysts under practical reaction conditions. The study of the reactivity of the PdO(101) surface could thus prove to be vital in understanding the general reactivity of supported Pd catalysts under oxidizing conditions. There is tremendous potential to conduct me aningful studies on the PdO(101) surface, one such study being the breakthrough work on propane oxidation included as part of this thesis. It is the first instance of alkane activa tion reported on a crystalline oxide surface under U HV conditions It has pa ved the way for an important class of reactions in heterogeneous catalysis Alkanes are available in plenty and being saturated hydrocarbons, are considered chemically inert, hence they are also known as ctivation of the C H bond of alkanes o n TM oxides could be the key for developing clean and efficient catalytic combustion systems and also synthesis of important organic compounds, through partial oxidation of the saturated hydrocarbons. This also opens up the alluring prospect of chain lengt hening of small hydrocarbons, starting right from methane. The adsorption of alkanes on clean TM surfaces has previously been extensively investigated  Dissociative chemisorption can occur either by a direct or a trapping mediated mechanism. When strongly activated by energizing the incident alkane molecules, the direct mechanism dominates and the alkane dissociates in a single gas surface collision. Whereas, under typical reaction conditions the vast majority of molecules have low kinetic energies and tend to molecularly adsorb instead. For a trapped molecule, a kinetic competition exists between dissociation and desorption, and since the molecule dissipates its energy to the surface du ring trapping, the existence of
18 low energy pa thway(s) is a prerequisite for C H bond cleavage. Facile trapping mediated dissociation has been reported only on a few metal surfaces, including the reconstructed Ir(110) [55 57] and Pt(110)  surfaces and defect sites of Pt(111)  The fact that no su ch dissociation has been observed on the close packed surfaces of the clean metals, suggests that low coordination sites on metallic Pt and Ir surfaces are more effective towards alkane activation. Very little has been reported previously on the adsorption and activation of alkanes on TM oxides. In the investigation of the adsorption and reaction of propane on a PdO(101) thin film using temperature p rogrammed reaction s pectroscopy (TPRS), i t is observed that propane undergoes facile C H bond cleavage below 200K in UHV, by a precursor mediated mechanism A strongly bound molecular state serves as the precursor to dissociation  Subsequent studies with the additional use of DFT calculations, have revealed that methane, ethane, propane and n butane bind on PdO(101) by forming dative bonds with coordinatively unsaturated (cus) Pd atoms [61, 62] The resulting strongly bound molecularly adsorbed species are analogous to alkane sigma complexes that are formed as intermediates in the general mechanism for C H bond activation of alkanes by mononuclear TM complexes. The f ormation of the strongly bound sigma complexes through dative bonding with the cus Pd atoms on the surface can account for the exceptional activity of PdO surfaces towards alkane activation. The interesting trend emerging out of these studies of alkane act ivation is that dissociation barriers of the molecularly bound species are strongly dependent on the l ocal coordinative environment of the metallic as well as oxidic surfaces. This is an important finding in understanding the catalytic activity of Pd under
19 oxidizing conditions. The next section will touch upon the experimental techniques used to perform the surface science studies included in this dissertation. Experimental Techniques Used This section gives a brief description of the experimental methods u sed during the course of my doctoral study. The experiments are conducted in two separate UHV chambers equipped with unique surface analytical equipment. Individual studies are conducted within the same UHV chamber, utilizing the in situ techniques availab le in that particular chamber. The STM study of Pt(111) oxidation was conducted in the chamber, det ails regarding which are provided in the next chapter Thus, atomic oxygen beams generated in situ by splitting molecular O 2 using a microwave plasma source in the STM chamber and an RF plasma source in the AO chamber, are used to achieve high concentratio n oxygen phases on the substrates. Surface analytical techniques employed in my studies include TPD/TPRS, AES, XPS, LEED and STM. Detailed description of the AO chamber is available [63, 64] STM theory and experimentation is provided in the next chapter on UHV STM Sputtering using Ar + ions, and exposure to O 2 NO 2 and atomic oxygen beams at elevated temperatures followed by thermal annealing were used for sample cleaning in UHV. The sample cleanliness and composition was determined by mainly using Auger electron spectroscopy (AES) in the STM chamber and x ray photoelectron spectroscopy (XPS) in the AO chamber. In AES, an electron beam (~ 2keV) is used to ionize one or more core levels in the surface atoms. Such an ionized atom could return to its electronic g round state by the core hole being replaced by an outer electron. Through a
20 radiationless Auger transition the available energy it transmitted to a second electron which escapes as an Auger electron with a characteristic kinetic energy. Analysis of the Auger energies provides elemental identification since no two elements have the same set of atomic binding energies. Information on AES theory and da ta for identification and interpretation is available [65, 66] In XPS, the surface to be analyzed is irradiated with soft x ray radiation with photons of known energy, which results in the liberation of electrons from their bound state with certain kinetic energy. The difference between the incident photon energy and the measured kinetic energy of the ejected photoelectrons gives the binding energy of the electron. Elemental analysis is possible using the detected binding energies since no two elements share the same se t of electronic binding energies [66, 67] Since any change in the binding energies is reflected in the photoelectron energies, XPS is also useful in detecting changes in the chemical environment of an atom. This is particularly useful in determining whether the metal surface is oxidized. Surface ordering of clean substrate and adlayers can be studied by low energy electron diffraction (LEED). It involves the bombardment of a surface with a low energy electron beam. Back diffracted electrons travel towards the LEED screen on radial trajectories. The electrons pass through a retarding field energy analyzer to enable a narrow range of elastically scattered electrons to be transmitted to the phosphorescent screen. Visual examination of the diffraction pattern and simple mathematical transformations of the spot profile reveals the ordering of the surface. Deta ils regarding theory and analysis of LEED spots is available 
21 Temperature programmed desorption (TPD), also referred to as thermal desorption spectroscopy (TDS) is used to study reactions between pure gases and clean solid surfaces. Typically, a gas is adsorbed on a model catalyst surface at low temperature in UHV after which the residual gas is pumped out. This is followed by heating the surface at a constant rate, while a quadrupole mass spectrometer (QMS) measures the current of an ion or ion group as a function of the surface temperature. The peaks in the ion current correspond to the rate of desorption of a particular species from the surface, as and when their binding energy thresholds are crossed. T he integrated area under a peak, normaliz ed with that of a known coverage, relates to the coverage of a particular species on the surface. In addition, the characteristic peak temperature, the shape of the peak and the way the peak temperature and shape of the peak changes with initial coverage r eveal information regarding the activation energy for desorption, frequency factors and reaction order for the desorption processes [68 70] When multiple reactant species ar e a dsorbed onto a clean surface or, one or more species are adsorbe d to a surface oxide, the reaction products are monitored during the temperature ramp. This method is known as temperature programmed reaction spectroscopy (TPRS). STM is a surface probe micro scopy technique that accesses electronic properties of conducting surfaces through the quantum mechanical tunneling effect. An atomically sharp metal tip is rastered across a sample surface with the tip sample separation being small enough to allow tunneli ng of electrons across the vacuum barrier upon application of a small bias voltage. The measured tunneling current is extremely sensitive to the sample tip separation and can in turn be controlled using a feedback loop, which
22 adjusts the tunneling gap by a pplication of suitable voltages on the piezoelectrics of the scan head of the microscope. This possibility of positioning the atomicall y sharp probe with controllable precision above a surface, within atomic separation distances, allows for near field imag ing Thus topological information can be obtained down to the atomic scale. Real space imaging of local phenomena at the atomic scale through STM can provide information regarding non periodic physical and electronic properties of surfaces. This is a key a dvantage of the STM, over the conventional surface science techniques which chiefly provide data on the substrate and reactants, av eraged over the entire surface. The following is a chapter on UHV STM which briefly describes the setting up of the apparatus STM theory, design considerations, key aspects of the system and concludes with a section on special operating procedures and maintenance. The next chapter is an STM study on Pt(111) oxidation and is followed by a propane oxidation study on PdO(101). Thi s is followed by conclusions, list of references and a bibliographic sketch on the author of this dissertation.
23 CHAPTER 2 SCANNING TUNNELING M ICROSCOPY IN UHV UHV STM System Development The first part of my research involved the setting up a UHV system capable of performing in situ Scanning Tunneling Microscopy along with other surface science capabilities. A home built UHV system called the AO chamber mentioned in the previous chapter, was already available for performing surface science experiments on model, catalytic, single crystal surfaces. The system has the capability to perform AES, XPS, TPD/TPRS, LEED, Electron Energy Loss Spectroscopy (EELS) and Low Energy Ion Scattering Spectroscopy (LEISS). My interest in the growth, properties and reactivity of interest in oxidation catalysis of late transition metals. Studies were being carried out using atomic oxygen beams to produce high coverage structures on Pt and Pd surfaces. In particular a study on the oxidation of Pt(111) with ato mic oxygen beams had just been completed and a precursor to oxide formation at intermediate oxygen concentrations and bulk like oxide growth at higher concentrations were observed using thermal des orption studies. Although these were very interesting findings, we felt the need to include atomic level investigations to our surface science capabilities in order to be able to further investigate the structure and properties of these new phases at the a tomic scale. A home built UHV system with in situ STM and other surface science capabilities was purchased from the Robert Madix group at Stanford University I spent the next few years assembling, modifying and testing the system with the aim of achieving atomic resolution of oxygen phases grown on platinum and palladium single crystal surfaces.
24 We installed a side chamber containing a microwave plasma source to achieve atomic oxygen beam capabilities in the STM chamber b rief details regarding which are p rovided in the experimental section of the next chapter on platinum oxidation Since our group was new to UHV STM, upon the invitation of Dr. Cynthia Friend and Dr. Robert Madix, I went to their surface science laboratory at Harvard University as a visitin g researcher for a few months and learnt the fundamentals of operating UHV STM systems and was also exposed to the troubleshooting and maintenance procedures T he knowledge gained allowed me to tackle problems specific to UHV STM systems ranging from minim um requirements of the individual components to operating procedures of the system as a whole. Considerable time and effort was required to properly set up the UHV STM system and to develop operating procedures. I undertook a complete overhaul of the syste m and systematically weeded out all possible sources of trouble and at the same time, identified and eliminated specific problems using troubleshooting techniques The last section of this chapter deals with the common proble ms I had to deal with, and the steps taken to overcome them Also mentioned are key aspects of the s pecial operating procedures developed for the system over the years that should be useful in operating the system. The following is a brief presentation of STM theory, followed by a descr iption of the STM imaging in UHV. This will be followed by a section which highlights the design and implementation aspects of a UHV STM system. The next section will briefly outline the key aspects of the system development. The concluding section deals w ith the common problems associated with UHV STM and special operating procedures.
25 Basis for Scanning Tunneling Microscopy In scanning tunneling microscopy, the three dimensional variation of charge density at a surface is probed via electron tunneling betw een a sharp tip and the sample. The potential in the vacuum region acts as a barrier to electron transfer between the two electrodes. Tunneling occurs as a result of the wavelike properties of particles in quantum mechanics. Unlike in classical mechanics, there is a finite probability that a particle, with kinetic energy lower than the potential energy of a barrier, could traverse the forbidden region and reappear on the other side. A simple schematic representing the phenomenon of quantum tunneling effect is presented in Figure 2 1. When the tunneling gap is small and the bias voltage is low, the relation of the tunneling current to the gap distance can be simplified to, (2 1) where, A = 1.025 (eV) 1/2 1 (a constant), is the average barrier height, V is the bias voltage and s is the separation between the two electrodes  This equation indicates that the tunneling current changes by an order of magnitude for every 1 change in tunneling gap, with the average barrier height taken as the work function, ~ 4 5 eV for most metals of interest Young et al.  observed this exponential dependence in 1971, by br inging a field emitter close to a metal surface and monitoring the current voltage characteristic. This work was reported as the first instance of metal vacuum  and Tersoff and Hamann were the first to apply it to STM  Detailed information on tunneling theory and STM theory is available in these references [71, 75 77]
26 Atomic resolution with STM is a direct result of the extreme sensitivity of the tunneling cu rrent to the sample tip separation. If a tip is atomically sharp, the tunneling current from the apex atom of the tip is an order of magnitude larger than that from a second layer atom. Since most of the tunneling occurs through one atom on the tip, atomic scale resolution in the lateral directions is possible. Binning and Rohrer invented scanning tunneling microscopy in 1981 by using a piezoelectric driver to accurately control the tip sample separation via a feedback loop. The ability to bring an atomical ly sharp tip within tunneling range of the sample surface in a controlled manner, and ability to use measured data to reposition the tip allowed them to obtain atomically resolved images of solid surfaces [78 83] Their method and initial results were validate d when select surface science groups were able to reproduce their images of the Si(111) 7 7 reconstruction [84 86] Ability to be used in surface science studies is a key factor for the success of the tech nique which earned its inventors the Nobel Prize in 1986. The unique properties of surfaces result from the reduced coordination of surface atoms and break in periodicity in the surface normal direction making them the center of a number of interesting ph enomena that neither have analogy to the bulk nor to the vacuum outside the surface. Thus, surface science studies impact the scientific development of diverse areas such as, catalysis, corrosion, tribology, adhesion and microelectronics by providing infor mation about both surface composition and surface chemistry. Real space imaging of local phenomena at the atomic scale through STM can provide information regarding non periodic physical and electronic properties of surfaces. This is a key advantage of STM over conventional UHV surface science techniques which involve excitation of the
27 surface using particles, photons, heat or high electrical field, followed by the detection of secondary emitted particles and photons from the surface [66, 87] Th ese methods usually probe depths of about 5 20 and provide information averaged over the surface whereas, by default, STM probes the outermost surface layer and the size and location of the scan area can be arbitrarily specified. As demonstrated by Besse nbacher  many dis puted surface structural models previously studied using various experimental techniques and theoretical calculations have been resolved by the direct imaging capability of STM. The following section on STM imaging in UHV provides a basic description of ST M imaging methods, and highlights the essential requirements of an UHV STM system. Scanning Tunneling Microscopy: Imaging in Ultra High Vacuum Tip Sample Approach/Retract Surface science experiments on conducting surfaces using STM are carried out in UHV chambers hosting many other in situ sample preparation and analysis techniques, hence requiring sample transfer. For this the sample needs to be decoupled from the microscope and once again brought into tunneling range for imaging. This is done using an a pproach mechanism which essentially needs to be able to move the sample far enough away from the microscope or vice versa, and then be able to again bring them close enough to within the scanning range. This involves a combination of manual positioning and computer controlled stepwise motion of the tip towards or away from the sample while constantly monitoring the tunneling current. Several factors like geometry, rigidity and mobility requirements of the device, coupled with reliability and speed need to b e factored in while selecting the type of approach mechanism. The devices for sample approach are divided into three categories, namely screw, clamp step and stick
28 slip types, and the selection is based upon the unique requirements of the particular micros copy design  Once in tunneling range, the tip position, in the lateral and vertical directions, is co ntrolled by applying suitable voltages to piezoelectrics on the scan he ad of the microscope. Figure 2 2 is a schematic representation of the principle of the scanning tunneling microscope. The tu nneling current which is depende nt on the tip sample separati on, is controlled using a feedback loop to adjust the vertical position of the tip. A preset tunneling current of about a nanoampere is achieved for small bias voltages of about a few millivolts to a volt, between the sample and the tip. The tip is rastere d across a preset scan area on the surface by applying suitable voltages to the microscope head while monitoring the tunneling current and controlling it when required. Imaging Modes When the scanning is done slow ly enough across a surface at a fixed bias voltage, tunneling current can be kept constant by readjusting the height of the tip, thus following the topography of the surface. A plot of the vertical position of the tunneling tip at each point of the scan is recorded in the form of a topography image This is referred to as constant current or topography mode and is particularly useful for scanning areas with step edges or surfaces with unknown topography. A schematic representation for constant current mode is giv en in Figure 2 3 Scan dimensions for the constant current mode r ange from tens to thousands of ngstroms. For small bias voltages, images taken in this mode represent contour maps of the local density of states (LDOS) near the Fermi level of the metal If the polarity of the sample is negati ve, then the states in the valence band are imaged. For a positive polarity of the sample the distribution of electronic states in the conduction band can be recorded. In case different chemical
29 species are present on the surface, the image contrast is fur ther influenced by the varying effective barrier height (work function) at different positions  Alternatively, once the tip is close enough to the sample, the feedback loop is switched off and the tunneling curr ent is measured while rastering the tip at a constant height over very small areas over which the surface is expected to be relatively flat. This method is called the constant height mode or simply, current mode, and small features are reflected in terms o f fluctuations in tunneling current, rather than in the tip height  A schematic representation of the constant he ight mode is shown in Figure 2 4 Typical scan dimensions for constant height mode are ten to hundred ngstroms and care should be taken to avoid tip sample contact which will result in tip damage. Other modes of scanning include the study of d ifferences in local barrier height by recording dI/ds via modulations of tip sample separation, and spectroscopy studies by obtaining dI/dU, via modulation of the bias voltage for a fixed tip separation [71, 75, 77] Challenges of UHV STM System D esign Atomic resolution imaging involves obtain ing a horizontal resolution of tenths of angstroms and a vertical resolution of hundredths of angstroms. The exacting tolerances for instruments capable of obtaining atomic resolution, place a number of constraints on the design to keep it free of vibratio nal, electrical and thermal noise. Vibrational noise is caused by many sources ranging from low frequency building noise to high frequency machine and acoustic noise. Fans used to cool electronic equipment, fluorescent lighting and other room noise includ ing people contribute to the vibrational noise. Electrical noise sources include the power supply, signal transmission and noise pick up by the tunneling wire. The feedback loop which consists of a PID controller can also introduce instabilities and has t o be carefully monitored during scanning. Studying
30 surface reactions involve imaging chemical intermediates which require the sample temperature to be controllable to low temperatures. This can be achieved by thermal contact of the sample with a liquid nit rogen reservoir and simultaneous heating to maintain specific temperatures. Small temperature fluctuations can result in thermal drift which can cause the STM to shift away from its previous location. Boiling liquid nitrogen in the reservoir though, sends vibration to the sample through the thermal contact. Therefore, care needs to be taken to provide excellent thermal contact with the liquid nitrogen reservoir while at the same time provide vibration isolation from it. Another important aspect of STM imagi ng is the importance of sample cleanliness due to the extreme sensitivity of the microscope. Special procedures including chemical treatment and high temperature annealing need to be followed for sample preparation and surfaces to be imaged using STM need to be periodically checked using surface analytic techniques such as AES, LEED and TPD to get an idea of the cleanliness and structure of the surface. Due to the complications involved in imaging due to tip condition, noise and extremely localized area ima ging by the microscope it is not prudent to depend only on the STM, but rather wise, to use it as an extension to the other tradi tional surface science methods. The next section, on experimental design and implementation, deals with theoretical considerat ions for vibration isolation and a description of the UHV STM system used during my experiments, enumerating the salient features of the system that allow for the extreme demands of a UHV STM system.
31 Design and Implementation Aspects of UHV STM System Vibr ation Isolation Vibration isolation is perhaps the most important aspect of an UHV STM system design. Since the microscope is used to probe atomic features on the surface, the vibration isolation needs to damp out the response of the system to external noi se, to the same level. For floor vibrations of a few thousand angstroms, the vibration isolation factor should be at least 10 5 in order to accurately position the tip within a few hundredths of angstroms from the surface. Fourier transform analysis of th e tunneling current recorded by the control unit yields information about the frequency components and the intensity of each frequency. An oscilloscope and a dynamic signal analyzer (Stanford Research Systems: SR785) were used to study the noise spectrum. This information is primarily used to eliminate noise sources corresponding to the particular frequencies until all the noise is removed and high quality images are produced periodically by the microscope. The vibration isolation primarily consists of maki ng the system mechanically rigid and decoupling it from the surroundings by using vibration isolation stages. Stability of the tunneling junction is analyzed by reducing the system to a few rigid objects with characteristic parameters and considering the m otion in the vertical direction. The tunneling device including the scanner, tip, sample holder and the sample, should be rigid and have a high resonant frequency and the vibration isolation stage should have a low resonant frequency so that the best (lowe st) response of the tunneling gap to external vibration is achieved in the intermediate frequencies. Since the UHV STM system (especially home built apparatus) needs constant repair and modifications to improve performance, it is essential to conceptually understand vibration isolation. As
32 will be shown in the implementation section, basic design and modifications made to the s ystem should always take into account the extremely sensitive nature of the apparatus and also the fact that materials used must be UHV compatible. The following is a theoretical analysis of the tunneling unit by itself, followed by the incorporation of a single vibration isolation stage, closely following the model and results of the theoretical considerations of STM design by Sang il Park and Robert Barrett  The tunneling unit consisting of the sample, sample holder, microscope hea d and tunneling tip can be modeled as a forced, damped, simple harmonic os cillator, as shown in Figure 2 5 The sample is the driving oscillator, the tip is the mass at the end of the spring, and the spring is the microscope body and sample holder. The equ ation of motion is written in the form: (2 2) w here m is the effective mass, is its (tip) vertical position b is the damping factor, k is the spring constant of the unit, and is the vertical displacement of the oscillator (sample) Using the definitions of overall resonance frequency and quality factor (Q), (2 3) equation (2 2) can be rewritten as, (2 4) W hen the STM frame is driven sinusoidally, i.e., the steady state harmonic solution can be used to obtain the transfer function for the
33 relationship between variation in tunneling gap distanc e and external vibration amplitude as, (2 5) w hich is plotted in Figure 2 6 for It is interesting to note three distinct regions in the plot. At higher frequen cies, where the transfer function is unity, the tunneling gap changes according to the vibration of the tunneling unit. When the unit is driven at the resonance frequency, it is the damping which prevents the oscillations from approaching infinity. At the resonant frequency, the asymptotic growth of the oscillations is limited by the damping. Note that when there is no damping the transfer function approaches infinity for vibrations at the resonant frequency In the lower frequency range, the transfer function drops exponentially with decrease in the driving frequency as When the resonant frequency is 1 2 kHz, the transfer function is less than 10 6 for driving frequencies below 1 2 Hz, whic h is sufficiently small for STM, provided floor vibrations are of the order of a few thousand angstroms. An isolation system is required for vibration frequencies above 10 Hz. A model for single stage vibration isolation is sho wn in Figure 2 7 Here the ma ss represents the entir e tunneling device of Figure 2 5 and is assumed to be rigid. The equation of motion is given by, (2 6 )
34 w here, is the external vibration. By defining parameters and the equation can be rewritten and the amplitude ratio of external vibration to that of the tunneling stage can be obtained as, (2 7) This is the transfer function between the externa l vibration and the isolation st age and is plotted in Figure 2 8 for three different values of damping. It is noted that for small damping the vibration isolation is more efficient at higher frequencies, but the oscillation at the r esonant frequency is large. Heavy damping reduces the amplitude of resonance oscillation, but does not attenuate the high frequency vibrations as effectively. The response of the tunneling gap to external vibrations, including both the tunneling unit and t he isolation stage is obtained by the magnitude of the product of T 1 and T 2 Figure 2 9 is a plot of the overall transfer function for again for the three different values of damping. It can be seen that the best isolation is for sma ll damping, in the intermediate frequency range where the overall transfer function becomes thus solely depend e nt on the ratio of the resonant frequencies of the tunneling stage and the isolation stage Vibrati on isolation systems are of two types namely, coiled spring suspension with magnetic damping and stack of metal plates with viton dampers between each pair of steel plates [77, 88] In most of the systems additional vibration isolation is provided through pneumatic
35 suspe nsion using laminar flow isolation legs, which provides very low resonant vibration frequencies (~1 Hz). To summarize, UHV STM systems are designed such that the entire system must have a low resonant frequency to prevent most vibrations from reaching the microscope, while the tunneling unit comprising the tip sample system should have a very high resonant frequency so that the low frequency vibrations that do reach it will not excite the resonant frequency of the microscope. Description and Salient Feature s of UHV STM Apparatus A very detailed description of the theory, design, construction and characteristics of the UHV STM apparatus used for my STM study is given by William Crew [89, 90] who built the original system during his doctoral study at Stanford University. Although, the system design is based on a design concept by Thomas Michely  it is itself unique because it allows imaging with atomic resolution while cooling continuously wi th liquid nitrogen and incorporates laminar flow isolation legs for vibrational isolation  In this section I will provide a basic description of the appa ratus and highlight the key features incorporated into the design, pertaining to the STM. Information and figures  Figure 2 10 is a schematic representing the UHV STM chamber. It is divided into a top and a bottom section. The top section contains all of the surface science techniques including the STM and the bottom section contains a large pump well which houses the ion and titanium sublimation pumps. The larg e pump well beneath the top half of the chamber lowers the center of mass to below the platform level of the chamber, thereby making it bottom heavy (like a pendulum) and less susceptible to hobby horse type oscillations. The top half is divided into two s tations. The first station is exclusively for STM, while and the second station is for LEED, AES, Ion Sputtering,
36 TPD/TPRS and gas dosing, radially distributed in a semicircle. To position the sample for various techniques, the sample manipulator moves alo ng the length of the chamber and rotates the sample up to 180 degrees. The sample manipulator is mechanically linked to a rotary, differentially pumped flange mounted on a motorized bellows. The sample is linearly translated by the motorized bellows and ro tated by turning the differentially pumped flange. length of the main chamber in Fig 2 9. The weight of the tube and thickness of its walls ensure a very low resonance frequency, which ensures that any noise conducted to it, is not amplified by coupling with its resonant frequency. The tube itself rests on 4 teflon buttons at either end of the main part of the chamber, which form a stable resting cradle for it. Teflon buttons minimize the number of moving parts and minimize vibration coupling between the chamber and the sample. A loose mechanical linkage between the rotary flange and the ma nipulator isolates the translation mechanism from the sample manipulator rod. Liquid nitrogen tubes that run down the center of the manipulator are isolated from the original feedthroughs by formed bellows. Electrical leads for sample heating, thermocouple bias etc. are secured directly to the sample manipulator rod. Separate leads are drawn inside the manipulator rod to reach the sample. Any vibration conducted through the feedthrough wires must excite the vibrational modes of the tube to continue to the sample. A cutoff section in the middle of the sample manipulator rod houses the sample, first vibration isolation stage, sample heater and cooling reservoir. The sample to be
37 imaged is held in place by a sample holder which consists of an STM ramp, a tanta lum spring, a sapphire insulator and a bottom copp er well. As shown in Figure 2 11 the sample is pressed up against the STM ramp by the use of a home made tantalum spring which is sandwiched between the sample and the copper well with a sapphire spacer in between for electrical isolation. The bottom copper well allows a sample heater to be inserted from below, such that no contact is made with sample holder assembly and the filament reaches as close as possible to the sample for radiative heating. The tant alum spring helps keep the sample holder assembly rigid while imaging. The sample holder assembly is secured from the sides and mounted on a large copper block with thermal insulation between them. Sample cooling is achieved via a flexible copper braid whi ch provides thermal contact between the copper well and a liquid nitrogen reservoir. The copper block is in turn attached to the sample manipulator rod using 4 viton legs, in the cutoff section of the manipulator. The weight of the copper block and the el asticity of the viton comprises the first vibration isolation stage with damping, whose resonance frequency is between 5 and 10 Hz [90, 92] A laminar flow vibration isolation system, manufactured by Newport Corporation, is used a s a second stage of vibration isola tion The vibration isolation system consists of four air legs and has a very low resonant freq uency of ~ 1 Hz (see Figure 2 12 ). This additional vibration isolation stage is essential in achieving atomic resolution while imaging with the UHV STM. The cutoff section of the sample manipulator, also houses a sample heater and a liquid nitrogen reservoir. The sample heater sticks into the sample holder well, from below, thus not making direct physical contact with the sample or the sampl e holder as
38 shown in Figure 2 11 The heater is secured directly to the manipulator rod using a cross plate and heats the sample radiatively. The liquid nitrogen reservoir is secured to the manipulator rod and is connected to the sample holder assembly using a flexible copper braid, to minimize the transmission of vibration arising from the boiling of liquid nitrogen in the reservoir. The schematic in Figure 2 13 represents a cross sectional view of the STM assembly. The scanning tunneling micro type, which was originally designed by Frohn et al.  It consists of a nickel plated aluminum disk and 4 piezoelect ric tubes. Of these, three offset piezos are placed in the center of the disk. As the name suggests, the offset piezos are chiefly involved in the movement of the microscope head for electronic approach (or retract) and also in positioning the scanner tip over specific areas to be scanned on the surface. This is achieved by applying specific voltages on each of the offset piezos, which sit on three different sections of the S TM ramp. Sapphire balls at the ends of the offset piezos, provide thermal isolation between the microscope and the sample assembly and also facilitate the inertial approach mechanism. When the STM head is sitting on the ramp, its rotation is coupled with m ovement up and down the ramp, as though it were walking up and down a spiral staircase. The tip length is selected such that tunneling range is achieved in the middle of the STM ramp. A long tip leads to instability while imaging and a short tip requires p rolonged motion down the STM ramp for approach, which could result in tip crashes. The tunneling tip sticks out at the bottom of the scan piezo and its movement in the lateral and vertical directions is controlled by applying suitable
39 voltages on the scan piezo during the STM imaging. When in tunneling range it is also possible to apply an offset voltage on the offset piezos, so that the entire microscope is moved up or down to achieve desired tunneling current for the set bias voltage in the unextended pos ition of the tunneling piezo. Typically the feedback loop response for topography scanning is best when imaging in this scenario since the full range of motion is possible for the scan tip while imaging. This is a key advantage of this STM system. importantly, thermal drift is reduced and the microscope can scan any section of the surface. When studying reactions or surface structure at temperatures different from that of the micr oscope, there must be very little heat transfer between the microscope and the sample holder. This is achieved by the limited contact of the sapphire balls which also double up as insulators. Because all of the piezos are aligned in the vertical direction, any thermal expansion or contraction that occurs is isolated to the Z direction and there is very little drift in the lateral directions. The drift that shows up in the vertical direction in the images obtained during scanning in the constant current mode can be easily removed during image analysis. A preamplifier converts the nanoampere tunneling current into a low noise voltage signal that can be read by the STM electronics. The electronics for the system including the preamplifier is fabricated by RHK E lectronics Inc. (STM 100), and interfaced to a PC using a 50 pin connector. An MS DOS operating system based software SPM 32 (RHK Technology Inc.) is used for approach/retract and data acquisition during STM imaging. Another software (RHK XPM) is used for data processing on a separate PC. Since it is a very complex electronic system and my chief aim is to understand how to operate the apparatus, I will
40 not focus on the theory and implementation of the electronics system in this dissertation. Some informatio n regarding the repairs I had to make are briefly mentioned in the next section. The concluding section of this chapter deals with the common problems encountered with UHV STM systems and how to deal with them for normal operation, troubleshooting and also modifications, including my own experiences. Common Problems w ith the UHV STM System and Remedies Having given an idea about the scope of the experiment, instrumentation requirements and special features of the actual system incorporated to deal with the challenges of UHV STM, I will now proceed to another important topic which deals with how to operate, repair and maintain a UHV STM system. Operating a UHV STM system is a very difficult task since it has imposing demands during normal operation, which hav e to be taken into account for the design of the system. The need to achieve unrealistic sensitivity of the tunneling junction, while at the same time providing stability, not only imposes novel design strategies, but, also calls for extreme operating proc edures. The delicate nature of the apparatus often leads to maintenance and repair work both inside and outside the UHV chamber. I will start from the microscope itself and proceed outwards, listing key problems and methods to counter them. Tunneling and S can Drive Wires Voltages are applied to the quadrants of each of the three offset piezos and also to ML) that are soldered directly on to the leads on the STM head using a special solder (95% tin and 5% silver). A similar wire is used to carry the tunneling current signal from the tip assembly. These wires break often and have to be soldered back on, or
41 replaced, which involves breaking the vacuum and taking the entire STM o utside the chamber. Thus they have to be tension free at all times to prevent breakage and also to reduce vibration transmission between the microscope and the STM manipulator, during imaging. In order to achieve this and also check for physical integrity, they are pulled along the length using tweezers, straightened out and shaped like a slightly curvy L. Once all the wires are in place and do not tend to vibrate for small disturbances, they also need to be checked for electrical conductivity before puttin g the unit back in vacuum. Sample and Tip Condition The sample and tip condition play a huge role in determining the quality of the STM images obtained. Even if everything else is working properly, a bad tip or dirty sample will result in the image being f illed with random noise. It will be difficult to identify the source of the problem if this is the case. Small voltage spikes applied to the tip while in tunneling range can help to discard some particle that is stuck to the tip. Regular in situ tip condit ioning greatly helps in maintaining and also improving the tip quality. A gold sample placed alongside the sample holder, allows for stripping the tip clean and shaping it using field induced evaporation. A field emission current of 3 4 microamperes is obt ained for bias voltage of a few hundred volts, when the gap distance is just out of tunneling range, and a large resistor (~40 M ohms) is added in series for tip conditioning. When the tunneling tip has had a hard crash on the surface or does not give good images for a period of time, it is replaced with a new tip. New Pt/Ir tips are either purchased (Veeco Probes model: PT ECM) or just made from the alloy wire (Nano Science Instruments: Pt 0.8 Ir 0.2 ) using a pair of scissors to cut a sharp edge from
42 the bott om end of a centimeter long piece. An S shaped bend is put on the top end of the wire and inserted into the hole at the bottom of the scan piezo. Special sample cleaning and high temperature annealing procedures need to be used to prepare the sample for S TM imaging. The procedure must be validated from time to time using AES to check sample cleanliness, and LEED to obtain structural information of the clean metal or adlayer to be imaged. Heating the sample to high temperatures (~1000 K) for cleaning or rea ction studies though, has adverse effects on the performance of the sample holder assembly. Its components are small by necessity, and the assembly has to be very rigid in order to keep the resonance frequency of the microscope high. The heating and coolin g cycles of the sample should not be accompanied by corresponding expansion/contraction of the entire sample holder assembly, which could result in fracture of the sapphire insulator or loosening of the sample holder assembly. Although the cooling braid at tached to the bottom copper well serves to achieve this, the sample holder assembly does become loose over a period of time and has to be regularly maintained and broken sapphire washers need to be replaced. The sapphire washer also gets coated by a tungst en film as a side effect of extremely high temperatures of the filament. This causes a breakdown in electrical isolation between the sample and the cooling line which grounds it. In order to circumvent this problem a shield was added to the sample holder a ssembly to protect the sapphire washer from being exposed to the tungsten vapor. The shield is shown in the schematic for the sampl e holder assembly in Figure 2 11 The tantalum spring used to hold the sample in place within the sample holder assembly has to be inspected and replaced from time to time, to again ensure rigidity of the unit.
43 Electronic Noise and Feedback O scillation The tunneling junction can pick up pick stray electromagnetic fields when the tip is no secured properly or the tunneling juncti on is not properly shielded. The tunneling wire also picks up noise due to vibrations and hence should be short and well supported mechanically. It is provided with ground shielding in the region where it passes through the STM manipulator. The tunneling t ip upon installation is inspected to see if it is secure and just long enough to get into tunneling range at the middle position on the STM ramp. Longer tips tend to be more unstable and shorter tips require the microscope to travel unnecessarily long dist ances down the ramp for tip approach. The preamplifier is a crucial part of the control electronics and needs to be of high quality. It needs to be a low noise type and its location should be close to the tunneling region. It converts the low level tunneli ng current into a high level, low impedence voltage signal that can be transmitted to the STM control electronics. The scan drive signals and offset voltages used have a high voltage range (+/ 130 volts) and are not susceptible to noise pick up. Testing o f the preamplifier performance is done with the tip out of tunneling range, while monitoring for any characteristic noise. Due to the extreme importance of the preamplifier and its sensitivity it is advisable to just replace it with a new one if it is not functioning properly instead of trying to repair it in the laboratory. The old preamplifier which came with the system was replaced with a new RHK IVP 200 preamplifier (Gain: 100 mV/nA). The control electronics itself is pretty robust and nothing usually g oes wrong with it. Rarely, due to voltage spike, a resistor or capacitor on the driver board may burn out and has to be replaced. There are quite a few BNC cables that run from the control electronics to the STM feedthroughs, and these have to be checked f or integrity and proper connections periodically.
44 60 Hz noise is very common in the noise spectrum and can be very hard to remove. This is chiefly because there are quite a few sources for the 60 Hz noise. It comes from improper grounding of the control el ectronics and can also be picked up by the tunneling wire from stray fields, when not properly shielded and also vibrations from fans and other equipment in the circuitry. A sample noise power spectrum show in Figure 2 14 indicates large noise in the measu red signal around 60 Hz. A noise of almost 100 pA is not permissible while measuring 1 nA. To avoid this problem a clean dedicated supply line must be used only for the STM system and all other appliances should be disconnected from this line even if they are switched off. Computers and printer network should definitely not operate on this line. To avoid a situation called, electronics should be run from a two pin connection circu mventing the grounding pin. In this particular STM system, the tunneling wire is already ground shielded with respect to the vacuum chamber, so the only grounding should be done through the chamber itself. Once the ground looping is eliminated, and all ele ctrical connections are checked, if we still have 60 Hz noise from the tunneling wire, the ground shielding has to be replaced. The biggest challenge in eliminating the 60 Hz noise is recognizing the various sources and pinpointing the existing noise to a particular source and taking remedial action. A noise power spectrum such as seen in Figure 2 15 in which the 60 Hz noise is just a few pA is sufficiently low when compared to the measured value of 1000 pA. This indeed is the general way any type of noise is tracked down and eliminated from the system.
45 Special Operating Procedures for UHV STM S ystem The operation procedure includes some extreme steps taken to avoid vibration transmission to the microscope. These include removing some sources of noise within the chamber and also isolating the entire chamber physically from vibration sources. All mechanical pumps and fans on the chamber or near the chamber are turned off. The turbomolecular pumps on the chamber are vented to atmosphere after switching them off All the cables for the various equipment on the chamber and hoses for backing pumps and water lines are removed. The cables connecting the preamplifier and the STM feedthroughs to the STM electronics are secured to the chamber frame so that their weight does not affect the system when it is floating on the pneumatic support. A time is chosen for scanning at which heavy machinery and compressors are not running nearby in the building. The mechanical pumps on other systems in the lab are placed on sponge pa ds to prevent their vibration being transmitted to the floor and in turn to the STM chamber. Control system oscillations should be avoided as they can result in tip crashes and poor imaging. The straightforward case for this type of oscillation is an impro per feedback setting. This is remedied by suitably reducing the control gain of the feedback settings. During regular use, the feedback settings are more or less optimized depending on the i maging parameters. Other factors can contribute to build up of fee dback oscillations. A contaminated tip or sample can contribute to an unstable tip sample junction even at low control gains. If the oscillations do not reflect corresponding decrease with lowering of gain settings, then the problem is most likely with the tip sample junction and better cleaning procedures need to be used for the tip and the sample. In some cases, mechanical vibrations can appear as feedback oscillations.
46 They are identified by applying a light mechanical shock in the form of a light tap t o the microscope and looking for vibrations in the tunneling curren t at the oscillation frequency. Thermal drift and piezoelectric hysteresis can affect the imaging. Non ideal mechanical motions due to varying temperatures across the microscope can result in distortion of the STM images. By using a symmetrical design of the microscope, the thermal drift in the vertical direction is cancelled to a large extent. Post processing can eliminate whatever thermal drift there is in the vertical direction. Thermal d rift in the lateral directions is monitored by successive imaging of an area with a characteristic feature and observing the distance over which the feature has moved during the time interval. Once the sample has reached the desired imaging temperature, th e microscope is allowed to sit on the STM ramp for at least half an hour before imaging, which reduces effects of thermal drift. When known surface structures are imaged, post processing analysis can account for thermal drift, but for unknown structures it is best to image when there is little to no thermal drift. Another factor which can contribute to a poor image with distortion in particular directions is p iezoelectric creep When a control voltage is suddenly applied to a piezoelectric element, it will immediately move approximately 95 % of the full extent of the desired movement. It will slowly move (creep) the remaining 5 % over a period of several minutes. This effect typically shows up in the Z direction while imaging across step edges in large area scans, making them look blurry instead of appearing as a sharp step edge. This effect is not serious in the Z direction for atomic scale imaging. The creep appears as additional drift in the X Y directions and can remain for several minutes after any large motion of the microscope
47 across the sample. This problem is taken care of by waiting for a suitable period of time before imaging after a large change in position across the sample. In spite of all the design considerations and special procedures for oper ation of the UHV STM, it is very difficult to give a standard operating procedure for running the system. It is a very hands on experiment and one needs to be on constant lookout to notice anything wrong with the imaging. With experience, one can tell what is wrong, by looking at the images themselves and suggest proper remedial action. In the worst case, a complete systematic troubleshooting has to be done both inside and outside the vacuum chamber to identify the fault and bring the system back to working order. The results obtained using the UHV STM system during a platinum oxidation study is presented in the next chapter.
48 Figure 2 1. Schematic representation of the phenomenon of quantum mechanical tunneling. Unlike in classical mechanics, there is a finite probability that a particle, with kinetic energy lower than the potential energy of a barrier, could traverse the forbidden region and reappear on the other side.
49 Figure 2 2 Schematic for STM operation using tunneling mechanism with feedback l oop. The tunneling tip is rastered in the X Y plane by applying drive voltages on the X and Y drive piezos of the scanner when the tip is placed close enough to the surface to be imaged to achieve tunneling current for a small bias voltage. The tunneling c urrent signal itself (constant height mode), or tip height adjustments needed in the Z direction required to maintain constant current using feedback (constant current mode), is recorded as the STM image.
50 Figure 2 3 Schematic showing the constant cu rrent scanning mode for scanning tunneling microscopy. The tunneling tip traces the topology of the surface to be imaged by maintaining the tunneling current at a preset value using a feedback loop. Figure 2 4 Schematic showing the constant height sc anning mode for scanning tunneling microscopy. The feedback is turned off and the tunneling current is measured as the tip is ratered across the surface without changing the tip height.
51 Figure 2 5 Schematic of t unneling stage of the microscope for vi bration analysis. The variation of the sample tip separation is response to forced oscillations of the isolation stage is viewed as an effective mass performing forced simple harmonic oscillations. The relative motion of the effective mass with respect to the isolation stage is calculated for forced oscillations of the isolation stage. Schematic based on the model presented in reference 
52 Figure 2 6 Transmission factor for vibration of the tunneling junction caused by forced oscillations of the vibration isolation stage as a function of the ratio of the driving frequency to the resonant frequency of the tunneling unit. Damping is given by For low vibration frequencies the transfer function drops exponentially with decrease in the driving frequency as At the resonant frequency, the asymptotic growth of the oscillations is limited by the damping. At higher frequencies, the tunneling gap changes according to the vibration of the tunneling unit.
53 Figure 2 7 Schematic of first isolation stage of STM for vibration analysis. The tunneling stage and isolat ion stage are considered together as a single block. The interest is to calculate the displacement of the vibration isolation stage for forced oscillations of the sample manipulator. Schematic based on the model presented in reference 
54 Figure 2 8 Transmission factor for vibration of the isolation stage caused by forced vibration of the sample manipu lator is plotted as a function of the ratio of the driving frequency to the resonant frequency of the isolation stage. The transmission factor is plotted for three different values of damping namely, and As the damping increases the amplification at the re sonant frequency decreases, but the isolation at higher frequencies decreases.
55 Figure 2 9 Overall transmission function for vibration of tunneling junction caused by forced vib rations of the sample manipulator through the first isolation stage and the tunneling unit. It is plotted as a function of the ratio of the driving frequency to the resonant frequency of the tunneling unit for three different values of damping and It is clearly seen from the plot that with a vibration isolation stage with low resonance frequency and low damping, vibrations to the system are attenuated sufficiently for driving frequencies between the resonant frequency of the tunneli ng unit and the isolation stage.
56 Figure 2 10 Schematic of UHV STM system. The chamber is welded onto a table frame using angled struts. The upper part of the chamber is a belljar which accommodates sample preparation and analysis. Stage 1 is exclusi vely for STM and Stage 2 consists of various preparation and analytic tools distributed radially around the chamber. The bottom part is a pump well which houses the ion and titanium sublimation pumps. The entire chamber is floated on four laminar flow vibr ation isolation legs during STM imaging. The sample manipulator rod housing the sample assembly is translated linearly using the motorized bellows and rotated using the differentially pumped rotary flange. Schematic is modified and updated from reference 
57 Figure 2 11 Schematic showing the sample holder assembly consisting of the STM ramp, sample, tantalum spring, sapphire ring insulator, and the bottom copper well. The heater is inserted into the sample holder from bel ow and does not make physical contact with the sample holder assembly. Thermocouple feedthroughs reach the sample through holes on the side of the STM ramp. The sample is electrically isolated from the copper well and biased during imaging. The copper well is connected to a LN reservoir for cooling (not shown here). Schematic is modified and updated from reference 
58 Figure 2 12 Vertical transmissibility for Newport I 2000 series laminar flow vibration isolation system t aken from Newport Vibration Control System Instruction Manual, copyright(1991), Newport Corporation. The extremely low resonant frequency (~1Hz) and small amplification near resonant frequency of the vibration isolation system is useful for atomic scale im aging.
59 Figure 2 13 Schematic depicting cross sectional view of the STM assembly. The tunneling wire is protected by a metal shield and ground shielding. The scan drive wires connecting the feedthroughs to the quadrants of the piezos are smoothed out and L shaped. The decoupled STM head sits on the STM ramp and the sample is exposed to the tunneling tip through the center cavity. The STM manipulator is mounted on a bellows for Z translation and a rotary platform for rotation. Schematic is modified and updated from reference 
60 Figure 2 14 Noise power spectrum indicating large noise around 60 Hz. Figure 2 15 Noise power spectrum indication reduced noise around 60 Hz.
61 CHAPTER 3 STM STUDY OF HIGH COVERAGE STRUCT URES OF ATOMIC OXYGEN ON P LATINUM (111): P(21) AND P LATINUM OXIDE CHAIN STRUCTURES Introduction Oxygen atoms can populate a variety of states or phases on transition metal surfaces during the course of oxidation, including phases that are intermediate to a dilute chemisorbed layer and the bulk oxide(s). The possibility that different surface oxygen states interact uniquely with reactant molecules is very interesting scientifically and poses a technological challenge for many applications of heterogeneous catalysis. In particular, strong variations in the reactivity among different oxygen species could cause the performance of a transition metal catalyst to depend sensitively on the extent of surface oxidation, and hence on reaction conditions. For example, Chen et al.  have recently found that several late transition metals exhibit dramatic increases in their activity toward CO oxidation in a narrow range of oxygen coverage near one monolayer. To explain this observation, Chen et al.  hypothesized that highly reactive oxygen states develop on the metal surfaces just prior to the onset of bulk oxidation Similarly, Gabas c h et al.  find that a precursor state to PdO formation on Pd(111) exhibits exceptionally high activity for CH 4 oxidation compared with other oxygen phases that develop on this surface. Of particular relevance to the present study is the Pt catalyzed oxidation of NO to NO 2 under oxygen rich co nditions, which is a critical step in the NO x storage/reduction system used in lean burn diesel engines to reduce NO x compounds to N 2  Recently, Mulla et al.  have presented evidence that the NO oxidation activity of Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr ., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
62 oxide supported Pt catalysts is higher for large Pt particles, which expose a large fraction of ( 111) facets, and that the activity diminishes under conditions favoring Pt oxide formation. In a related computational study, Ovesson et al.  report that NO oxidation is endothermic on clean Pt(111) and only becomes favorable at sufficiently high surface coverages of atomic oxygen due to a weakening in the O Pt(111) b ond strength with increasing coverage. Taken together, these studies suggest that NO oxidation on Pt(111) is quite sensitive to the surface concentration of oxygen atoms, and occurs at optimum rates in a relatively narrow window of oxygen coverage. While t hese are interesting and important findings, the origin for the strong dependence of the surface reactivity on the oxygen coverage is unclear, largely because the exact nature of the oxygen states that develop on Pt(111) at high coverage has not been fully resolved. The interactions of O 2 with Pt(111) have been extensively investigated under ultrahigh vacuum (UHV) conditions. The dissociative chemisorption of O 2 is facile on Pt(111) at temperatures as low as 150 K, but effectively ceases at an atomic oxyge n coverage of 0.25 ML due to kinetic limitations in O 2 bond activation in the presence of chemisorbed atomic oxygen. Note that we define 1 ML as equal to the Pt(111) surface atom density of 1.51 10 15 cm 2 [6 9] It is well established that ox ygen atoms bind at fcc hollow sites on Pt(111) below 0.25 ML, and arrange into ordered p(2 2) domains [17 21] At these low coverages, oxygen atoms recombinatively desorb during temperature programmed desorption (TPD) experiments to produce a single, broad 3 that shifts toward lower temperature with increasing initial coverage. T he desorption peak appears at about 730 K for an initial atomic
63 oxygen coverage of 0.25 ML. Researchers have shown that the O 2 TPD spectra obtained from Pt(111) are best reproduced by assuming that desorption is a second order process and that the activati on energy for desorption decreases with increasing coverage [7, 18, 19] which is indicative o f repulsive interactions between adsorbed oxygen atoms. Calculations based on density functional theory (DFT) also predict that the binding energy of oxygen atoms chemisorbed on Pt(111) decreases with increasing coverage [25, 26] Atomic oxygen coverages as high as 0.75 ML can be generated on Pt(111) in U HV through the dissociative chemisorption of NO 2 [18, 22] and even higher coverages can be attained using more aggressive oxidants such as O 3  and atomic oxygen [19, 24] The results of O 2 TPD experiments suggest that oxygen atoms sequentially 1 2 3 ) on Pt(111) as the oxygen coverage increases to 0.75 ML [18, 19, 22] Above 0.25 ML, the O 2 TPD spectra initially exhibit a second desorption feature ( 2 ) centered at about 630 K that intensifies as the coverage 1 ) at about 560 K then becomes evident, and primarily intensifies with increasing coverage from 0.50 to 0.75 ML. T he TPD results demonstrate tha t the O Pt(111) binding energy weakens significantly with increasing oxygen coverage However, spectroscopic measurements show that the vibrational  and electronic [18, 19] properties of atomic oxygen on Pt(111) differ only slightly among the states that develop below 0.75 ML. These observations imply either that oxygen atoms continue to o ccupy the same adsorption site or bind to different sites but maintain similar properties up to a coverage of 0.75 ML.
64 Past experiments using low energy electron diffraction (LEED) provide only limited information about the structural properties of the hi gh concentration states of atomic oxygen on Pt(111) [18, 19, 23] For example, beyond 0.25 ML LEED images continue to exhibit a (22) pattern th at fades only gradually as the oxygen coverage increases to 0.50 ML. This observation has lead researchers to suggest two possible structures for the 2 state. Firstly, a (22) LEED pattern would result at 0.50 ML if oxygen atoms continue to bind at fcc hollow sites and arrange into three rotationally degenerate p(21) domains. Alternatively, oxygen atoms could begin to bind at hcp hollow sites above 0. 25 ML and arrange into a p(22) structure containing two oxygen atoms per unit cell, with the final structure resembling a honeycomb. According to DFT calculations [25, 26] the p(21) structure should be preferred over the honeycomb structure, mainly because the oxygen binding energy on an hcp site is abo ut 50 kJ/mol weaker than on an fcc site [26, 94] However, TPD spectra obtained from P t(111) covered with a mixture of 18 O and 1 6 O atoms provide evidence that oxygen atoms adsorb on multiple sites at coverages above 0.25 ML, and hence imply that the honeycomb structure is preferred over the p(21)  L ess is known about the specific 1 state As the coverage increases above 0.50 ML, the (22) LEED pattern disappears and the substrate spots become dimmer, indicating that formation of the 1 state generally disrupts the long range order of the surface We have also reported that the 1 state begins to transform to three dimensional PtO 2 domains above 0.75 ML, and that this transformation involves sig nificant local restructuring of the surface, as evidenced by rapid deterioration of the Pt(111) LEED pattern as well as explosive desorption of O 2
65 during TPD  Considerin g the substantial restructuring that occurs during oxide formation, it is reasonable to expect that adsorption into the 1 state could induce some degree of local restructuring of the surface Pt atoms as well. As far as we know, only DFT calculations have provided information about the possible configurations of oxygen atoms in the 1 state. According to DFT calculations by Legare  oxygen atoms prefer to populate tetrahedral sites in the P t(111) subsurface at coverages above 0.50 ML rather than continuing to adsorb on surface sites H ow ever, more recent DFT calculations predict that direct oxygen diffusion into the tetrahedral sites has energy barriers between 150 and 300 kJ/mol  which implies that subsurface penetration is kinetically hindered at moderate surface temperature. Based on more recent computations, Getman et al.  conclude that oxygen binding on fcc surface sites continues to be favorable above 0.50 ML. However, the TPD results by Jerdev et al.  provide convincing evidence that oxygen atoms begin to bind in distinct sites or phases above 0.25 ML. Indeed, these discrepancies signal a need to clarify the oxygen states that develop on Pt(111) a t high coverage. In the present study, we used STM to investigate the structural properties of Pt(111) at oxygen coverages up to 0.75 ML. We find that oxygen atoms arrange into p(21) domains between 0.25 and 0.50 ML, with desorption from these ordered do mains producing the 2 feature observed in TPD. We also find that population of the 1 TPD state coincides with the appearance of bright protrusions and chains within p(21) domains, which we attribute to a Pt oxide chain compound that resides nearly 2 a bove the (111) surface plane. As the oxygen coverage increases, the Pt oxide chains assemble into an interconnected network of Y shaped structures consisting of
66 ~20 long branches. Overall, the STM results demonstrate the Pt oxide chain formation occurs d uring the early stages of oxidation, and that long range effects associated with surface stresses play a central role in determining the structural characteristics of the chain compound on Pt(111). Experimental Detail s The experiments were carried out in an ultrahigh vacuum (UHV) apparatus with a typical base pressure of 2 x 10 10 Torr. The UHV chamber is equipped with a scanning tunneling microscope, a four grid LEED optics, a quadrupole mass spectrometer (QMS) used for TPD experiments and an ion sputter source. The design and performance of this system has been described in detail elsewhere [97, 98] type microscope uses RHK STM 100 control electronics and a Pt/Ir tip. We typically prepared the tip for scanning via field induced evaporation onto a gold film (~4 uA, 2 0 min) in vacuum. The STM scan dimensions were determined based on the Pt(111) p(2 2) O structure, and heights were estimated from the known monatomic Pt(111) step height of 2.3 The Pt(111) single crystal used in this study is a circular disk (10 mm x 1.5 mm) that is cut and polished to within 0.4 of the (111) plane. The sample was radiatively heated by a tungsten filament and cooled through a thin copper braid connected to a coolant reservoir. The temperature of the sample was monitored using a type K thermocouple and a linear temperature ramp for TPD experiments was achieved using a temperature controller interfaced to a programmable DC power supply. Initially, sample cleaning consisted of repeated cycles of sputtering with 500 eV Ar + ions and O 2 trea tments, followed by annealing to 950 K. Subsequent cleaning involved routinely exposing the sample at 573 K to NO 2 followed by heating to 950 K. We considered the
67 sample to be clean when we could no longer detect impurities with AES, observed a sharp (1 1 ) LEED pattern and observed large, flat terraces in STM images. We recently connected a single stage differentially pumped beam chamber to the main UHV chamber to allow sample dosing with oxygen atom beams. This apparatus generates beams of oxygen atoms by partially dissociating pure O 2 (Praxair, 99.999%) continuously supplied to a small discharge chamber at the end of a microwave plasma source. Species exit the discharge chamber through a small aperture, and then enter a quartz tube (30 cm long, 6 mm ID) t hat directly views the sample surface held in the main UHV chamber. A mechanical shutter attached to the plasma source controls the exposure of the Pt(111) sample to the oxygen atom beam. Additional details of the microwave source are given elsewhere [19, 99] A 500 l/s turbomolecular pump and a titanium sublimation pu mp evacuate the beam chamber. This combination of pumps and the small conductance of the quartz tube produce a pressure difference of more than three orders of magnitude between the beam chamber and the UHV chamber during source operation. In most of the experiments, we generated adlayers with various atomic oxygen coverages by background dosing NO 2 with the sample held at fixed temperature. The NO 2 supplied from the vendor (Linde Gas, 99.5%) was used without further purification. We conducted O 2 TPD exper iments by facing the sample toward the QMS ionizer and then heating the sample at a constant rate of 1 K/s while monitoring the 32 amu signal. To estimate the initial atomic oxygen coverages, we scaled the area under each O 2 TPD trace with that obtained fo llowing a saturation exposure of NO 2 and assuming that this exposure produces an atomic oxygen coverage of 0.75 ML as reported previously
68 [18, 22, 100, 101] Toward the end of this study, we prepared a few oxygen covered surfaces by exposing the Pt(111) sample to an oxygen atom beam. These results are presented at the end of the manuscript. Notably, the STM images and TPD traces obtained from the same oxygen coverages prepared using NO 2 versus atomic oxygen beams are indistinguishable within the uncertainty limits of our measurements. Results and Discussion Thermal Desorption o f O Pt(111) Figure 3 1 sho ws O 2 TPD spectra obtained after generating atomic oxygen coverages on Pt(111) by dissociatively adsorbing NO 2 with the surface held at 450 K. Since NO desorbs promptly from Pt(111) at 450 K, the resulting surfaces are covered exclusively with atomic oxyge n. The TPD spectra shown in Figure 3 1 are very similar to those reported in prior UHV studies that have employed NO 2  O 3  and gas phase oxygen atoms  as oxidants. Since the O 2 TPD spectra from Pt(111) have been discussed in detail previously [18, 19] we provide only a brief description here. 3 ) arising from the recombinative desorption of oxygen atoms bound on fcc hollow sites and arranged in p(2 2) domains. Above 0.25 ML, a second desorption feature, denoted 2 3 feature, and increases in intensity with increasing coverage to 0.50 ML. Once the coverage exceeds about 0.40 ML, a third desorption feature denoted as 1 begins to appear as a shoulder near 580 K in the O 2 TPD spectra. Above 0.50 ML, 2 feature intensifies slightly without an appreciable shift in the peak temperature. In 1 feature grows sharply as the coverage increases above 0.50 ML, evolving into a distinct peak centered at about 560 K once the coverage reaches 0.75
69 1 desorption feature does populate slightly at 450 2 state, but 1 state mainly increases above 0.50 ML. Similar behavior in the development of the oxygen layer on Pt(111) prepared at 450 K using oxygen atom beams, has previously been re ported  Oxygen atoms populating the 1 state do not arrange into domains with sufficient long range order to be detected with LEED. Consequently, the structural characte ristics 1 state, and indeed the fundamental nature of this oxygen state on Pt(111) have remained unclear. STM Study o f O Pt(111) System We used STM to characterize the changes in surface structure that occur as the oxygen atom coverage increases to 1 2 states d evelop above 0.25 ML. Figure 3 2A shows a large scale STM image obtained after adsorbing 0.30 ML of oxygen atoms on Pt(111). A single step edge runs diagonally across the image and small islands sp arsely decorate the terraces and the step. We observe similar islands on clean Pt(111), and attribute these features to Pt domains. While our annealing procedure is apparently too mild to fully eliminate these structural features, the Pt islands do not det ectably influence the structural transformations that occur with increasing oxygen coverage ( vide infra ). The large scale image appears very similar to those obtained from the clean surface, and hence reveals that the Pt(111) substrate maintains large flat terraces (> 500 ) and a smooth surface morphology after the adsorption of 0.30 ML of oxygen atoms. The magnified image in Figure 3 2 B shows the familiar p(22) structure with equal spacing of oxygen atoms in the three high symmetry directions of the hex agonal (111) surface. The average values of the periodicity in atomic corrugation measured in the A,
70 B and C directions shown in Figure 3 2 B are 5.3 5.4 and 5.2 which are close to 2a, where a is the Pt(111) lattice constant of 2.77 The measured values are not identical to the theoretical value of 5.54 due to minor distortion in the images caused by thermal drift and experi mental limitations. Figure 3 2 B also shows a unit cell drawn for the p(2 2) structure. Although the hexagonal p(2 2) O struc ture is symmetric with respect to the occupied and empty oxygen adsorption sites, the dark spots appear more well defined in our STM images, which is similar to previous observations [20, 21] Simulations of STM images of the p(2 2) O structure on both Pt(111)  and Pd(111)  also predict that the oxygen atoms appear as depressions when the ST M tip is terminated by a metal atom. Thus, we believe that oxygen atoms of the p(2 2) structure appear as dark spots in the STM images obtained in the present study. Finally, the Z corrugation in our STM images is about 0.4 along the high symmetry direct ions of the p(2 2) structure, which corresponds to 17% of the height of a monatomic step on Pt(111). The 2 State: p (21) Striped Structure Upon initially increasing the atomic oxygen coverage above 0.25 ML at 450 K, STM images exhibit regions with a striped structure in the background and bright protrusions 5 to 6 in diameter distributed randomly throughou t the striped structure. As discussed in detail below, the number of protrusions increases with increasing coverage at 450 K, and the protrusions also begin to aggregate into clusters and chains. According to the TPD data (Figure 3 1), the 2 stat e mainly populates the surface at oxygen coverages from 0.25 to 0.50 ML; however, as noted above, small amounts of the 1 state also form below 0.50 ML. We therefore attribute the background striped structure to the 2 state and the protrusions to the 1 state. In order to confirm this
71 assignment, and also obtain better resolved images of the striped structure, we prepared oxygen adlayers by dissociatively adsorbing NO 2 on Pt(111) at a surface temperature of 573 K. Since this preparation temperature lies between the 1 and 2 peak temperatures in TPD (Figure 3 1), the 1 state should populate negligibly at 573 K, thereby allowing us to isolate the 2 state for characterization with STM. Indeed, the TPD spectra following NO 2 exposures at 573 K (not shown) exhibit only the 2 and 3 peaks, and the bright protrusions are no longer evident in the STM images. Figure 3 3 shows STM images obtained after adsorbing 0.40 ML of oxygen atoms on Pt(111) at 573 K. The large scale image (Figure 3 3 A ) shows a step edge in the top rig ht corner and terraces with a similar smooth morphology as observed at lower coverage. The magnif ied images shown in Figure 3 3B and Figure 3 3 C reveal that the striped structure arises from oxygen atoms arranged into a p(2 1) structure, wit h the micrograp hs in Figure 3 3 B and Figure 3 3 C corresponding to p(2 1) domains at different orientations. The measured corrugation of the p(2 1) structure exhibits periodicities of 5.9 3.3 and 5.5 along the high symmetry A, B and C directio ns, respectively (Figu re 3 3 C ). Also, the perpendicular distance between the close packed rows of the (2 1) structure is about 4.6 which is close to twice the spacing of 2.4 (= 3a/2) between adjacent rows of Pt atoms on the Pt(111) surface. We note that the A, B and C dir ections are identical to the high symmetry directions of the p(2 2 ) structure shown in Figure 3 2 B In fact, one can imagine the p(2 1 ) structure shown in Figure 3 3 C as identical to the p(2 2 ) structure shown in Figure 3 2 B but with oxygen atoms added bet ween the already existing oxygen atoms along the B direction, thus forming the characteristic (2 1) rows. Indeed, the periodicity measured along the B direction is
72 about half of that measured along the A and C directions and hence close to the Pt(111) latt ice constant of 2.77 Similar to the p(2 2) structure, the oxygen atoms of the p(2 1) also seem to image as depressions (see the below discussion). Finally, we estimate the Z corrugation along the high symmetry directions of the p(2 1) structure to be ab out 0.50 which is 23% of the monatomic step height. Overall, these STM results demonstrate that oxygen atoms arrange into p(2 1) domains at coverages between 0.25 and 0.50 ML on Pt(111), with desorption from these domains giving rise to the 2 state obs erved in TPD (Figure 3 1). The 1 State: One Dimensional (1D) Pt Oxide Chains Figure 3 4 shows STM images obtained after adsorbing 0.46 ML of oxygen atoms on Pt(111) at 450 K. According to TPD (Figure 3 1), oxygen atoms populate both the 1 and 2 states under these adsorption conditions, though neither state is saturated. The large scale micrograph in Figure 3 4 A shows that formation of the 1 state, even in relatively small amounts, causes considerable roughening of the Pt(111) terraces. At slightly hi gh er magnification (Figure 3 4 B ), the STM images exhibit bright protrusions distributed randomly across the terraces. The protrusions exhibit varying shapes and sizes in Figure 3 4 B Many of the protrusions are symmetric, but chain like structures are also e vident. Although their shapes vary, all of the protrusions have a similar apparent height of 1.7 which corresponds to 72% of the height of a monatomic step on the Pt(111) substrate. Figure 3 4 C and Figure 3 4 D show atomically resolved images of terrace areas that contain relatively few of the protrusions seen in the larger scale images. Figure 3 4 C shows coexisting p(22) and p(21) domains, where the close packed direction of the p(21) structure runs diagonally across the image, starting from the botto m right. A
73 bright protrusion can be clearly seen within the p(2 1) domain (arrow in Figure 3 4 C ). Close inspection reveals that the protrusion has dimensions similar to the atomic scale features of the 21 structure, and is positioned nearly along a bright row of the 21 domain, with slight offset toward the adjacent dark row. As elaborated below, we attribute the bright protrusions to Pt atoms that displace out of the surface plane as oxygen atoms adsorb onto sites between the close packed oxygen rows of t he p(21) structure. We note that this interpretation is consistent with attributing the alternating bright and dark rows in the STM image to close packed Pt and O atom rows, respectively. Figure 3 4 D shows that the oxygen induced protrusions begin to agg regate at aligned with the high symmetry directions of the p(21) structure. For example, the bright s tructures marked in Figure 3 4 D consist of chains that align parallel t o the close packed (21) rows seen at the top of the image, and connect with branches that propagate roughly perpendicularly to the close packed oxygen rows. The protrusions continue to cluster as the oxygen coverage increases. Figure 3 5 shows an image wi th numerous aggregates on Pt(111) covered with 0.58 ML of oxygen atoms. The p(21) structure is evident in this image, though it appears only faintly in the background with the close packed rows running diagonally across the image, starting from the bottom left. Here, the aggregates exhibit a range of shapes and sizes, and appear largely disordered. However, several of the bright features clearly consist of chains that have aggregated to form larger structures, with these structures exhibiting a characteris tic size of 20 to 25
74 The oxygen induced protrusions and chains on Pt(111) are consistent with a one dimensional (1D) Pt oxide compound that develops within the p(21) structure. While such structures have not been observed previously during oxygen adsor ption on Pt(111), chain structures have been observed along the low coordination sites of Pt(110)  and the stepped Pt(332) surface  after oxyge n adsorption. For example, Helveg et al.  find that high c overages of oxygen atoms generate bright chains along the ridges of Pt(110) in STM images, and show with DFT that the bright features arise from Pt atoms that buckle out of the surface by as much as 0.8 as a result of coordinating with multiple oxygen at oms. DFT predicts that strengthening of the O Pt bonds compensates the energy penalty caused by displacing the Pt atoms out of the surface plane, and results in energetically favorable structures. In a related study, Wang et al.  find that 1D PtO 2 chains form along the step edges of Pt(332) where these chains consist of Pt atoms with fourf old coordination with oxygen atoms. Since bulk PtO 2 is octahedral with sixfold O Pt coordination, the 1D PtO 2 chains represent an intermediate oxide compound. Using DFT, Hawkins et al.  have recently determined that chain formation is also energetically preferred when oxygen atoms are added to the p(21) O structure on Pt(111). Figure 3 6 shows a representat ion of an initial chain structure predicted to form after adding oxygen atoms to a p(21) layer. The extra oxygen atoms form a chain between the close packed oxygen rows of the p(21) structure and cause Pt atoms in an adjacent row to displace out of the s urface plane by about 1.7 (Figure 3 6, blue circles), which is in excellent agreement with the STM results. The large outward displacement of the Pt atoms and the oxygen clustering is clearly indicative of Pt oxide
75 compound formation. In particular, the DFT results suggest that formation of a Pt oxide chain may be viewed as a chemical reaction involving the formation of new Pt O bonds and cleavage of Pt Pt bonds between the first and second layers of the solid. Notice that the Pt atoms in the chain struct ure have three and four fold coordination with oxygen atoms. Friend and coworkers have also observed surface compound formation on Au(111) after adsorbing electronegative species, including S  O [105, 106] and Cl  Similar to these cases, strong charge transfer likely serves as the driving forc e for Pt oxide chain formation on Pt(111). Interestingly, the Pt oxide compound on Pt(111) bears remarkable similarity to the chain structures observed on Pt(110)  and Pt(332)  Taken together, these results provide strong evidence that oxygen addition to p(21) O domains on Pt(111) tends to initially produce a 1D Pt oxide compound rather than continuing to populate a new chemisorbed phase or sub surface sites as suggested by others [26, 96] The Saturated 1 State: Chain Branchi ng and Formation of a n Interconnected Network Figure 3 7 shows STM images obtained after adsorbing 0.71 ML of oxygen atoms on Pt(111) at 450 K using an oxygen atom beam. This coverage is slightly below the highest coverage that can be attained using NO 2 under vacuum conditions. The l arge scale image in Figure 3 7 A exhibits a high density of chain like protrusions on the Pt(111) terraces, giving the terraces a roughened appearance, while the step edges appear relatively smooth. At hig her magnification (Figure 3 7B and Figure 3 7 C ), we find that the chains predominantly align along three principle directions oriented at 120 relative to one another, clearly reflecting the three fold symmetry of the Pt(111) surface. The chains also inter
76 as well as hexagons, with each branch consisting of two to three side by side chains. A few bright domains can also be seen throughout the branched network which appear to be a new oxide st ructure. We estimate that each chain is about 1.7 in height, which is identical to the chain height observed at lower coverage. This similarity is consistent with the TPD data as it implies that oxygen atoms continue to form the same Pt oxide chain compo und (i.e., the 1 state) as the oxygen coverage increases from 0.50 to 0.75 ML. Apparently, however, it becomes favorable for the chains to organize into branched structures as the oxygen density increases. The chains that form at 0.71 ML are characterize d by a narrow distribution of lengths, with each branch of the Y structure having a characteristic length between 19 and 24 In other words, the chains show a strong tendency to propagate by about 7 to 9 Pt(111) lattice constants before branching in a ne w direction. The magnifie d image in Figure 3 7 D further reveals that one branch of the Y structure aligns parallel with the close packed oxygen rows of the local p(21) structure, with the other two branches running parallel to the equivalent directions of the underlying Pt(111) lattice. Finally, we estimate that the chain structures occupy approximately 50% of the surface area at a total oxygen coverage of 0.71 ML. Assuming that p(21) domains cover the remaining surface area, the area estimate suggests th at the chains have a local atomic oxygen coverage close to 1 ML, which is equal to the oxygen density within the 1D Pt oxide chain structure predicted by DFT (Figure 3 6). The close agreement between the predicted and estimated oxygen coverages within the chains strengthens the conclusion that the branched structures are comprised of the same 1D Pt oxide chains that form initially within the p(21) structure.
77 High resolution images provide further evidence that the 1D Pt oxide chains comprise the legs of t he branched structure, and reveal additional details of the atomic structure. Referring to the structure in Figure 3 8, the branches that run parallel to the A and B directions each consist of two chains about 21 long, while the branch in the C direction consists of three chains with lengths of 19.5 We also estimate that the chains aligned parallel with the A direction are separated by about 4.5 and each contains seven atoms with a spacing of about 3.1 The heights of the Pt oxide chains that comp rise an individual branch differ by about 30%; the center chain is clearly the tallest within a three chain branch (Figure 3 8, C direction). The chain separation of 4.5 agrees very well with the measured spacing between close packed oxygen rows of the p (21) structure (Figure 3 3), which is theoretically equal to 4.8 (= 3a). This agreement suggests that two to three individual 1D Pt oxide chains (Figure 3 6) close pack to form the legs of the branched structure. Also, the measured interatomic spacing of ~3.1 along a chain is considerably greater than the Pt(111) lattice constant of 2.77 implying that the chains experience significant compressive stress. The Pt Pt bond length in PtO 2 is reported to lie between 3.1 and 3.2 [108, 109] For this range of spacings, a Pt oxide ch ain requires only about 1% strain to become commensurate with Pt(111) at chain lengths ranging from 7a to 9a, which equals 19.4 to 24.9 Since the chain lengths observed in the present study fall exactly within this range, we conclude that achieving comm ensurability with the Pt(111) lattice dictates the specific Pt oxide chain lengths in the branched structures. The observed chain branching and development of an interconnected superstructure are characteristic of stress induced processes. Superstructures can be
78 expected when a two fold symmetric reconstruction occurs on a three fold symmetric surface, and are generally associated with equilibrating uniaxial stresses along three equivalent crystallographic directions. A good example of such a superstructur e is the herringbone reconstruction of Au(111)  Interestingly, surface chloride formation on Au(111) produces a similar honeycomb superstructure as observed here, though on a smaller length scale  The compressive stress experienced by an individual Pt oxide chain provides a large driving force for the chains t o connect and establish branch points that relieve this stress. While the Pt(111) surface acts as a template that dictates the formation of Y shaped structures, this chain arrangement also provides a branch point that minimizes the uniaxial stresses acting along the chains. As seen in Figure 3 7, the vast majority of branching points in the network involve three chains connected at 120 relative to one another. Thus, the tendency for the chains to assemble into Y shaped structures, rather than other configu rations, is a strong indication that stress relief is dominant in determining the geometric characteristics of the chain network. While the present results provide detailed information about the Pt oxide chain network, several aspects of this structure re quire further investigation. For example, chain pairing could perturb the structures of the individual chains, as evidenced by the observed height differences among the chains that comprise the branched structures (Figure 3 8). We also do not know the exte nt to which Pt oxide chain growth or branching alters the oxygen atom positions in nearby p(21) domains, particularly close to the chain (21) phase boundary. Because of these uncertainties, we are reluctant to propose a more detailed model of the Y struc tures. DFT studies are currently underway to further characterize the Pt oxide chain network.
79 Implications o f Pt Oxide Chain Formation Our discovery of the Pt oxide chain network on Pt(111) has several implications. Firstly, the formation of Pt oxide chai ns below 0.50 ML can explain TPD results by Jerdev et al.  which indicate that 18 O and 16 O atoms occupy different binding sites when these specie s are adsorbed sequentially on Pt(111) to reach total coverages above 0.25 ML. This observation is difficult to rationalize if one assumes that chemisorbed oxygen atoms populate only fcc sites. Also, the observation of chains below 0.50 ML demonstrates tha t Pt oxide compound formation becomes favorable on Pt(111) in the early stages of oxidation, and does not require chemisorbed oxygen atoms to accumulate in large quantities or oxygen atoms to occupy subsurface sites. As far as we know, surface oxide compou nds form on all of the late transition metal surfaces that have been studied in detail  and typically develop at sub monolayer oxygen coverages. This observation provides strong ev idence that surface oxide formation is a common step in the oxidation of late transition metal surfaces, and indeed reinforces the more general idea that surface metal atoms play a dynamic role in mediating adsorbate surface interactions. As mentioned abov e, charge transfer likely provides a strong driving force for Pt oxide chain formation on the Pt(111) surface. Further studies using DFT should be helpful for clarifying the underlying factors that cause Pt oxide chain formation to become favored over chem isorption with increasing oxygen coverage on Pt(111). Observing that Pt oxide chains form on the close packed (111) surface, in addition to the low coordination sites of Pt(110)  and Pt(332)  suggests that this compound plays a general role in determining the oxidation and catalytic behavior of Pt surfaces and nanoparticles. For example, in a prior study, we presented evidence that
80 1 state observed in TPD is a distinct precursor state to the formation of bulk like PtO 2 on Pt(111)  The present results confirm this interpretation and identify the precursor as the Pt oxide chain compound. Since polycrystal line Pt exposes large fractions of (111) facets as well as low coordination sites, Pt oxide chains are likely to mediate PtO 2 formation on Pt nanoparticles as well. Furthermore, oxygen covered Pt(111) is apparently most active for NO oxidation [10, 16] and possibly other reactions, at coverages for which the chain network is well developed. Thus, characterizing the reactive behavior of Pt oxide chains could provide important insights for understanding P t catalyzed oxidation reactions in oxygen rich environments. Summary We used STM to investigate high coverage structures of atomic oxygen that develop on Pt(111) up to a coverage of 0.75 ML. We find that oxygen atoms arrange into p(21) domains above a cov erage of 0.25 ML, and associatively desorb to yield the 2 1 desorption feature in TPD, bright protrusions and chains about 1.7 in height begin to appear within p(21) domains at a coverage of about 0.40 ML. We attribute these features to a 1D Pt oxide chain compound that buckles out of the surface plane, and forms as oxygen atoms occupy sites between the close packed oxygen rows of the p(21) structure. The Pt oxide chain compound has four fold Pt O coordination, and an oxygen density of approximately 1 ML. As the total coverage approaches 0.75 ML, the Pt oxide chains develop into an interconnected network of Y shaped branched structures with regions locally resembling a honeycomb. The bran ches of the Y shaped structure consist of two to three side by side Pt oxide chains, each with lengths between about 19 and 24 which corresponds
81 to 7 to 9 Pt(111) lattice constants. We suggest that uniaxial strain causes the chains to select specific le ngths and achieve commensurability along the three close packed directions of the Pt(111) substrate. The observed chain branching and formation of an interconnected superstructure is likely governed by stress relief mediated in large part by the three fold branch points of the Y structures. Overall, the present results demonstrate that formation of a Pt oxide chain compound occurs during the early stages of Pt(111) oxidation, and that long range effects dictate the structural characteristics of the chain ne twork. These findings may have important implications for understanding the mechanisms for oxidation as well as the catalytic properties of Pt surfaces in oxidizing environments.
82 Figure 3 1. O 2 TPD spectra obtained from Pt(111) with initial atomic oxyg en coverages of 0.07, 0.11, 0.29, 0.41, 0.47, 0.60, 0.63, 0.67, 0.71 and 0.75 ML. Each coverage was generated by exposing Pt(111) to NO 2 with the surface held at 450 K. A constant heating rate of 1 K/s was used to obtain each TPD spectrum. Reprinted with p ermission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
83 A B Figure 3 2. STM images of Pt(111) covered with 0.30 ML of oxygen atoms prepared by dissociatively adsorbing NO 2 at 450 K. A ) Large flat terraces (> 1000 ) with a step edge running from the bottom left to top right of image ( 200 mV and 1.06 nA, 1501 1501 ). B ) p(2 2) structure with the three high symmetry directions and the unit cell indicated ( 201 mV, 0.97 nA, 100 100 ). Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
84 A B C Figure 3 3. STM images of Pt(111) covered with 0.40 ML of oxygen atoms prepare d by dissociatively adsorbing NO 2 at 573 K. A ) Large flat terrace (> 500 ) with a step edge in top right corner of the image ( 201 mV, 1.02 nA, 499 500 ). B ) Magnified image of a p(2 1) domain ( 201 mV, 0.99 nA, 49.8 49.8 ). C ) Magnified image o f the p(2 1) structure with the three high symmetry directions and the unit cell indicated ( 201 mV, 0.97 nA, 100 100 ). Note that the slightly hazy vertical strips in the middle of image (B ) a nd on the right side of image (C ) arise from a scanning ar tifact. Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
85 A B C D Figure 3 4. STM images of Pt(111) covered with 0.46 ML of oxygen atoms prepared by dissociatively adsorbing NO 2 at 450 K. A) Large terraces (> 500 ) with rough morphology, and two step edges running from the bottom left to the top right ( 200 mV, 0.47 nA, 1501 1501 ). B ) Large terrace (> 500 ) with bright protrusions and chains decorating the surface (200 mV, 1.03 nA, 500 500 ). C ) Coexisting p(2 2) and p(2 1) domains with a bright protrusion evident in the latter ( 200 mV, 0.99 nA, 100 100 ). D ) Linear aggregates of protrusions within a p(2 1) domain ( 200 mV, 1.01 nA 102 100 ). Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
86 Figure 3 5. STM image of of Pt(111) covered with 0.58 ML of oxygen atoms prepared by disso ciatively adsorbing NO 2 at 450 K. Aggregation of chains is evident along the high symmetry directions of the faint p(2 1) stripes seen running diagonally across the image starting at the bottom left ( 199 mV, 0.94 nA, 200 200 ). Reprinted with permiss ion from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
87 Figure 3 6. Top and side view illustrations of a Pt oxide chain structure that forms on the Pt(111) p(21) O surface as determined from DFT. The atomic positions are shown only approximately. Large grey circles represent Pt atoms and yellow circles represent oxygen atoms in the p(2 1) structure. The blue and red circles represent Pt and O atoms of the chain compound. The side view il lustrates that the Pt oxide chain resides above the (111) surface plane. Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
88 Figure 3 7. STM images of Pt(111) co vered with 0.71 ML of oxygen atoms prepared at 450 K. A ) Large terraces (> 500 ) with a rough morphology, and step edges spanning the image ( 201 mV, 1.07 nA, 1000 10 00 ). B ) Network of branched structures with Y shapes ( 200 mV, 1.01 nA, 200 200 ). C ) Magnified view of the branched network ( 200 mV, 1.06 nA, 100 100 ). D ) Image of an isolated Y structure showing that one of the braches aligns with the closed packed rows of t he background p(21) structure ( 49.8 mV, 0.98 nA, 75.1 75.1 ). Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
89 Figure 3 8. High resolution image of a hexagonally shaped chain superstructure (200 mV, 0.99 nA, 49.8 49.8 ). Each branch consists of two to three Pt oxide chains, each between 19 to 22 long Reprinted with permission from S.P. Devarajan, J.A. Hinojosa Jr., and J.F. Weaver, Surface Science, 602 (2008) 3116. Copyright 2008, Elsevier
90 CHAPTER 4 FACILE C H BOND CLEAVAGE AND DEEP OXIDATION OF PROPANE ON A PALLADIUM OXIDE (101) THIN FILM Introduction Understanding alkane activation on metal based catalysts is central to developing more efficient and clean methods for generating electrical power through ca talytic combustion as well as for transforming saturated hydrocarbons to more useful forms. Because initial C H bond cleavage is typically rate determining in the reactive processing of alkanes by solid catalysts, considerable effort has been devoted to ch aracterizing both the molecular and dissociative adsorption of alkanes on clean transition metal surfaces  In contrast, however, relatively little work has been reported on the adsorption and activation of alkanes on transition me tal oxides. This is mainly because most TM oxides are expected to interact only weakly with alkanes and are typically unsuitable as catalysts for alkane processing. Palladium oxide (PdO) is an important exception as this oxide has been identified as the a ctive surface in the catalytic combustion of methane and other alkanes by Pd based catalysts. In fact, oxide supported Pd has been the traditional catalyst of choice for catalytic methane combustion because it performs exceptionally well under lean conditi ons due to the formation of PdO [32 43] Unfortunately, however, the fundamental understanding of PdO surface chemistry is rather limited due largely to difficulties in generating well defined PdO surfaces for model studies. In the present work, we investigated the adsorption and reactions of propane on a PdO(101) thin film Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society.
91 that was grown on Pd(111) under ultrahigh vacuum (UHV) conditions. We present evidence that propane activation by C H bond cleavage is high ly facile on the PdO(101) surface and leads to the complete oxidation of propane. Previous studies provide important insights for understanding the dissociative chemisorption of alkanes on transition metal surfaces. These studies reveal that the dissociat ive chemisorption can occur by two distinct mechanisms, namely, a direct and a trapping mediated mechanism  In the direct mechanism, the gaseous alkane molecule dissociatively chemisorbs during its initial collision with the solid surface. The direct mechanism dominates when alkane dissociation is strongly activated, and reaction must therefore be promoted by energizing the incident alkane molecules. In the trapping mediated mechanism, the alkane first adsorbs into a molecular stat e on the surface and then dissociates. In this process, the molecularly adsorbed, or so called trapped state, acts as a precursor to dissociative chemisorption, and a kinetic competition between dissociation and desorption from the molecularly adsorbed sta te determines the net rate of dissociative chemisorption. Trapping into the molecular state affords the alkane with more time, compared with a single gas surface collision, to explore the potential energy surface. However, since the alkane molecule dissipa tes its energy to the surface during trapping, dissociation from a molecularly adsorbed state requires the existence of a low energy pathway(s) for C H bond cleavage on the potential energy surface. Dissociation by the direct mechanism is more common than the trapping mediated mechanism since cleavage of the strong C H and C C bonds of alkanes is typically energetically demanding. However, trapping mediated dissociation is reported to be
92 facile on a few metal surfaces, including the reconstructed Pt(110)  and Ir(110) [55 57] surfaces, and also defect sites of Pt(111)  Dissociation is termed facile in these systems because the activation energy for dissociation of the molecularly adsorbed alkane is lower than that for desorption. This leads to an appar ent, negative activation energy with respect to the gas phase zero, and causes the net rate of dissociative chemisorption to decrease with increasing surface temperature. Since the molecular binding energy of a given alkane shows only small variation among different metal surfaces, the facile dissociation observed for the (110) surfaces of Pt and Ir appears to be a consequence of lower energy barriers for C H bond cleavage compared with other surfaces. Nevertheless, the extent to which the initial molecule surface interaction contributes to lowering the dissociation barrier remains an open question. The dissociative chemisorption of alkanes on well defined oxide surfaces is a relatively unexplored area, but may have characteristics that are analogous to alk ane activation by transition metal complexes in solution. This seems plausible considering the more localized nature of the chemical bonds in an oxide compared with a metal. For complexes with high electron density, low valent metals, alkane activation typ ically occurs by oxidative addition of the metal atom across a C H bond, which results initially in an alkyl hydride compound [111 113] There is s ubstantial evidence that an alkane sigma complex is an intermediate in the pathway for C H bond cleavage in many oxidative addition reactions of transition metal complexes with alkane molecules [111 113] An alkane sigma complex results from a donor acceptor interaction between the alkane molecule and the metal atom (i.e. a dative bond). While the dative bond is relatively weak (~20 to 60 kJ/mol), it can alter the characteristics of the C H bond in a
93 way that makes the bond easier to cleave, in addition to prolonging the time that the alkane molecule spends in close proximity to the metal center. In principle, similar complexation could occur during t he interaction between an alkane molecule and an oxide surface. In this case, alkane dissociation on an oxide surface may occur by a precursor mediated process where an alkane surface complex acts as the precursor to dissociation. In this study, we invest igated the adsorption and reactions of propane on PdO(101) using temperature programmed reaction spectroscopy (TPRS), and present evidence of facile C H bond cleavage of propane at surface temperatures below 200 K, followed by complete oxidation of the hyd rocarbon fragments above 400 K. From measurements of product yields as a function of surface temperature, we show that the dissociative chemisorption of propane on PdO(101) is well described by a precursor mediated mechanism with a negative, apparent activ ation energy. Finally, we find that propane adsorbs into a strongly bound molecular state on PdO(101), and present evidence that this state serves as the precursor to dissociation. Experimental Methods Previous studies [64, 114] provide details of the three level UHV chamber utilized for the present experiments. The Pd(111) crystal employed in this study is a circular disk (8 mm ~1 mm) spot welded to W wires and attached to a copper sample holder in thermal conta ct with a liquid nitrogen cooled reservoir. A type K thermocouple spot welded to the backside of the crystal allows sample temperature measurements. Resistive heating, controlled using a PID controller that varies the output of a programmable DC power supp ly, supports maintaining or linearly ramping the sample temperature from 81 K to 1250 K. Initially, sample cleaning consisted of sputtering with
94 600 eV Ar + ions at a surface temperature of 900 K, followed by annealing at 1100 K for several minutes. Subsequ ent cleaning involved routinely exposing the sample held at 856 K to an atomic oxygen beam for several minutes, followed by flashing the sample to 923 K to desorb oxygen and carbon oxides. As discussed previously  we limited the sample temperature to 923 K to maintain oxygen saturation in the subsurface reservoir, and thereby ensure rep roducibility in preparing the PdO(101) thin films used in this study. We considered the Pd(111) sample to be clean when we could no longer detect contaminants with X ray photoelectron spectroscopy (XPS), obtained sharp low energy electron diffraction (LEED ) patterns consistent with the Pd(111) surface, and did not detect CO production during flash desorption after oxygen adsorption. A two stage differentially pumped chamber attached to the UHV chamber houses the inductively coupled RF plasma source (Oxford Scientific Instruments) utilized to generate beams containing oxygen atoms for this study  This system produces gaseous oxygen atoms by partially dissociating pure O 2 (BOC gases 99.999%) continuously supplied to a s mall discharge chamber at the end of the plasma source. All oxygen atom beams used in this study were generated using an RF power of 120 W and an O 2 flow rate that establishes a pressure of 3 10 6 Torr in the first pumping stage of the beam chamber. Unde r these conditions, we estimate that 20% of the inlet O 2 dissociates in the plasma. To ensure uniform impingement of the oxygen beam across the sample surface, we positioned the sample approximately 50 mm from the end of the quartz tube that serves as the final beam collimating aperture, and with a 45 rotation with respect to the tube axis. The combination of this sample positioning and beam composition corresponds to an atomic oxygen flux with a lower bound of 0.02
95 ML/s at the sample surface, where 1 ML i s defined as the atomic density of 1. 53 10 15 cm 2 of bulk terminated Pd(111)  We produce a PdO(101) thin film on Pd(111) by exposing the metal sample at 500 K to an ~12 ML dose of gaseous ox ygen atoms supplied in a beam. This procedure generates a high quality PdO(101) film that has a stoichiometric surface termination, contains ~3.0 ML of oxygen atoms and is about 12 thick  The structure of the PdO(101) surface is discussed in detail below. Propane (Matheson, 99.993 %) was delivered to the sample from a calibrated beam doser at an incident flux of approximately 8.72 10 4 ML/s We set the sample to doser distance to about 50 mm to ensure uniform impingem ent of the propane across the sample surface. Due to this large distance, the direct flux of propane at the sample surface was only about 2.5 times higher than the background flux. After the propane exposures, we conducted TPD experiments by positioning th e sample in front of the QMS and then heating at a constant rate of 1 K s 1 until the sample temperature reached 923 K. The PdO thin films completely decompose when heated to 923 K so it was necessary to prepare fresh PdO films for each propane adsorption experiment. We estimated propane coverages by scal ing the integrated desorption spectra of mass 29 amu by an integrated TPD spectrum collected from a monolayer of propane adsorbed on Pd(111) at 110 K, and assuming that the monolayer saturates at 0.205 ML on Pd(111). This value corresponds to the saturation coverage of the propane monolayer on Pt(111) as determined from calibrated molecular beam experiments by Carlsson and Madix  Kao and Madix  have shown that the binding energy of propane is similar on Pd(111) and Pt(111) so it seems reasonable to assume that the
96 propane monolayer saturates at the same coverage on these surfaces. We estimated the prop ane incident flux at the sample surface by fitting the initial portion of the propane uptake curve on Pd(111) with a linear function, and assuming an initial trapping probability of unity. Since the propane binding energy on Pd(111) is approximately 45 kJ/ mol ( vide infra ), the trapping probability of propane molecules with thermal kinetic energies (~2.5 kJ/mol) should be close to unity  Both direct and background components contribute to the total flux estimated from the uptake curve. While the initial uptake curve is linear up to at least 0.06 ML, the linear fit has an intercept of about 0.006 ML. As elaborated below, we a ttribute the non zero intercept to an approximately constant background exposure of propane that occurs after terminating the beam flow into the chamber. We find that propane molecules react with the PdO(101) surface to produce CO 2 and H 2 O which desorb du ring temperature programmed reaction spectroscopy (TPRS) experiments. To estimate the CO 2 desorption yields, we conducted TPRS experiments of CO oxidation on Pd(111) covered initially with adsorbed oxygen atoms. We performed these experiments by adsorbing less than 0.25 ML of oxygen atoms on Pd(111), and then exposing the surface to large doses of CO with the sample held at 100 K. In subsequent TPRS experiments, only CO and CO 2 evolve from the surface, indicating that atomic oxygen was the limiting reactant The measured CO 2 TPRS peak areas were then set equal to the initial oxygen coverages, which were determined in separate TPD experiments conducted immediately before the CO oxidation experiments. As discussed previously, we estimate atomic oxygen coverage s on Pd(111) by scaling integrated O 2 TPD spectra with those obtained after exposing
97 Pd(111) at 300 K to a saturation dose of O 2 and assuming that the O 2 exposure generates an atomic oxygen coverage of 0.25 ML. Results and Discussion Structure of PdO(101 ) Thin Film o n Pd(111) Figure 4 1 depicts the structure of the stoichiometric PdO(101) surface that we examined in this study. Bulk crystalline PdO has a tetragonal unit cell and consists of square planar units of Pd atoms fourfold coordinated with oxygen atoms. The bulk terminated PdO(101) surface is defined by a rectangular unit cell of dimensions a = 3.04 and b = 6.14 where the a and b lattice vectors coincide with the  and [ 101] directions of the PdO crystal, respectively. The PdO(101) PdO su rface consists of alternating rows of threefold or fourfold coordinated Pd or O atoms that run parallel to the a direction shown in Figure 4 1. Thus, half of the surface O and Pd atoms are coordinatively unsaturated (cus) and likely to be more active than the fourfold coordinated atoms for binding adsorbed molecules. The rows of cus Pd atoms and cus O atoms are also adjacent to one another, and hence establish a unique binding environment for adsorbing molecules. The areal density of each type of coordinati vely distinct atom of the PdO(101) surface is equal to 35% of the atomic density of the Pd(111) surface. Hence, the coverage of cus Pd atoms is equal to 0.35 ML (monolayer), and each PdO(101) layer contains 0.7 ML of Pd atoms and 0.7 ML of O atoms. Given t hat the film contains ~3.0 ML of oxygen atoms, we estimate that the PdO(101) film on Pd(111) consists of between four and five layers and has a total thickness of ~12 In a prior study  we found that the PdO(101) struc ture aligns with the close packed directions of the Pd(111) substrate, and would expand by 0.46%
98 and 3.4% in the a and b directions to achieve commensurability with the metal substrate, which corresponds to unit cell dimensions of a = 3.06 and b = 6.35 TPRS from Propane Saturated PdO (101) Figure 4 2 shows TPRS spectra of masses 2, 18, 28, 29 and 44 amu obtained after exposing the PdO(101) film to a saturation dose of propane at 85 K. Desorption of propane from the PdO(101) surface gives rise to the pe aks centered at 120 and 190 K in the TPD spectra of masses 28, 29 and 44 amu as well as the broad region separating these peaks. We made this assignment after determining that the relative intensities of the low temperature peaks in the 28, 29 and 44 amu T PD spectra match those of the C 3 H 8 fragmentation pattern measured with our mass spectrometer. By comparison with propane TPD spectra obtained from a saturated monolayer on Pd(111), we estimate that ~0.15 ML of C 3 H 8 desorbs from the PdO(101) surface after s aturation at 85 K. The low desorption temperatures strongly suggest that propane is initially adsorbed on PdO(101) in molecular form, and desorbs to yield the peaks at 120 and 190 K. In addition to the peaks below 200 K, the 44 amu desorption trace exhib its a sharp peak centered at 463 K. A similar, though much smaller feature is also apparent at 463 K in the 28 amu desorption trace. In contrast, the 29 amu desorption trace shows no signs of a peak at 463 K. We therefore attribute the desorption peak at 4 63 K to CO 2 and conclude that a fraction of the adsorbed propane is oxidized by reaction with the PdO(101) surface. Cracking of CO 2 in the ionizer to CO can explain the small feature observed at 463 K in the 28 amu trace. We estimate that ~0.11 ML of CO 2 desorbs from the propane saturated PdO(101) surface during TPD, and hence that ~0.04 ML of the adsorbed C 3 H 8 oxidizes. This corresponds to 20% of the propane initially adsorbed on the oxide surface at saturation. We note that oxidation of ~0.04 ML of propa ne would
99 consume 0.4 ML of oxygen atoms, which is slightly greater than the amount of cus O atoms in the PdO(101) surface. Since propylene readily oxidizes on PdO surfaces  we checked the propane gas for propylene contamination by conducting T PD measurements with propane exposed Pd(111). Propylene partially dehydrogenates on Pd(111), liberating H 2 during TPD  We observed only propane desorption during the TPD experiments with Pd(111) and therefore conclude that propylene contamination of the propane gas was negligible. The H 2 O (18 amu) desorption trace provides further evidence that propane oxidizes on the PdO(101 ) surface. Firstly, based on the amount of CO 2 produced, and assuming that each oxidized propane molecule produces stoichiometric amounts of CO 2 and H 2 O, we estimate that ~0.15 ML of water desorbs during TPRS of the propane saturated surface. This estimate of the H 2 O yield agrees well with values determined by comparing the H 2 O desorption trace (Figure 4 2) with TPD spectra obtained from water covered PdO(101), for which the coverages have been previously determined  The 18 amu desorption trace exhibits peaks at 345 and 463 K, with the latter being more intense (Figure 4 2). We attribute the peak at 345 K to the oxidation of adsorbed H atoms by the PdO surface, where the H atoms are produced by the dissociation of propane at temperatures below 345 K. We reached this conclusion by considering the results of separate TPD experiment s in which we observe a water desorption peak at 345 K from hydrogen covered PdO(101). The H 2 O TPD peak at 463 K must therefore arise from a reaction limited process. Indeed, the coincidence of the H 2 O and CO 2 peaks at 463 K is a strong indication of a rea ction limited process wherein the oxidation of a hydrocarbon fragment(s) produces H 2 O and CO 2 above their
100 desorption temperatures. In TPRS experiments of propylene oxidation on PdO covered Pd(100), Yun et al.  also observe reaction limited deso rption of H 2 O and CO 2 at a similar peak temperature (T des = 490 K) as observed here, and no other reaction products except for small amounts of CO. Thus, both propylene and propane oxidation on PdO surfaces appear to occur by a direct pathway that involves rapid oxidation of the carbon backbone once a threshold temperature is surpassed during TPRS. This similarity implies that a common reaction step limits the rate of oxidation of both hydrocarbon species. The relative yield of reaction to desorption limit ed H 2 O suggests that propane dissociation initially generates adsorbed propyl groups, rather than other fragments, and that complete oxidation of the propyl groups produces the reaction limited water at 463 K. Integration of H 2 O TPRS spectra obtained from propane saturated PdO(101) indicates that the amount of water desorbing in the 463 K peak is between 6.1 and 6.8 times greater than the amount of water desorbing in the 345 K peak. Dissociation of a propane molecule by single C H bond cleavage would produc e a C 3 H 7 group and an H atom. If all of the propyl groups oxidize by a pathway that yields CO 2 and H 2 O above 345 K and all of the adsorbed H atoms oxidize to H 2 O below 345 K, then the ratio of the high to low temperature H 2 O peak areas would equal seven. S ince the measured value of this peak ratio is only slightly less than seven, it is reasonable to conclude that propane dissociation initially occurs by cleavage of a single C H bond, and that complete oxidation is the dominant fate of the resulting propyl species. We note that adsorption of small amounts of H 2 or H 2 O (< 0.01 ML) from the background gas can contribute to the H 2 O peak at 345 K, causing the intensity of this peak to increase
101 relative to the reaction limited H 2 O peak at 463 K. This may explain why the measured peak ratio is slightly less than seven. Although we monitored mass fragments of several potential reaction products in selected TPD experiments, we identified only CO 2 and H 2 O as reaction products resulting from the interaction between di ssociated propane and the PdO(101) surface. For example, we were unable to detect partially oxidized species such as 1 propanol, 2 propanol, propionaldehyde and acetone, or hydrocarbon species other than propane. Furthermore, the 28 amu trace does not exhi bit distinct features that could be associated with CO desorption, and H 2 desorption is negligible during TPD (Figure 4 2). Our inability to detect gaseous products of partial oxidation, particularly CO and H 2 further supports the conclusion that complete oxidation of dissociated C 3 H 8 is highly favored on PdO(101) under the conditions examined. The data also suggests that adsorbed propyl groups do not undergo appreciable dehydrogenation on PdO(101) below about 345 K, i.e., the desorption temperature of de sorption limited water. If dehydrogenation did occur below 345 K, then the relative yield of reaction to desorption limited water would likely be lower than seven and/or H 2 desorption would be observed. An interesting possibility is that the propyl groups remain stable on PdO(101) over a relatively wide temperature range, conceivably up to the onset of CO 2 production. Such stability would provide opportunities to characterize the properties of adsorbed propyl groups on PdO(101). Further studies are clearly needed to fully characterize the stability of adsorbed propyl groups and the mechanism(s) governing their oxidation on PdO(101).
102 The overall scenario that emerges from the data is that propane adsorbs molecularly on PdO(101) and that a fraction of the mol ecularly adsorbed propane dissociates by C H bond cleavage to produce adsorbed propyl groups and H atoms. The H atoms react with the PdO surface to produce H 2 O that desorbs near 345 K, while the propyl groups undergo complete oxidation, with CO 2 and H 2 O de sorbing above about 400 K. Intriguingly, the results strongly suggest that adsorbed propane molecules dissociate below their peak desorption temperature (< 200 K), but that reactions between the resulting propyl species and the PdO(101) surface generate ga seous products only above about 400 K. This observation demonstrates that propane dissociation on PdO(101) and subsequent surface reduction occur in temperature regimes that are sufficiently distinct that these reactive processes can be investigated separa tely. As elaborated below, we studied the dissociation of propane on PdO(101) at temperatures between 250 and 300 K, and find that the dissociation kinetics are well described by a trapping mediated mechanism with a negative, apparent activation energy, wh ich can be estimated from temperature dependent rate data. Propane Adsorption on Pd O (101) as a Function o f Coverage Fig ure 4 3 A shows a series of propane TPD spectra obtained as a function of coverage following propane adsorption on the PdO(101) thin film at 85 K. We calculated the total coverages listed in the figure by adding the amount of propane that desorbs to that which oxidizes to CO 2 and H 2 O. This estimate assumes that complete oxidation is the only reaction pathway for the dissociated propane. Pro pane initially desorbs in a single peak ( 1 ) that shifts from 200 to 190 K with increasing propane coverage. A second TPD feature becomes evident at about 155 K as the coverage increases above
103 0.05 ML, and apparently develops into a distinct peak ( 2 ) at 1 20 K once the coverage reaches 0.12 ML. The peak at 120 K continues to grow as the coverage increases to saturation at 0.19 ML, which is equivalent to 0.81 C atoms per surface Pd atom. A shoulder at 110 K is also evident in the TPD spectrum obtained from t he propane saturated PdO surface. Notice that the 1 peak at 190 K continues to intensify slightly as the low temperature desorption features develop above 0.05 ML. This behavior suggests that propane molecules only partially fill the strongly bound 1 sta te before starting to bind into less stable configurations at a surface temperature of 85 K. Figure 4 3 B shows a series of propane TPD spectra obtained after preparing different propane coverages on Pd(111) at 85 K. Propane molecules initially physisorb o n the metal surface, and desorb in a narrow peak centered at 153 K during TPD. Desorption from the second layer gives rise to the smaller peak at about 115 K, and begins to occur at a total coverage above about 0.14 ML. As discussed in the Experimental sec tion, we assume that the propane monolayer on Pd(111) saturates at a coverage of 0.205 ML, which is the value for propane on Pt(111)  The saturat ion coverage is determined by both intermolecular interactions and the molecule surface interaction, and is relatively low, in part, because propane molecules adsorb in a flat lying geometry on close packed metal surfaces  Using t he Redhead equation to obtain a rough estimate, we calculate that the binding energy of propane on Pd(111) is 44.9 kJ/mol for a desorption pre factor of 10 15 s 1 The value of the pre factor was chosen based on detailed analysis of alkane TPD spectra by Ta it et al. [123, 124] These workers used an inversion optimization method to estimate desorption kinetic parameters of molecularly adsorbed alkanes, and report pre factors of 10 1 4.8 and 10 15 .6
104 s 1 for propane desorption from Pt(111)  and MgO(100)  respectively. While the lower value is pr obably more appropriate for Pd(111), we chose 10 15 s 1 to make a reasonable comparison of binding energies estimated for propane on both Pd(111) and PdO(101), as discussed below. Comparing the propane TPD spectra obtained from the metal and the oxide surfa ces illuminates key features of the interaction of propane molecules with the PdO(101) surface. Firstly, the propane saturation coverages are nearly the same on the metal and the oxide surfaces (0.205 vs. 0.19 ML). This similarity suggests that both the 1 and 2 desorption peaks arise from propane molecules adsorbed directly on the PdO(101) surface, and hence occupying two distinct first layer states. In this case, the 70 K difference in the TPD peak temperatures would indicate that propane molecules encou nter significantly different binding environments in the 1 vs. 2 states. For example, the 1 state could correspond to molecules that interact strongly with cus Pd and/or O atoms, while the 2 state could arise from molecules that experience weaker inter actions with the coordinately saturated surface atoms. If we assume that only molecules bound in the 1 state react to produce CO 2 and H 2 O, then we estimate that the propane coverages in the 1 and 2 states are approximately the same at 0.09 ML when the p ropane layer is saturated at ~0.19 ML. Interestingly, the densities of threefold and fourfold coordinated surface atoms are also equal (~0.70 ML) on the PdO(101) surface. Thus, it is reasonable to conclude that half of the propane molecules ( 1 ) interact m ainly with cus atoms of the surface while the other half ( 2 ) interact mainly with coordinatively saturated atoms.
105 A particularly intriguing difference between the propane TPD spectra obtained from the metal versus the oxide is that the 1 peak temperatur e is about 40 K higher than the peak desorption temperature of the propane monolayer on Pd(111) (Figure 4 3). In general, the polarizability of metals is greater than oxides so one expects stronger physisorption bond strengths on metals. Consistent with th is expectation, Tait et al. report binding energies of 29 and 41.5 kJ/mol for propane physically adsorbed on MgO(100) and Pt(111), respectively [123, 124] Using the Redhead equation and a pre factor of 10 15 s 1 we estimate binding energies of 35.7 and 57.2 kJ/mol for propane adsorbed in the 2 and 1 states on PdO(101), respectively. From the Redhead analyses, we estimate that the binding energy of propane in the 2 state is about 9 kJ/mol lower than that of propane physisorbed on Pd(111). Thus, the low binding energy of the 2 state is indicat ive of propane molecules that are physically adsorbed in a first layer state on the PdO surface, where the binding is weaker than on the metal due to a smaller electric polarizability of PdO compared with Pd. By a similar argument, we conclude that the st rong binding of propane in the 1 state is inconsistent with a physically adsorbed state and therefore assert that the corresponding propane PdO(101) interaction involves electrostatic interactions between the molecule and the oxide surface in addition to the dispersive interaction. A possibility is that the propane molecule forms a dative bond with the cus Pd atoms of the surface, resulting in a bound state that is analogous to so called alkane sigma complexes that have been observed in studies with mononu clear TM compounds  The bonding in an alkane sigma complex involves electron donation from a C H bond(s) into empty orbitals of the metal atom, and, in some cases, back donation from the metal to the
106 alkane LUMO. Bond strengths of alkane sigma complexes are reported to lie in a range from roughly 20 to 50 kJ/mol [111, 112] In addition to dative bonding, hydrogen bonding between the hydrogen atoms of propane and the cus O atoms of oxide could contribute to the molecule surface i nteraction as well. Each of these interactions could enhance the molecule surface binding in the 1 state beyond that of propane physisorbed on the metal. Desorption a nd Reaction Yields: Selective Dissociation from t he 1 State An evaluation of the propan e desorption and reaction yields suggests that propane 1 binding state on PdO(101). Figure 4 4 shows the yields of propane that oxidizes, propane that desorbs in the 1 peak and the total propane yield, which we define as the amount of propane that oxidizes plus the total amount of 1 propane increase steadily with propane exposure up to an exposure of about 0.085 ML, at which point both curves exhibit an abrupt decrease in slope and thereafter increase more gradually with exposure. In contrast, the total propane yield continues to increase smoothly as the propane exposure increases beyond 0.085 ML, and starts to plateau only above an exposure of about 0.17 ML. As seen in Figure 4 2, propane adsorption into the weakly bound 2 state(s) is primarily responsible for the total propane yield increasing above an exposure, or equivalently, a total coverage of ~0.085 ML. The similar evolution of the 1 d esorption yield and the reacted propane suggests that the 1 molecular state acts as the precursor for propane dissociation on PdO(101) under the conditions examined.
1 07 Preliminary experiments of propane adsorption on other surface oxygen phases on Pd(111) provide additional support that the 1 state is the precursor to facile propane dissociation. We conducted TPRS measurements after adsorbing propane on the Pd 5 O 4 oxide  as well as meta stable surface oxides  that form on Pd(111) at oxygen coverages below about 0.70 ML. On these oxides, we observe no evidence for propane dissociation or oxidation and find that molecular propane ev olves only in a single TPD peak at 133 K, which indicates that the propane surface interaction is weaker on the surface oxides than in the 1 state on PdO(101). Indeed, we observe propane dissociation only when PdO(101) domains are present on the Pd(111) s urface, and the 1 TPD peak at 190 K can be detected. Several factors could enable the 1 state to act as a precursor to propane dissociation. Firstly, a propane molecule bound in the 1 state may adopt a geometric configuration that is amenable to C H bo nd scission, i.e. the 1 configuration could lie on the dissociation coordinate. The strong molecule surface interaction in the 1 state H bond, making it easier to break. For example, agostic bonding shifts electro n density out of the C H bond, and the resulting charge depletion could facilitate H atom transfer to a neighboring O anion of the surface. An agostic interaction can also involve back donation of electrons into the LUMO of the propane molecule that would weaken the C H bond. Further studies on the bonding characteristics of the 1 state are clearly needed to address these questions. It is an intriguing possibility that bonding in the 1 state is similar to the agostic bonding in an alkane sigma complex. Th is analogy could aid in developing a deeper understanding of the electronic factors that facilitate C H bond activation by TM oxide surfaces.
108 Kinetic Analysis o f Propane Dissociati on: Trapping Mediated Mechanism The present data provides convincing evidenc e that propane dissociation on PdO(101) occurs by a trapping mediated mechanism wherein dissociation occurs from a molecularly adsorbed state and a kinetic competition between dissociation and desorption determines the net rate of dissociative chemisorptio n. Furthermore, the observation of propane dissociation at low temperature is characteristic of a facile system for which the energy barrier for dissociation is lower than that for desorption. In this case, the apparent activation energy for dissociation i s negative, and the net rate of dissociation consequently decreases with increasing surface temperature. Trapping mediated dissociation has been reported previously for alkane adsorption on certain transition metal surfaces  In ge neral, trapping mediated dissociation is thought to dominate over direct dissociation when low energy pathways for dissociation are accessible during the lifetime of a molecularly adsorbed species. The dissociative chemisorption of an alkane molecule by a trapping mediated process may be represented by the following kinetic scheme, RH(g) RH(ad) R(ad) + H(ad), where RH(g) represents a gaseous alkane molecule, RH(ad) represents a molecularly adsorbed alkane, R(ad) represents an adsorbed alkyl group and H(ad) represents an adsorbed hydrogen atom. The reaction equation written above considers that alkane dissociation occurs by cleavage of a single C H bond. The kinetic analysis that follows has been previously shown to accurately describe the trapping med iated dissociative chemisorption of alkanes on metal surfaces  In the kinetic description, the rate of
109 flux of gaseous alkane molecules at the surface, multiplied by the probability for alkane trapping into the molecular state, In addition, alkane desorption and dissociation are each treated as elementary, first order processes governed by the rate coefficients k d and k r respectively. Dissociation and desorption provide competing pathways for alkane molecules to depart the mo lecularly adsorbed state. Note that the r of dissociative chemisorption is given by r r = k r [RH], where [RH] is the surface coverage of molecularly adsorbed species. This expression neglects any dependence of the dissociation rate on the coverage of R(ad) and H(ad), and is therefore strictly valid only in the limit of zero coverage of these species. Similarly, one may invoke the pseudo steady state approximation to evaluat e the coverage of molecularly adsorbed species in the limit of zero coverage of RH(ad). This simplification leads to the following expression for the initial probability for alkane dissociative chemisorption, (4 1) This expression shows that the initial dissociation probability is given by the initial probability for alkane molecular adsorption multiplied by the branching probability for dissociation from the molecularly adsorbed state. Assuming that the rate coefficients depend on temperature according to the Arrhenius equation, the model also predicts that the initial dissociation probability will decrease with increasing surface temperature if the activation energy for dissociation is lower than that for desorption. By rearrangin g the above expression for S o one may obtain the following equation, (4 2)
110 where T s is the surface temperature, R is the universal gas constant, d and r represent the pre exponential factors, and E d and E r represent the activat ion energies for desorption and dissociation, respectively. Thus, given values of and S o as a function of temperature, an Arrhenius plot of should be linear if propane dissociation on PdO(101) follows the trapping mediated kinetic model. Furthermore, if the model describes the data accurately, evaluation of the Arrhenius construction will provide estimates of the apparent pre factor and activation energy ( E r E d ) for alkane dissociative chemisorption. We c onducted propane adsorption experiments at temperatures from 250 to 300 K to determine if the trapping mediated model accurately describes the kinetics of propane dissociation on the PdO(101) thin film. We performed these experiments above 250 K since this 1 peak temperature observed in propane TPD spec tra (Figure 4 1 and Figure 4 2 A ). In this case, the molecular propane desorption rate should be sufficiently high to maintain very low coverages of molecular propane during ad sorption. This condition must be satisfied to justify using equation (4 1) to calculate the initial dissociation probability S o We chose 300 K as the highest temperature for the experiments in order to minimize desorption of water during propane adsorptio n. Recall that the H 2 O TPD peak at 345 K results from the oxidation of adsorbed H atoms, generated by propane dissociation, by reaction with the PdO surface. By preventing H 2 O desorption, we avoid removing O atoms from the PdO surface during propane uptake and hence prevent potential changes in the surface reactivity that may accompany surface reduction.
111 We exposed the PdO surface to various doses of propane at fixed surface temperature, and then performed TPRS experiments to estimate the yield of CO 2 an d hence dissociated C 3 H 8 for each surface temperature and propane exposure considered. Given estimates of the coverage of propyl groups resulting from propane dissociation, we applied the propyl mole balance equation to estimate the initial propane dissoc iation probability at different temperatures. We generated low propyl coverages in these experiments since the dissociation probability is approximately independent of the propyl coverage at low coverage, and may thus be set equal to the initial dissociati on probability. In this case, integration of the mole balance equation yields a linear relationship between the coverage of propyl groups and the propane exposure as given by the equation To determine S o as a function of surface temperature, we prepared plots of the propyl coverage [ R ] versus the propane exposure, = t for each surface temperature considered where t is the time of the gas exposure. A linear fit of each such uptake plot then gives the initial dissociation probab ility S o at each temperature at which the uptake data was obtained. This analysis requires estimates of the propane exposure as well as the propyl coverage. As discussed above, we estimated the propyl coverages from the CO 2 desorption yields by assuming th at all of the propyl groups react to produce CO 2 Also, we estimate a propane incident flux of 8.72 x 10 4 ML/s from the initial portion of an uptake curve obtained for molecular propane adsorption on Pd(111) at 85 K, and assuming that the initial trapping probability is unity. We utilized the same propane beam conditions in all of our experiments within the limits of experimental uncertainty.
112 Figure 4 5 shows representative plots of propyl coverages versus propane exposure obtained at several surface temperatures. C onsistent with the expected limiting behavior, each initial uptake curve is well approximated by a linear function so the slopes of these curves may be equated with the corresponding initial dissociation probability. We note that the exposure scale for eac h curve in Figure 4 5 is shifted by a constant amount (~ 0.013 ML), which causes the intercepts to pass more closely through the origin (within 0.001 ML). We also obtain a non zero intercept in the linear fit to the uptake curve obtained for molecular pr opane adsorption on Pd(111). This shift originates from a background propane exposure that the crystal experiences after cessation of the propane beam flow into the chamber. Apparently, slow desorption from cold surfaces causes the propane partial pressure in the UHV chamber to decay slowly after the propane flow from the beam is terminated. From measured traces of the propane partial pressure versus time, we find that the background propane exposure, occurring after termination of the beam flow, varies by no more than 10% among the exposures conducted. This observation is consistent with our finding that a constant shift causes all of the uptake curves shown in Figure 4 5 to pass closely through the origin. Since the background exposure is approximately the same among the measurements, it should only introduce small errors in our estimates of the initial dissociation probability. Figure 4 6 shows the initial dissociation probability as a function of the surface temperature determined from the propyl uptake curves. The analysis predicts that the initial dissociation probability decreases smoothly with increasing surface temperature, with S o ranging from 0.491 at 250 K to 0.209 at 300 K. The decrease in S o with
113 increasing surface temperature indicates that the apparent activation energy for propane dissociation on PdO(101) is negative. As shown in Figure 4 7, a plot of versus reciprocal temperature is accurately represented by a linear function ( R 2 = 0.98), indicating that propane dissoc iation on PdO(101) is well described by the trapping mediated model represented in equations (4 1) and (4 2). From the linear fit shown in Figure 4 7, we estimate that the apparent kinetic parameters for propane dissociation on PdO(101) are given by E d E r = 16.2 kJ/mol and d / r = 2588, respectively. The large value of the desorption pre factor d relative to the dissociation pre factor r implies a more constrained transition structure for dissociation than desorption, which is expected. The d / r ratios reported previously for ethane and propane dissociation on Ir(110) are also greater than unity (~400 vs. 200, respectively)  though smaller than that estimated for propane dissociation on PdO(101). The compositi onal and structural heterogeneity of PdO(101) may impose stricter geometric requirements for propane dissociation on the oxide compared with dissociation on pure metal surfaces. Comparison w ith Other Investigations Recent computational studies of the disso ciative chemisorption of methane on PdO surfaces may provide insights for understanding the facile dissociation of propane on PdO(101) observed in the present study. Li et al.  employed density functional theory (DFT) calculations to study CH 4 dissociation on PdO(100), PdO(001) and PdO(110) surfaces. These researchers predict that the dissociation pathway on each of the PdO surfaces involves a methane molecule bound initially to a Pd atom followed by H atom transfer to a nearby O atom of the surface, producing CH 3 Pd and O H moieties.
114 Similar to our findings with propane, these authors predict a negative, apparent dissociati on barrier of E r E d = 12.5 kJ/mol for the favored dissociation pathway on PdO(001). However, methane dissociation on the PdO(100) and PdO(110) surfaces involves larger energy barriers. In a more recent study, Blanco Rey and Jenkins  also predict that methane dissociation on PdO(100) involves methane bound on a Pd at om and H transfer to a surface O atom. Interestingly, the dissociation pathways predicted in these DFT studies are consistent with suggestions made earlier by Fujimoto et al.  based on rate data obtained from supported Pd catalysts. Specifically, these workers asserted that methane dissociation on PdO surfaces occurs at O vacancy/O atom pairs, and involves methane bindi ng on the O vacancy (i.e., Pd site) and H atom transfer to the neighboring O site. The PdO(101) surface appears to have an ideal structure for C H bond activation of alkane molecules. As seen in Figure 4 1, the stoichiometric PdO(101) surface consists of neighboring rows of cus Pd atoms and cus O atoms, with these atoms representing half of the total number of surface atoms. Considering the DFT studies discussed above, the favored pathway for alkane activation on PdO(101) likely involves an alkane molecule binding to a cus Pd atom, followed by H atom transfer to an adjacent cus O atom, resulting in adsorbed R Pd and O H groups. In this case, the negative, apparent energy barrier for propane dissociation would indicate that C H bond activation is facile on a cus Pd atom/cus O atom pair of the PdO(101) surface. The large concentration of cus site pairs, together with their intrinsic reactivity, may therefore be responsible for the exceptionally high activity of the PdO(101) surface toward propane activation.
115 The high concentration of cus Pd/cus O site pairs appears to be a unique characteristic of the PdO(101) facet, and imparts this surface with distinct reactive properties. Most of the low index facets of PdO possess only one type of surface atom, and hence Pd/O pairs would be present only in small concentrations on the majority of PdO surfaces under oxygen rich reaction conditions. The stoichiometric PdO(100) surface is an exception, and is predicted by DFT to be the most stable facet of bulk PdO  ; the same study reports that PdO(101) is the second most stable facet. The stoichiometric PdO(100) surface does possess cus O atoms (three fold coordinated), but all of the Pd atoms are coordinatively saturated and thus relatively inert toward adsorbed molecules. Accordingly, Blanco Rey and Jenkins  have recently predicted that CH 3 and CH 4 species bind more strongly on the cus O sites of stoichiometric PdO(100) than on the Pd atoms. Such a site preference seems unlikely for alkane molecules on PdO(101), given the availability of cus Pd atoms on this surface and the potential for dative bonding. However, he possibility that propoxy species form during or after C 3 H 8 activation on PdO(101) should be further i nvestigated experimentally. DFT investigations would be quite useful for identifying the preferred sites for alkyl and alkane binding on PdO(101), and for clarifying the nature of the strong alkane PdO(101) interaction observed for propane. Importantly, wh ile the (100) facet of bulk PdO may be slightly more stable than the (101) facet  the oxidation of Pd(111) preferentially produc es PdO(101)  As such, PdO(101) should be prevalent on the surfaces of supported Pd catalysts under realistic reaction conditions. Thus, compared with other facets of PdO, the reactivity of PdO(101) should be more represe ntative of the reactivity of supported Pd catalysts in practical applications of alkane oxidation.
116 Summary We investigated the adsorption and reactions of propane on a PdO(101) thin film using temperature programmed reaction spectroscopy, and find that pro pane dissociation by C H bond cleavage is highly facile on this surface and leads to the complete oxidation of the adsorbed propane. At 85 K, propane saturation of the PdO(101) surface produces two distinct states that desorb at 120 and 190 K during TPD. B ased on a comparison with propane TPD spectra obtained from Pd(111), we attribute the 190 K peak to propane molecules that interact strongly with cus Pd and/or cus O atoms of the surface, possibly through an agostic interaction. The 120 K peak is consisten t with propane molecules physically adsorbed on the surface in a first layer state. The TPRS data shows that a fraction of the molecularly adsorbed propane dissociates and then undergoes complete oxidation above about 400 K. In TPRS spectra, we observe the desorption of H 2 O at 345 K and the simultaneous desorption of H 2 O and CO 2 at 463 K. The water peak at 345 K arises from reaction between the PdO surface and H atoms produced by propane dissociation, while the peaks at 463 K originate from the reaction lim ited production of H 2 O and CO 2 by oxidation of adsorbed hydrocarbon fragments. The relative yield of reaction to desorption limited H 2 O is approximately seven, which is strong evidence that molecularly adsorbed propane dissociates by single C H bond cleava ge. Since no other reaction products could be detected, we conclude that the dissociated propane undergoes complete oxidation with high selectivity on PdO(101) for the conditions studied. We also conclude that the initial C H bond cleavage occurs below 200 K since propane desorbs in peaks at 120 and 190 K during TPD. As far as we know, this study is the first to report facile C H bond cleavage of an alkane on a crystalline oxide surface under UHV conditions.
117 We also present evidence that the dissociative c hemisorption of propane on PdO(101) occurs by a precursor mediated mechanism wherein a molecularly adsorbed state of propane acts as the precursor to initial C H bond cleavage. From CO 2 desorption yields measured as a function of the propane exposure and t he surface temperature, we estimate that the initial dissociation probability of propane on PdO(101) decreases from 49% to 21% with increasing surface temperature from 250 to 300 K. We find that the temperature dependent dissociation probabilities are accu rately described by a simple precursor mediated kinetic model, and estimate that the apparent activation energy for propane dissociation is E r E d = 16.2 kJ/mol. Finally, we find that the CO 2 yield correlates with the amount of propane desorbing at 190 K but is invariant with increases in the amount of propane desorbing at 120 K. This observation suggests that the more strongly bound molecular state serves as the precursor to the initial activation of propane on PdO(101). The precursor mediated pathway f or C H bond cleavage of propane on PdO(101) is analogous to a general mechanism for C H bond activation of alkanes by mononuclear TM complexes in which alkane sigma complexes serve as a key intermediates. This similarity could have implications for clarify ing the electronic and geometric factors that facilitate alkane activation on TM oxide surfaces.
118 Figure 4 1. Schematic representation of PdO(101) thin film structure. A) Top view. B) Side view. The orange and blue atoms represent O and Pd atoms respec tively. Rows of threefold coordinated (cus) Pd and O atoms are indicated. The a and b directions correspond to the  and [ 101] crystallographic directions of PdO. Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society
119 Figure 4 2. TPD spectra of masses 2, 18, 28, 29 and 44 amu obtained from a PdO(101) thin film after exposing the surface to a saturation dose of propane at a substrate tem perature of 85 K. The TPD spectra were obtained using a constant heating rate of 1 K/s. Propane gives rise to the desorption features below 200 K in the 28, 29 and 44 amu traces, while CO 2 is responsible for the peak at 463 K in the 44 amu trace. Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society
120 Figure 4 3. Prop ane TPD spectra obtained as a function of init ial propane coverage at 85 K. A) TPD from PdO(101) surface. B ) TPD from Pd(111) surface. The TPD spectra were obtained using a constant heating rate of 1 K/s. Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society A B
121 Figure 4 4. Desorption and reaction yields as a function of propane exposure. The yields were estimated from integrated TPRS spectra that were collected after adsorbing propane on t he PdO(101) film at 85 K. The figure shows the yields of propane that oxidizes, propane that desorbs in the 1 TPD peak at 190 K and the total amount of adsorbed propane, which we define as the total desorption yield ( 1 + 2 ) plus the yield of propane tha t oxidizes Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society
122 Figure 4 5. Propyl group coverage as a function of the propane exposure for adsorption at surface temperatures of 258, 266, 274 and 300 K. The propyl coverages were calculated from the CO 2 yields determined from TPRS spectra, where we assumed that all of the adsorbed propyl groups react to produce CO 2 during heating. The dash ed lines are linear fits to each isothermal yield exposure curve. Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society
123 Figure 4 6 Initi al dissociation probability of propane on PdO(101) as a function of the surface temperature from 250 to 300 K. The dissociation probabilities were determined from linear fits of the propyl yield versus propane exposure curves at each temperature shown. The dashed curve is a plot of equation (4 1) using the kinetic parameters determined from analysis of the data (see text). We estimate an error of 0.02 for the calculated S o values. Reprinted with permission from J.F. Weaver, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society
124 Figure 4 7. Arrhenius construction derived from the precursor mediated kinetic model discussed in the text and using values of S o determined from the experimen tal data. The dashed line is a linear fit of the experimental data. The linearity ( R 2 = 0.98) demonstrates that the precursor mediated model accurately describes the measured dependence of S o on surface temperature. Reprinted with permission from J.F. Weav er, S.P. Devarajan and C. Hakanoglu, Journal of Physical Chemistry C 113 (2009) 9773. Copyright 2009, American Chemical Society
125 CHAPTER 5 CONCLUSIONS The study of platinum group metals under oxidizing conditions reveals important information which could h elp in designing improved catalytic systems that are widely used in various industrially relevant applications. Pt and Pd oxidation is studied in UHV using strong oxidizing agents like NO 2 and atomic oxygen beams to produce high coverage oxygen phases on s ingle crystal model catalytic surfaces. The oxidic phases that are produced in UHV conditions are analyzed using surface sensitive techniques such as STM, AES, LEED, XPS and TPD/TPRS, to reveal information regarding structure and reaction properties. The f irst part of my doctoral study is the setting up of a UHV STM apparatus for the study of oxidation reactions on TM surfaces. An STM study on the oxidation of Pt(111) revealed useful information on the growth of various oxygen phases on the surface with in creasing oxygen concentration using strong oxidizing agents like NO 2 and atomic oxygen. The STM study of Pt(111) oxidation reveals that beyond the well known 0.25 ML coverage, the oxygen atoms continue to occupy fcc hollow sites forming three rotationally degenerate (2 1) domains, conforming to the three symmetry directions of the Pt(111) substrate. This finding puts an end to a long standing debate regarding the occupancy of adsorption sites in the 0.25 0.50 ML range. Beyond 0.50 ML coverage, a surface oxi de phase with Pt O coordination, similar to that of bulk PtO is observed. The compound starts to grow as particles within the (2 1) chemisorbed phase even below the 0.50 ML coverage and agglomerates to form branched Y shaped structures with Pt oxide chains about 19 24 long. These structures interconnect to form a honeycomb super structure at the saturation coverage of 0.75 ML.
126 The observation of Pt oxide chain formation on Pt(111) along with previous results on low coordination sites of Pt(110) and Pt(33 2) suggest a common trend in Pt oxidation [28, 29] Recent results from studies on oxidation of Pt under near ambient conditions  and also electrochemical oxidation  reveal that oxidation of Pt(111) proceeds via a PtO like surface oxide. Thus three entirel y different scenarios pinpoint to the same oxidation mechanism for Pt(111), which is the most abundant close packed surface on the Pt nanoparticles typically employed in industrial catalysis. Similar results for Pt(111) oxidation have also been predicted u sing DFT calculations  It is also conjectured that oxide nucleation occurs at corner and edge sites which can spread to the low index facets to explain decrease in oxygen reduction reaction activity per surface Pt atom in nanoparticles as compared to single crystals for electrodes in polymer electrolyte membrane fuel cells  Thus, it would be really interesting to study the onset of bulk like oxide formation on Pt(111) using STM to find out the role played by step edges and defects. Future work on platinum oxidation can also include scanning tunneling spectroscopy to probe the various phases grown on the surface. Since oxide formation is an inhibitor for many reactions which use Pt catalysts, it should be worthwhile to study the inhibition of oxide growth on Pt(111) and Pt nanoparticles by incorporating other metal clusters such as Au  or study the oxida tion of bimetallics such as Pt Pd  The second part of my study deals with the reactivity of a thin film PdO(101) surface grown on Pd(111) surface using an atomic oxygen beam, towards propane oxidation. The TPRS study revealed that C H b ond breaking of the alkane is facile on the thin film oxide at temperatures below 200 K, in UHV, via the formation of a strongly
127 bound molecular species. The strongly bound molecular species acts as a precursor to C H bond breaking and its formation is att ributed to the availability of coordinatively unsaturated Pd atoms of the surface. Further studies by the Weaver Group at University of Florida, with the additional use of DFT calculations, have revealed that methane, ethane, propane and n butane bind on PdO(101) by forming dative bonds with coordinatively unsaturated (cus) Pd atoms [61, 62] The activation of the C H bond is attributed to the weakening of the bonds due to a strongly bound sigma complex formation. The interesting trend emerging out of these studies of alkane activation is that dissociation barriers of the molecularly bound species are strongly dependent on the local coordinative environment on the metallic as well as oxidic surfaces. This is an important finding in understanding the catalytic activity of Pd under oxidizing conditions. Recently, the UHV STM system has been utilized to characterize and study the structural changes that accompany the isothermal decomposition of the PdO(101) thin film grown on Pd(111) which is used for t he alkane oxidation studies  Future studies could include STM investigation of alkane activation on the PdO(101) surface by producing the alkane adlayer at low temperature and imaging the surface while holding it at various temperatures to get snapshots of the surface before and after reaction and also after the oxidation.
128 LIST OF REFERENCES  E. Lundgren, A. Mikkelsen, J.N. Andersen, G. Kresse, M. Schmid, P. Varga, J. Phys.: Condens. Matter, 18 (2006) R481.  N. Seriani, F. Mitte ndorfer, J. Phys.: Condens. Matter, 20 (2008) 184023.  R.E.R. Colen, J. Christoph, F. Pea, H.H. Rotermund, Surf. Sci., 408 (1998) 310.  J. Dicke, H.H. Rotermund, J. Lauterbach, Surf. Sci., 454 456 (2000) 352.  M.S. Chen, Y. Cai, Z. Yan, K.K. Gath S. Axnanda, D.W. Goodman, Surf. Sci., 601 (2007) 5326.  N.R. Avery, Chem. Phys. Lett., 96 (1983) 371.  C.T. Campbell, G. Ertl, H. Kuipers, J. Segner, Surf. Sci., 107 (1981) 220.  P.D. Nolan, B.R. Lutz, P.L. Tanaka, J.E. Davis, C.B. Mullins, J. C hem. Phys., 111 (1999) 3696.  H. Steininger, S. Lehwald, H. Ibach, Surf. Sci., 123 (1982) 1.  S. Ovesson, B.I. Lundqvist, W.F. Schneider, A. Bogicevic, Phys. Rev. B, 71 (2005) 115406.  N. Takahashi, H. Shinjoh, T. Iijima, T. Suzuki, K. Yamazaki, K. Yokota, H. Suzuki, N. Miyoshi, S. i. Matsumoto, T. Tanizawa, T. Tanaka, S. s. Tateishi, K. Kasahara, Catal. Today, 27 (1996) 63.  R. Burch, J.A. Sullivan, T.C. Watling, Catal. Today, 42 (1998) 13.  H. Hamada, Y. Kintaichi, M. Inaba, M. Tabata, T. Yoshinari, H. Tsuchida, Catal. Today, 29 (1996) 53.  H. Ohtsuka, Appl. Catal. B, 33 (2001) 325.  C. Yokoyama, M. Misono, J. Catal., 150 (1994) 9.  S.S. Mulla, N. Chen, L. Cumaranatunge, G.E. Blau, D.Y. Zemlyanov, W.N. Delgass, W.S. Epling, F .H. Ribeiro, J. Catal., 241 (2006) 389.  J.L. Gland, B.A. Sexton, G.B. Fisher, Surf. Sci., 95 (1980) 587.  D.H. Parker, M.E. Bartram, B.E. Koel, Surf. Sci., 217 (1989) 489.  J.F. Weaver, J. J. Chen, A.L. Gerrard, Surf. Sci., 592 (2005) 83.  J. Wintterlin, R. Schuster, G. Ertl, Phys. Rev. Lett., 77 (1996) 123.
129  J. Wintterlin, S. Vlkening, T.V.W. Janssens, T. Zambelli, G. Ertl, Science, 278 (1997) 1931.  J. Segner, W. Vielhaber, G. Ertl, Isr. J. Chem., 22 (1982) 365.  N.A. Saliba, Y.L. Tsai, C. Panja, B.E. Koel, Surf. Sci., 419 (1999) 79.  C.R. Parkinson, M. Walker, C.F. McConville, Surf. Sci., 545 (2003) 19.  A. Eichler, F. Mittendorfer, J. Hafner, Phys. Rev. B, 62 (2000) 4744.  P. Lgar, Surf. Sci., 580 (2005) 137. [ 27] D.I. Jerdev, J. Kim, M. Batzill, B.E. Koel, Surf. Sci., 498 (2002) L91.  S. Helveg, H.T. Lorensen, S. Horch, E. Lgsgaard, I. Stensgaard, K.W. Jacobsen, J.K. Nrskov, F. Besenbacher, Surf. Sci., 430 (1999) L533.  J.G. Wang, W.X. Li, M. Borg, J. Gustafson, A. Mikkelsen, T.M. Pedersen, E. Lundgren, J. Weissenrieder, J. Klikovits, M. Schmid, B. Hammer, J.N. Andersen, Phys. Rev. Lett., 95 (2005) 256102.  D. Friebel, D.J. Miller, C.P. O'Grady, T. Anniyev, J. Bargar, U. Bergmann, H. Ogasawara, K.T Wikfeldt, L.G.M. Pettersson, A. Nilsson, Phys. Chem. Chem. Phys., 13 (2011) 262.  D.J. Miller, H. berg, S. Kaya, H. Sanchez Casalongue, D. Friebel, T. Anniyev, H. Ogasawara, H. Bluhm, L.G.M. Pettersson, A. Nilsson, Phys. Rev. Lett., 107 (2011) 19550 2.  R.B. Anderson, K.C. Stein, J.J. Feenan, L.J.E. Hofer, Ind. Eng. Chem., 53 (1961) 809.  R.J. Farrauto, M.C. Hobson, T. Kennelly, E.M. Waterman, Appl. Catal. A, 81 (1992) 227.  R.J. Farrauto, J.K. Lampert, M.C. Hobson, E.M. Waterman, Appl. Ca tal. B, 6 (1995) 263.  R. Burch, F.J. Urbano, Appl. Catal. A, 124 (1995) 121.  R. Burch, F.J. Urbano, P.K. Loader, Appl. Catal. A, 123 (1995) 173.  J.G. McCarty, Catal. Today, 26 (1995) 283.  J.N. Carstens, S.C. Su, A.T. Bell, J. Catal., 17 6 (1998) 136.  A.K. Datye, J. Bravo, T.R. Nelson, P. Atanasova, M. Lyubovsky, L. Pfefferle, Appl. Catal. A, 198 (2000) 179.
130  P. Salomonsson, S. Johansson, B. Kasemo, Catal. Lett., 33 (1995) 1.  R.S. Monteiro, D. Zemlyanov, J.M. Storey, F.H. Rib eiro, J. Catal., 199 (2001) 291.  H. Gabasch, K. Hayek, B. Kltzer, W. Unterberger, E. Kleimenov, D. Teschner, S. Zafeiratos, M. Hvecker, A. Knop Gericke, R. Schlgl, B. Aszalos Kiss, D. Zemlyanov, J. Phys. Chem. C, 111 (2007) 7957.  C.F. Cullis, B.M. Willatt, J. Catal., 83 (1983) 267.  V.A. Bondzie, P.H. Kleban, D.J. Dwyer, Surf. Sci., 465 (2000) 266.  H. Gabasch, A. Knop Gericke, R. Schlgl, M. Borasio, C. Weilach, G. Rupprechter, S. Penner, B. Jenewein, K. Hayek, B. Kltzer, Phys. Chem. Chem. Phys., 9 (2007) 533.  K. Nakao, S. I. Ito, K. Tomishige, K. Kunimori, Catal. Today, 111 (2006) 316.  E.E. Tornau, V. Petrauskas, G. Zvejnieks, Catal. Today, 116 (2006) 62.  C.J. Zhang, P. Hu, J. Am. Chem. Soc., 123 (2001) 1166.  G. Zh eng, E.I. Altman, J. Phys. Chem. B, 106 (2002) 1048.  G. Zvejnieks, V.N. Kuzovkov, V. Petrauskas, E.E. Tornau, Appl. Surf. Sci., 252 (2006) 5395.  I. Nakai, H. Kondoh, T. Shimada, A. Resta, J. Andersen, T. Ohta, J. Chem. Phys., 124 (2006) 224712. [ 52] J. Rogal, K. Reuter, M. Scheffler, Phys. Rev. B, 69 (2004) 075421.  H.H. Kan, J.F. Weaver, Surf. Sci., 602 (2008) L53.  J.F. Weaver, A.F. Carlsson, R.J. Madix, Surf. Sci. Rep., 50 (2003) 107.  A.V. Hamza, H.P. Steinruck, R.J. Madix, J. Chem Phys., 86 (1987) 6506.  C.B. Mullins, W.H. Weinberg, J. Chem. Phys., 92 (1990) 4508.  P.D. Szuromi, J.R. Engstrom, W.H. Weinberg, J. Chem. Phys., 80 (1984) 508.  P.D. Szuromi, J.R. Engstrom, W.H. Weinberg, J. Phys. Chem., 89 (1985) 2497.  J.F. Weaver, M.A. Krzyzowski, R.J. Madix, Surf. Sci., 393 (1997) 150.  J.F. Weaver, S.P. Devarajan, C. Hakanoglu, J. Phys. Chem. C, 113 (2009) 9773.
131  J.F. Weaver, J.A. Hinojosa Jr, C. Hakanoglu, A. Antony, J.M. Hawkins, A. Asthagiri, Catal. Today, 160 (2011) 213.  J.F. Weaver, C. Hakanoglu, J.M. Hawkins, A. Asthagiri, J. Chem. Phys., 132 (2010) 024709.  A.L. Gerrard, Interactions of atomic oxygen with Pt(111) and nitrided Si(100), Ph.D. Thesis, in: Chemical Engineering, University of Flori da, 2005.  A.L. Gerrard, J. J. Chen, J.F. Weaver, J. Phys. Chem. B, 109 (2005) 8017.  L.E. Davies et al., Handbook of Auger Electron Spectroscopy, 2 ed., Physical Electronics Division, Perkin Elmer, Eden Prairie, MN, 1976.  G. Ertl, J. Kppers, Low Energy Electrons and Surface Chemistry, VCH Veragsgessellschaft, Weinheim, Federal Republic of Germany, 1985.  C.D. Wagner, G.E. Muilenberg, Handbook of x ray photoelectron spectroscopy: a reference book of standard data for use in x ray photoelec tron spectroscopy, Physical Electronics Division, Perkin Elmer, Eden Prairie, MN, 1979.  D.A. King, Surf. Sci., 47 (1975) 384.  R.I. Masel, Principles of Adsorption and Reaction on Solid Surfaces, Wiley Interscience, New York, 1996.  P.A. Redhe ad, Vacuum, 12 (1962) 203.  H. J. G ntherodt, R. Wiesendanger (Eds.), Scanning Tunneling Microscopy I: General Principles and Applications to Clean and Adsorbate Covered Surfaces, Springer Verlag, Berlin Heidelberg, 1992.  R. Young, J. Ward, F. Sci re, Phys. Rev. Lett., 27 (1971) 922.  J. Bardeen, Phys. Rev. Lett., 6 (1961) 57.  J. Tersoff, D.R. Hamann, Phys. Rev. B, 31 (1985) 805.  C. Bai, Scanning Tunneling Microscopy and Its Applications, 2nd revised ed., Berlin: Springer Verlag, Shang ai: Scientific and Technical Publishers 1999.  F. Besenbacher, Rep. Prog. Phys., 59 (1996) 1737.  J. Stroscio, W.J. Kaiser (Eds.), Scanning Tunneling Microscopy, Academic Press, San Diego, CA, 1993.  G. Binnig, H. Rohrer, Helv. Phys. Acta, 55 ( 1982) 726.  G. Binnig, H. Rohrer, Surf. Sci., 126 (1983) 236.
132  G. Binnig, H. Rohrer, C. Gerber, E. Weibel, Appl. Phys. Lett., 40 (1982) 178.  G. Binnig, H. Rohrer, C. Gerber, E. Weibel, Phys. Rev. Lett., 49 (1982) 57.  G. Binnig, H. Rohrer, F. Salvan, C. Gerber, A. Baro, Surf. Sci., 157 (1985) L373.  G. Binnig, H. Rohrer, C. Gerber, E. Stoll, Surf. Sci., 144 (1984) 321.  R.S. Becker, J.A. Golovchenko, B.S. Swartzentruber, Phys. Rev. Lett., 54 (1985) 2678.  R.M. Feenstra, A.P. Fei n, Phys. Rev. B, 32 (1985) 1394.  R.M. Tromp, R.J. Hamers, J.E. Demuth, Phys. Rev. Lett., 55 (1985) 1303.  H. Bubert, H. Jenett (Eds.), Surface and Thin Film Analysis: Principles, Instrumentation, Applications, Wiley VCH Verlag, Weinheim, Federal R epublic of Germany, 2002.  C. Gerber, G. Binnig, H. Fuchs, O. Marti, H. Rohrer, Rev. Sci. Instrum., 57 (1986) 221.  W.W. Crew, Monitoring Surface Reactions Using Scanning Tunneling Microscopy, Ph.D. Thesis, in: Chemical Engineering, Stanford Unive rsity, 1995.  W.W. Crew, R.J. Madix, Rev. Sci. Instrum., 66 (1995) 4552.  T.A. Land, T. Michely, R.J. Behm, J.C. Hemminger, G. Comsa, J. Chem. Phys., 97 (1992) 6774.  M. Okano, K. Kajimura, S. Wakiyama, F. Sakai, W. Mizutani, M. Ono, J. Vac. Sc i. Technol. A, 5 (1987) 3313.  J. Frohn, J.F. Wolf, K. Besocke, M. Teske, Rev. Sci. Instrum., 60 (1989) 1200.  P.J. Feibelman, Phys. Rev. B, 56 (1997) 10532.  J.M. Hawkins, J.F. Weaver, A. Asthagiri, Phys. Rev. B, 79 (2009) 125434.  R.B. Ge tman, Y. Xu, W.F. Schneider, J. Phys. Chem. C, 112 (2008) 9559.  W.W. Crew, R.J. Madix, Rev. Sci. Instrum., 66 (1995) 4552.  R.J. Madix, W.W. Crew, X.C. Guo, Appl. Surf. Sci., 113 114 (1997) 539.  J.F. Weaver, H.H. Kan, R.B. Shumbera, J. Phys.: Condens. Matter, 20 (2008) 184015.  M.E. Bartram, R.G. Windham, B.E. Koel, Surf. Sci., 184 (1987) 57.
133  D. Dahlgren, J.C. Hemminger, Surf. Sci., 123 (1982) L739.  P. Sautet, Surf. Sci., 374 (1997) 406.  I.S. Tilinin, M.K. Rose, J.C. Dun phy, M. Salmeron, M.A. Van Hove, Surf. Sci., 418 (1998) 511.  M.M. Biener, J. Biener, C.M. Friend, Surf. Sci., 601 (2007) 1659.  B. Min, A. Alemozafar, M. Biener, J. Biener, C. Friend, Top. Catal., 36 (2005) 77.  B.K. Min, X. Deng, D. Pinnad uwage, R. Schalek, C.M. Friend, Phys. Rev. B, 72 (2005) 121410.  W. Gao, T.A. Baker, L. Zhou, D.S. Pinnaduwage, E. Kaxiras, C.M. Friend, J. Am. Chem. Soc., 130 (2008) 3560.  T.M. Pedersen, W. Xue Li, B. Hammer, Phys. Chem. Chem. Phys., 8 (2006) 1 566.  N. Seriani, W. Pompe, L.C. Ciacchi, J. Phys. Chem. B, 110 (2006) 14860.  E. Lundgren, et al., J. Phys.: Condens. Matter, 18 (2006) R481.  C. Hall, R.N. Perutz, Chem. Rev., 96 (1996) 3125.  A.E. Shilov, G.B. Shul'pin, Chem. Rev., 9 7 (1997) 2879.  J.A. Labinger, J.E. Bercaw, Nature, 417 (2002) 507.  H.H. Kan, R.B. Shumbera, J.F. Weaver, J. Chem. Phys., 126 (2007) 134704.  H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci., 602 (2008) 1337.  H.H. Kan, R.B. Shumbera, J.F. Weaver, J. Chem. Phys., 126 (2007) 134704.  H. Gabasch, W. Unterberger, K. Hayek, B. Kltzer, G. Kresse, C. Klein, M. Schmid, P. Varga, Surf. Sci., 600 (2006) 205.  A.F. Carlsson, R.J. Madix, J. Phys. Chem. B, 104 (2000) 12237.  C. L. K ao, R.J. Madix, J. Phys. Chem. B, 106 (2002) 8248.  Y. Yun, J. Wang, E.I. Altman, J. Catal., 253 (2008) 295.  D. Stacchiola, L. Burkholder, W.T. Tysoe, Surf. Sci., 542 (2003) 129.  H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495.
134  L.S. Tait, Z. Dohnlek, T.C. Campbell, D.B. Kay, J. Chem. Phys., 122 (2005) 164708.  L.S. Tait, Z. Dohnlek, T.C. Campbell, D.B. Kay, J. Chem. Phys., 125 (2006) 234308.  E. Lundgren, G. Kresse, C. Klein, M. Borg, J .N. Andersen, M. De Santis, Y. Gauthier, C. Konvicka, M. Schmid, P. Varga, Phys. Rev. Lett., 88 (2002) 246103.  J. Klikovits, E. Napetschnig, M. Schmid, N. Seriani, O. Dubay, G. Kresse, P. Varga, Phys. Rev. B, 76 (2007) 045405.  S.A. Soulen, R.J. Madix, Surf. Sci., 323 (1995) 1.  H. Y. Li, Y. L. Guo, Y. Guo, G. Z. Lu, P. Hu, J. Chem. Phys., 128 (2008) 051101.  M. Blanco Rey, S.J. Jenkins, J. Chem. Phys., 130 (2009) 014705.  K. i. Fujimoto, F.H. Ribeiro, M. Avalos Borja, E. Iglesia, J. Catal., 179 (1998) 431.  J. Zhang, K. Sasaki, E. Sutter, R.R. Adzic, Science, 315 (2007) 220.  B. Lim, M. Jiang, P.H.C. Camargo, E.C. Cho, J. Tao, X. Lu, Y. Zhu, Y. Xia, Science, 324 (2009) 1302.  J.A. Hinojosa Jr, J.F. Weaver, Surf. Sci. 605 (2011) 1797.
135 BIOGRAPHICAL SKETCH Sunil Poondi D evarajan was born in 1979, in Chennai, India to Shri. P.C. Devarajan and Smt. S udha Devarajan. He is the middle son amongst three brothers and grew up in Chennai and attended Padma Seshadri Bala Bhav an Senior Secondary School, Nungambakkam, affiliated to the Central Board of Secondary Education of India and graduated in 1997. He attended the prestigious Indian Institute of Technology Madras, located in Chennai, India and graduated in 2001 with a Ba ch elor of Technology degree in chemical e ngineering. He joined the graduate program at the Department of Chemical Engineering University of Florida, Gainesville, in 2001. He received a Maste r of Science degree in 2004 and continued to pursue a doctoral degr ee under the guidance of Dr. Jason Weaver. He set up an Ultra High Vacuum system at the Weaver laboratory and specialized in conducting surface science experiments in Ultra High Vacuum using modern surface analytical techniques, to study growth, properties and reactivity of oxygen phases on transition metal model catalytic surfaces. Sunil received his Ph.D. i n Chemical Engineering in 2012 from the College of Engineering, University of Florida.