<%BANNER%>

The Development and Reactivity of Various Oxide Phases Grown on Platinum (100) and Palladium (111)

Permanent Link: http://ufdc.ufl.edu/UFE0024348/00001

Material Information

Title: The Development and Reactivity of Various Oxide Phases Grown on Platinum (100) and Palladium (111)
Physical Description: 1 online resource (181 p.)
Language: english
Creator: Kan, Heywood
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: 100, 111, atoms, catalysis, chemisorption, deuterium, dpm, film, h2o, heterogeneous, hot, iss, leed, nonthermal, oxidation, oxygen, palladium, pd, platinum, pt, surface, thin, tpd, uhv, ultrahigh, vacuum, water, xps
Chemical Engineering -- Dissertations, Academic -- UF
Genre: Chemical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Platinum group metals are excellent catalysts for the oxidation of hydrocarbons and CO under oxygen-rich conditions. However, the fundamental mechanisms governing oxide growth and reactivity still remain unclear due to difficulties associated with oxidizing these metals under the ultrahigh vacuum conditions necessary for most high precision surface analytical techniques. Over the past few years, my colleagues and I have elucidated the mechanisms for the growth and reactivity of platinum and palladium oxide by utilizing gas-phase oxygen atoms beams to simulate high pressure oxidizing environments while still operating under clean ultrahigh vacuum conditions. This document contains the research results of the following five group efforts I have led during my doctoral studies at the University of Florida: 1) Hot precursor reactions during the collisions of gas-phase oxygen atoms with deuterium chemisorbed on Pt(100), 2) Adsorption and abstraction of oxygen atoms on Pd(111): Characterization of the precursor to PdO formation, 3) An ordered PdO(101) thin film grown on Pd(111) in ultrahigh vacuum, 4) Adsorption of water on a PdO(101) thin film: Evidence of an adsorbed HO-H2O complex, and finally 5) Pd(111) oxidation kinetics: Temperature and oxygen pressure effects on thin film growth, thickness, and morphology. Key findings from each of these works include 1) confirmation of non-thermal hot precursor reactions between gas-phase oxygen radicals with adsorbed deuterium on platinum, 2) the development of a novel method for characterizing oxide phase reactivity during oxidation, 3) the discovery of the first method for producing a well-ordered palladium oxide single crystalline surface in ultrahigh vacuum, 4) the first ever study of water on a well-ordered palladium oxide thin film, and 5) the discovery that oxide thin film growth kinetics, and final film thickness and morphology all depend on both oxygen pressure and substrate temperature. Collectively, these studies demonstrate the utility of gas-phase oxygen atoms beams generated in ultrahigh vacuum in characterizing the development and reactivity of oxide phases grown on platinum and palladium.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Heywood Kan.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Weaver, Jason F.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-05-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024348:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024348/00001

Material Information

Title: The Development and Reactivity of Various Oxide Phases Grown on Platinum (100) and Palladium (111)
Physical Description: 1 online resource (181 p.)
Language: english
Creator: Kan, Heywood
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: 100, 111, atoms, catalysis, chemisorption, deuterium, dpm, film, h2o, heterogeneous, hot, iss, leed, nonthermal, oxidation, oxygen, palladium, pd, platinum, pt, surface, thin, tpd, uhv, ultrahigh, vacuum, water, xps
Chemical Engineering -- Dissertations, Academic -- UF
Genre: Chemical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Platinum group metals are excellent catalysts for the oxidation of hydrocarbons and CO under oxygen-rich conditions. However, the fundamental mechanisms governing oxide growth and reactivity still remain unclear due to difficulties associated with oxidizing these metals under the ultrahigh vacuum conditions necessary for most high precision surface analytical techniques. Over the past few years, my colleagues and I have elucidated the mechanisms for the growth and reactivity of platinum and palladium oxide by utilizing gas-phase oxygen atoms beams to simulate high pressure oxidizing environments while still operating under clean ultrahigh vacuum conditions. This document contains the research results of the following five group efforts I have led during my doctoral studies at the University of Florida: 1) Hot precursor reactions during the collisions of gas-phase oxygen atoms with deuterium chemisorbed on Pt(100), 2) Adsorption and abstraction of oxygen atoms on Pd(111): Characterization of the precursor to PdO formation, 3) An ordered PdO(101) thin film grown on Pd(111) in ultrahigh vacuum, 4) Adsorption of water on a PdO(101) thin film: Evidence of an adsorbed HO-H2O complex, and finally 5) Pd(111) oxidation kinetics: Temperature and oxygen pressure effects on thin film growth, thickness, and morphology. Key findings from each of these works include 1) confirmation of non-thermal hot precursor reactions between gas-phase oxygen radicals with adsorbed deuterium on platinum, 2) the development of a novel method for characterizing oxide phase reactivity during oxidation, 3) the discovery of the first method for producing a well-ordered palladium oxide single crystalline surface in ultrahigh vacuum, 4) the first ever study of water on a well-ordered palladium oxide thin film, and 5) the discovery that oxide thin film growth kinetics, and final film thickness and morphology all depend on both oxygen pressure and substrate temperature. Collectively, these studies demonstrate the utility of gas-phase oxygen atoms beams generated in ultrahigh vacuum in characterizing the development and reactivity of oxide phases grown on platinum and palladium.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Heywood Kan.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Weaver, Jason F.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-05-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024348:00001


This item has the following downloads:


Full Text

PAGE 1

THE DEVELOPMENT AND REACTIVITY OF VARIOUS OXIDE PHASES GROWN ON PLATINUM (100) AND PALLADIUM (111) By HEYWOOD HON-WAI KAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Heywood Hon-Wai Kan 2

PAGE 3

To Cynthia 3

PAGE 4

ACKNOWLEDGMENTS Thanks go out to my graduate advisor, Professor Jason Weaver, for his guidance throughout the years both inside and outside the lab; Professor Aravind Asthagiri for his many insightful comments and assistance with the DFT calculations presented he rein; Jose Hinojosa, Sunil Devarajan, Brad Shumbera, and Robert Colmyer for their helpful discussions and experimental assistance; and the United States Department of Energy for funding. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................8LIST OF FIGURES .........................................................................................................................9ABSTRACT ...................................................................................................................... .............12 CHAPTER 1 INTRODUCTION ................................................................................................................ ..14Background and Motivation ...................................................................................................14Experimental Techniques Used ..............................................................................................202 HOT PRECURSOR REACTIONS DURIN G THE COLLISIONS OF GAS-PHASE OXYGEN ATOMS WITH DEUTERIUM CH EMISORBED ON PLATINUM (100) .........23Introduction .................................................................................................................. ...........23Experimental Methods .......................................................................................................... ..26Results .....................................................................................................................................28Thermal Desorption of D2O and D2 ................................................................................28Reactions of Gaseous O Atoms with D-Saturated Pt(100) at 90 K .................................30Post-Reaction TPD Spectra .............................................................................................32Effective Cross Sections for D2 and D2O Production .....................................................36Variation of D2 and D2O Desorption Rates with Initial D Coverage ..............................38Influence of Surface Temperature on the Initial D Coverage Dependence .....................39Discussion .................................................................................................................... ...........42Summary ....................................................................................................................... ..........453 ADSORPTION AND ABSTRACTION OF OXYGEN ATOMS ON PALLADIUM (111): CHARACTERIZATION OF THE PRECURSOR TO PA LLADIUM OXIDE FORMATION ..................................................................................................................... ....54Introduction .................................................................................................................. ...........54Experimental Methods .......................................................................................................... ..56Results and Discussion ........................................................................................................ ...58Oxygen Coverage Calibration and Dissolution into the Bulk .........................................58Thermal Desorption of O2 ...............................................................................................59LEISS Measurements ......................................................................................................61Direct Abstraction of 18O/Pd(111) by Gaseous 16O Atoms .............................................63Kinetic Model of 18O Abstraction and Exchange ............................................................66Summary ....................................................................................................................... ..........72 5

PAGE 6

4 A PALLADIUM OXIDE (101) THIN FI LM GROWN ON PALLADIUM (111) IN ULTRAHIGH VACUUM ......................................................................................................80Introduction .................................................................................................................. ...........80Experimental Methods .......................................................................................................... ..81Results and Discussion ........................................................................................................ ...81Summary ....................................................................................................................... ..........865 ADSORPTION OF WATER ON A PALLA DIUM OXIDE (101) THIN FILM: EVIDENCE OF AN ADSORBED HO-H2O COMPLEX .....................................................91Introduction .................................................................................................................. ...........91Experimental Methods .......................................................................................................... ..93Computational Methods ......................................................................................................... .95Experimental Results and Discussion .....................................................................................97Structure of the PdO(101) Thin Film on Pd(111) ...........................................................97Water Desorption from Pd(111) and PdO(101) ..............................................................97TPD of Water on PdO(101) as a Function of Coverage ................................................100Uptake of Water into Different Adsorbed States ..........................................................103Analysis of Water TPD Spectra ....................................................................................104Computational Results and Discussion ................................................................................109Adsorbate Pair Formation ..............................................................................................111High Coverage Configurations: Trimers and Saturation ...............................................114Summary ....................................................................................................................... ........1186 PALLADIUM (111) OXI DATION KINETICS: TEMPE RATURE AND OXYGEN PRESSURE EFFECTS ON THIN FILM GROWTH, THICKNESS, AND MORPHOLOGY ..................................................................................................................131Introduction .................................................................................................................. .........131Experimental Methods .......................................................................................................... 134Results and Discussion ........................................................................................................ .136Precursor TPD Feature ..................................................................................................136Gas-Phase Oxygen Atom Uptake Curves ......................................................................137Low Energy Electron Diffraction ..................................................................................139X-ray Photoelectron Spectroscopy ................................................................................140Oxygen Uptake Temperature Dependence ....................................................................141Annealing the Surface Oxide .........................................................................................143The Precursor State: A Second Look ............................................................................144Bulk Oxide Growth .......................................................................................................146The Kelvin Equation and Particle Growth Model .........................................................148Flux Effects on TPD State Population ...........................................................................150Simplified Oxide Growth Model ...................................................................................151Simulation vs. Experiment ............................................................................................153Controlling Oxide Thin Film Thickness ........................................................................154Summary ....................................................................................................................... ........155Appendix: Precursor to the Oxide Precursor State ...............................................................156 6

PAGE 7

7 CONCLUSIONS ................................................................................................................. .170 LIST OF REFERENCES .............................................................................................................171BIOGRAPHICAL SKETCH .......................................................................................................181 7

PAGE 8

LIST OF TABLES Table page 5-1 Adsorption energies of H2O and OH species on PdO(101) predicted with DFT. ...........130 8

PAGE 9

LIST OF FIGURES Figure page 2-1 TPD spectra of D2O and D2 from Pt(100) .........................................................................472-2 Normalized direct product monitoring spectra of D2O and D2 ..........................................482-3 Yields of D2, D2O and O2 that evolve in TPD after exposing deuterium-saturated Pt(100) held at 90 K to an oxygen atom beam. ..................................................................492-4 TPD spectra of D2O obtained as a function of th e atomic oxygen exposure to the deuterium-saturated Pt(100) surface held at 90 K. ............................................................502-5 Direct product monitoring spectra of D2O and D2 measured for different initial D atom coverages on Pt(100).................................................................................................512-6 Direct product monitoring spectra of D2O obtained for initial deuterium coverages of 0.44 ML, 0.65 ML and 1.15 ML and surface temperatures of 90 K and 150 K. ...............522-7 Desorption yields of D2O and D2 obtained during surface exposure to the oxygen atom beam shown as a function of the initial D atom coverage for surface temperatures of 90 K and 150 K.. ......................................................................................533-1 O2 TPD spectra (heating rate = 1 K/s) obtained after expos ing Pd(111) held at 500 K to an oxygen atom beam. ...................................................................................................753-2 LEISS spectra obtained after adsorption of oxygen on Pd(111). .......................................763-3 Schematic of the key stages/oxygen phases in the proposed mechan ism for the initial formation of PdO on Pd(111) ............................................................................................773-4 Reactive scattering meas urements of isotopic oxyge n abstraction by gas-phase oxygen atoms on Pd(111). .................................................................................................783-5 Schematic of the elementary steps consider ed in the kinetic model for abstraction of 18O atoms adsorbed on the 2D oxide by 16O atoms incident from the gas-phase. .............793-6 Arrhenius plot of the maximum slope obt ained in the direct rate measurements during the period over which the abstraction rate incr eases as oxygen atoms populate the precursor state. .............................................................................................................794-1 XPS Pd 3d5/2 spectra obtained from clean Pd( 111) and a 3.3 ML PdO layer grown on Pd(111) at 500 K using an oxygen atom beam. .................................................................884-2 LEED images obtained from clean Pd (111) and a 3.3 ML PdO layer grown on Pd(111) at 500 K, and then annealed to 675 K... ...............................................................884-3 LEED simulation results and identification of PdO(101). .................................................89 9

PAGE 10

4-4 Model representations of the stoichiometric PdO(101) surface. .......................................905-1 Top and side view of the PdO(101) thin film structure. ..................................................1205-2 H2O TPD spectra obtained from Pd(111) and PdO(101) after generating initial water coverages of 1.2 and 1.3 ML, respectively, at a substrate temperature of 85 K. .............1215-3 H2O TPD spectra obtained from PdO(101) at initial water coverages ............................1225-4 Difference H2O TPD spectra obtained by subtrac ting a TPD spectrum obtained after saturating the 1 peak with 0.175 ML of water from a TPD spectrum. ...........................1235-5 Uptake of H2O into various adsorbed states plotte d as coverage within a state versus the total H2O coverage .....................................................................................................1245-6 TPD inversion analysis of 1. ...........................................................................................1255-7 TPD inversion analysis of 2. ...........................................................................................1265-8 Top and side view of the most stable DFT identified configurations ..............................1275-9 Top view of (a) H2O-H2O (82 KJ/mol) (b) OH-OH (66.7 KJ/mol) (c) HO-H2O (96.1 KJ/mol) (d) H2O-HO (70.3 KJ/mol) pairs at 0.50 MLcus coverage. ...............................1285-10 Energy pathway for the formation of adsorbed trimers on PdO(101) .............................1296-1 TPD spectra obtained after exposure of Pd(111) held at 100 K to a gas-phase oxygen atom beam with an incident flux of 0.0132 ML s-1 and a temperature ramp rate of 1 K s-1. ................................................................................................................................1576-2 Total oxygen uptake curves plotted as a f unction of total oxygen atom dose in units of ML for the different fluxes and substrate temperatures ..............................................1586-3 LEED images obtained with a beam en ergy of 63 eV for various oxygen coverages generated at 500 K or 650 K. ...........................................................................................1596-4 XPS Pd 3d5/2 spectra collected from Pd(111) fo r various oxygen coverages generated at 650 K or 500 K. ............................................................................................................ 1606-5 Temperature Effects on Oxygen Uptake. .........................................................................1616-6 LEED images, using a beam energy of 63 eV, obtained from a 0.75 ML coverage generated at 500 K before and after annealing to 675 K. ................................................1626-7 TPD spectra obtained after exposing the Pd5O4 surface oxide to gas-phase oxygen atoms at a substrate temperature of 100 K. ......................................................................1636-8 Extrinsic oxygen concentra tion obtained by taking the O/Pd ISS peak ratio, plotted as a function of total oxygen concentration.. ...................................................................164 10

PAGE 11

6-9 The Kelvin equation plotted as a functi on of particle radius, outlining the three effective pressure ranges ..................................................................................................1656-10 TPD spectra obtained after oxidizing Pd(111) at a subs trate temperature of 650 K with an incident flux of 0.002 ML/s and 0.0174 ML/s. ...................................................1666-11 Comparison of results generated from si mulations with experimental data for low temperature and high temperat ure oxidation kinetics. .....................................................1676-12 Schematic representation of particle grow th illustrating the effects of pressure and substrate temperature on final oxide film thickness ........................................................1686-13 Total oxygen uptake plotted as a functi on of total dose of gas-phase oxygen atoms with an incident flux of 0.0283 ML/s, conducted at 500 K and 100 K. ..........................169 11

PAGE 12

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE DEVELOPMENT AND REACTIVITY OF VARIOUS OXIDE PHASES GROWN ON PLATINUM (100) AND PALLADIUM (111) By, Heywood Hon-Wai Kan May 2009 Chair: Jason F. Weaver Major: Chemical Engineering Platinum group metals are excellent catalys ts for the oxidation of hydrocarbons and CO under oxygen-rich conditions. However, the fund amental mechanisms governing oxide growth and reactivity still remain unclear due to difficulties associated with oxidizing these metals under the ultrahigh vacuum conditions necessary for mo st high precision surface an alytical techniques. Over the past few years, my colleagues and I have elucidated the mechanisms for the growth and reactivity of platinum and pa lladium oxide by utilizing ga s-phase oxygen atoms beams to simulate high pressure oxidizing environmen ts while still operati ng under clean ultrahigh vacuum conditions. This document contains the research results of the following five group efforts I have led during my docto ral studies at the University of Florida: 1) Hot precursor reactions during the collisions of gas-phase oxygen atoms w ith deuterium chemisorbed on Pt(100), 2) Adsorption and abst raction of oxygen atoms on Pd( 111): Characteri zation of the precursor to PdO formation, 3) An ordered PdO( 101) thin film grown on Pd(111) in ultrahigh vacuum, 4) Adsorption of water on a PdO(101) th in film: Evidence of an adsorbed HO-H2O complex, and finally 5) Pd(111) oxidation kinetics: Temperature and oxygen pressure effects on thin film growth, thickness, and morphology. Key findings from each of these works include: 1) confirmation of non-thermal hot precursor reacti ons between gas-phase oxygen radicals with 12

PAGE 13

13 adsorbed deuterium on platinum, 2) the developm ent of a novel method for characterizing oxide phase reactivity during oxidation, 3) the discove ry of the first method for producing a wellordered palladium oxide single crys talline surface in ul trahigh vacuum, 4) th e first ever study of water on a well-ordered palladium oxide thin film, and 5) the discovery that oxide thin film growth kinetics, and final film thickness and morphology all depend on both oxygen pressure and substrate temperature. Colle ctively, these studies demonstrate the utility of gas-phase oxygen atoms beams generated in ultrahigh vacuum in characterizing the develo pment and reactivity of oxide phases grown on platinum and palladium.

PAGE 14

CHAPTER 1 INTRODUCTION Background and Motivation Palladium is an excellent heterogeneous oxi dation catalyst for oxygen rich applications such as the catalytic oxidation of methane [1-11] or car bon monoxide [12-19]. However, due to the high cost of palladium, ther e are substantial financial in centives to develop oxidation catalysts with enhanced activity and resiliency while also utilizing cheaper and more abundant materials. Unfortunately, with current trends leaning more towards alloys and composites and away from traditional oxide s upported platinum group metals, hi storically employed trial-anderror based methods have become insufficient to tackle the mounting complexity of catalyst development as the problem compounds with ever y additional elemental ingredient considered. Indeed, the sheer volume of possible compositional permutations to be tested makes the daunting challenge of developing more cost effective catalysts very cost ineffective, and therefore an increasing number of researchers are attempting a first principles approach towards development of novel catalysts from the ground up through design. A first principles approach to catalyst de sign, however, requires a first principles understanding of the functionality of model catalysts, and theref ore fundamental surface science studies of the mechanisms of palladium catalyzed oxidation will enable the development of next generation heterogeneous oxidation catalysts. Id eally these fundamental surface science studies would take place under conditions id entical to those found in indus trial applications. However, this is not possible because of a variety of e xperimental limitations. For example, palladiums highly valued activity towards car bon monoxide oxidation results in highly troublesome levels of experimental uncertainty even in the presence of the trace amounts of carbon monoxide found under ambient conditions. Furthermore, most accu rate surface analytical techniques utilize 14

PAGE 15

electron or ion beams which requir e long mean free paths to be m easured or utilized, and their negligible mean free paths at pressures ex ceeding molecular flow conditions make most experiments utilizing ion or electron beams impossible. This pressure gap between industrial and e xperimental conditions forces researchers to extrapolate results obtained unde r ultrahigh vacuum through a pr essure differential of over twelve orders of magnitude in hopes of predic ting catalyst behavior under realistic conditions. The pressure gap is currently a contentious poi nt amongst those in the catalysis scientific community since certain aspects of catalytic activity may not be effectively captured ex situ (i.e. in vacuum) [20] based on the following three points: 1) thermodynamic and kinetic barriers may restrict the formation of certain industrially relevant phases th at do not develop under ultrahigh vacuum conditions, 2) reaction selectivity can be a strong function of pr essure and therefore reaction pathways which may be insignificant under ultr ahigh vacuum may in fact be dominant under realistic pressures, and 3) heterogeneously catalyzed reactions represent non-equilibrium systems and low pressure studies which focus entire ly on static and stable structures effectively miss the dynamical processes taking place on strained catalyst surfaces undergoing spatiotemporal organization under industrially re levant conditions. In other words, a full first principles understanding of a catalytic surface re quires detailed knowledge of not only the stable or metastable structures, but also detailed knowledge of the transition states which are difficult to study ex situ These points have driven some in the surface science community to abandon ultrahigh vacuum studies altogether in favor of ambient pressure in situ studies, which are limited to expensive high intensity electron or optical spec troscopic methods. Indeed, a couple of recent works have made substantial progress in elucid ating the mechanisms for palladium 111 oxidation 15

PAGE 16

by utilizing ambient pressure x-ray photoele ctron spectroscopy based on high intensity synchrotron radiation [21-23]. Howe ver, these high pressure studies are still over two orders of magnitude below industrial conditions. Furthermore, even these higher chemical potentials are still not sufficient to generate measurable rates of oxidation under industrially relevant temperatures, therefore forcing researchers to increase substrate temperatures possibly beyond industrially relevant temperatures until measurable rates of oxidation can be detected. They are in essence exchanging the cont entious pressure gap for an apparently n on-contentious temperature gap when in fact, both pressure an d temperature gaps are (o r should be) matters of contention. Since rates of reaction of interest to hetero geneous catalysis scale with a power law of pressure and exponentially with temperature, sele ctivity is typically a much stronger function of temperature than pressure. And if selectivity is a str onger function of temperature than pressure, then experiments conducted under realistic temperat ures are also (if not more) important for an accurate understanding of the behavior of catalyst under realistic conditions. In this dissertation, I will demonstrate that it may be possible to conduct both realistic temperature and pressure studies while still maintaining the clean ultrahig h vacuum conditions necessary to enable the use of the most precise analytical equipment currently available to surface science through the use of gas-phase oxygen atom beams. Let us now di scuss how and why this approach works. Previous researchers who have developed ex situ methods for studying oxidation utilized either high pressure molecular oxygen cells or ba ckfill doses with stronger oxidants such as NO2, however both methods introduce unacceptabl e levels of background carbon monoxide upon admission into the vacuum chamber. Indeed, a prominent challenge specifically hampering oxidation studies involve the pr oduction of highly reactive clean -off gases which comes as a 16

PAGE 17

result of the admitted oxidant not only oxidizing the sample of interest, but also all surfaces exposed to the oxidant. Since most vacuum chambers are made of stainless steel, oxidation of the chamber walls generates gaseous carbon monoxide which can severely compromise the accuracy of the experiments even in trace qua ntities. Indeed, a major hurdle limiting ex situ experiments involves the so-called clean-off reaction that inevitably follows an oxidizing exposure which can potentially react away an oxide phase prio r to examination. In fact, a recent study of palladium oxidation by Han et al. [24,25] who conducted ex situ experiments while utilizing high pressure oxygen cells failed to detect the forma tion of any oxide despite evidence of the contrary obtained from in situ experiments [22,23], which was possibl y a result of the carbon monoxide clean-off reaction which reduced the surface prio r to analysis. Therefore, before accurate ex situ oxidation studies can be conducted, we must fi rst develop a method for producing oxides under ultrahigh vacuum while also limiting exposure of the chamber walls to the oxidant, and the best way to do this is through the use of molecular beams utilizing aggressive oxidants, and no other gas-phase oxidant has a higher chemical potential than gas-phase oxygen atoms. The use of gas-phase oxygen atom beams eff ectively addresses the first pressure gap concern by simultaneously increasing chemical poten tial while also reducing kinetic barriers for oxidation. The begin with, the recent development of commercially availa ble ultrahigh vacuum compatible radio frequency inductively coupled non-thermal oxygen plasma sources has allowed us to assemble one of the first gas-phase oxygen atom beams in ultrahigh vacuum which enables accurate ex situ studies of oxidation. Since the rate limiti ng step for oxidation is the dissociation of the molecular oxygen bond as is shown in Ch apter 6, utilizing gas-phase oxygen atom beams essentially moves the rate limiti ng step from the surface to the pl asma chamber where the rate of dissociation can literally be dialed in. Furthermore, gas-phase oxygen atoms are expected to 17

PAGE 18

access the same reaction channels experienced by molecular oxygen during oxidation of the substrate, therefore effectively addressing the fi rst pillar of the pressure gap problem by both lowering the kinetic activation ba rriers to zero while simultaneously raising the chemical potential of the oxygen atoms. Removing the activ ation barrier for adsorption enables oxidation studies at virtually any su bstrate temperature and has allowed us to be the first to ever study just the effects of substrate temperature on oxide growth kinetics and pha se development without complications stemming from temperature de pendencies on molecular oxygen dissociative adsorption rates. The second pressure gap concern involving the inability accurate ly sample reaction pathways of industrial relevance under ultrahi gh vacuum conditions can also be effectively addressed through the use of ga s-phase oxygen atom beams. If we assume that the sticking probability of molecular oxygen is purely a func tion of oxygen coverage or exposed substrate phase, then the effective pressure established by the gas-phase oxygen atom beam is expected to scale linearly with gas-phase oxygen atom beam flux, enabling control of effective oxygen pressure through control of the incident gas-phase oxygen atom flux. Therefore, control of both effective pressure and substrate temperature a llows us to study mech anisms under realistic effective pressures and temperatures allowing us to study industrially relevant reaction selectivities. Finally, the third pressure gap concern can be addressed through the use of gas-phase oxygen atom beams since we have discovered a method of isotherm ally studying the reactivity of various phases during oxidation through the us e of reactive scattering measurements in conjunction with isotopic labeling as detailed in Chapter 3, enabling in situ experiments to be conducted accurately ex situ Furthermore, while the structur e of transition states cannot be 18

PAGE 19

entirely determined from reac tive scattering measurements, th e use of gas-phase oxygen atom beams do not prohibit the use of other nondestructive in situ methods such as electron or photon spectroscopies (although we were not equipped to conduct these e xperiments ourselves). Indeed, through the use of gas-phase oxygen atom beams, it is possible to both oxidize the substrate at low absolute pressures and temperatures, effect ively bridging the pressure gap while also enabling the use of virtually all surface analytical techniques. Aside from the pressure gap, however, there is also the materials gap. Since surface scientists can only extract elementary rate para meters from elementary surfaces, fundamental surface science studies must take place on well-d efined single miller index planes forcing researchers to extrapolate the behavior of macroscopic polycrystalline surfaces from the collective behavior of single miller index planes [20]. Considering how there are an infinite number of ways one can slice a crystal, it would technically take an infinite number of experiments to effectively capture the behavior of every possibly exposed surface of a material under industrial conditions. However, the fractiona l representation of the various miller index surfaces represented in polycrystalline catalyst powders is determined by the relative thermodynamic stability of each miller index plan e, and therefore a close examination of the properties of the close packed planes, such as the Pd(111) surface, provides substantial insight into the overall activity of the powder. This, of course, assumes that the catalytically active surface is the metal, which is not the case for pa lladium [10,26]. The studies of single crystalline oxide surfaces of palladium oxide have been limited primarily due to difficulties with producing single crystalline oxide surfaces in ultrahigh vacuum until now. This dissertation contains details on our discovery of a method to producing sing le crystalline palladium oxide surfaces in 19

PAGE 20

ultrahigh vacuum, which represen ts the first step towards an understanding of the catalytic activity of palladium oxide under industrial conditions. This dissertation contains five studies I have led during my time as a doctoral student at the University of Florida department of chemical engineering. Chapter 2 details the first study conducted with the newly installed high flux gasphase oxygen atom beam, and demonstrates its potential for in situ reactive scattering measurements of the thermal and nonthermal oxidation of deuterium chemisorbed on Pt(100). Chapter 3 details results of reactive s cattering measurements of the abstraction of isotopically labeled oxygen from the Pd(111) surface with gas-phase oxygen atoms, and represents the first ever in situ characterization of the reactivity of oxide phases during oxidation conducted in ultrah igh vacuum. Chapter 4 details our discovery of a method of producing the first ever single crystalline palladi um oxide surface in ultrahigh vacuum. Chapter 5 then details the results of our study of the inte ractions of water with a single crystalline palladium oxide surface, which drastically affect s the catalytic activity of palladium oxide. Finally, Chapter 6 details a comprehensive study of the growth and kinetics of Pd(111) oxidation, and the temperature and oxygen pressure effects on oxide film growth, thickness, and morphology. Experimental Techniques Used Temperature programmed reaction spectrosc opy (TPRS) or temperature programmed desorption (TPD) measures the rate of reaction and desorption of an adsorbate (or decomposition of an oxide) during a linear temperature ramp. The desorption products are monitored with a quadrupole mass spectrometer (QMS), which first i onizes incident gas molecules then utilizes oscillating electrostatic fields to sort the molecules by mass pr ior to detecti on. The rate of desorption is then plotted as a function of subs trate temperature, which can be analyzed to 20

PAGE 21

determine kinetic rate parameters such as activation energy and fre quency factors through a variety of methods [27]. Furthermore, because th e QMS measures a desorption rate, it is possible to determine the total initial coverage of an adsorbate by integrating the TPD spectrum and comparing the integrated are to the integrated area obtained after deso rption of a known amount of adsorbate (i.e. a saturation coverage). All cove rages presented in this dissertation have been determined using this method. While TPD allows us to measure the total adsorbate concentrati on of the surface, ion scattering spectroscopy (ISS) or low energy ion scattering spectroscopy (LEISS) probes only the extrinsic concentration of an adsorbate, which is particularly helpful for oxide phase identification. With ISS, ions of a set kinetic en ergy are scattered from substrate, with the final kinetic energies measured with a hemispherical analyzer. The composition of the substrate can then be determined from the distribution of fina l ion kinetic energies si nce the efficiency of energy transfer from the incident ion to the subs trate atom increases as the masses of the ion and atom approach each other based off of Newtonian mechanics. For a singl e adsorbate system one would expect to find a bimodal distribution of ion energies, with ions that have lost a lot of energy representing ions collidi ng with lighter absorbates, and ions with little energy loss representing ions that had collided with h eavy metal atoms. We can then get a qualitative measurement of the adsorbate (i n this study, oxygen) cont ent of the surface by taking a ratio of the two peak intensities. Since ISS uses relatively high energy ions, there is to some extent some surface sputtering occurring during any scan, enabling a form of depth profiling by plotting the peak ratio as a function of time. X-ray photoelectron spectroscopy (XPS), assesses total adso rbate concentr ation without substantially damaging the surface, and represents an excellent method for detecting evidence of 21

PAGE 22

22 oxidation. With XPS, incident x-rays excite el ectrons via the photoelectric effect from the substrate into the gas phase which are then detect ed with a hemispherical analyzer to determine their final kinetic energies. By subtracting the kinetic energy of the emitted electrons from the photon energy of the x-rays, it is possible to determine the bindi ng energy of the orbital the emitted electron originated from. From the binding energy of the originating orbital, we can then determine if the atom is chemically bonded (i.e. ox idized) or not. XPS is particularly well-suited for characterizing oxidation since it is very sensitive to subtle shifts in electronegativity, and therefore can detect if a metal at om is chemically reacting with an oxygen atom to form an oxide. Low energy electron diffraction (LEED) works by bombarding a substrate with a coherent beam of electrons, followed by monitoring the di ffraction pattern produced when the scattered electrons strike a phosphorescent screen similar to a cathode ray television screen. Therefore, LEED is a useful tool for non-destructive determination of surface or dering, and therefore, identification of oxide phases that develop during oxidation. Furthermore, it is possible to detect substrate phase transitions by monitoring the LEED pattern during temperature ramps. Finally, reactive scattering measurements, sometimes referred to as direct product monitoring (DPM) spectroscopy, is a technique to directly monitor under isothermal conditions the reaction products produced in situ during a gas dose. This techni que is especially useful for differentiating between thermal and nonthe rmal surface reactions as will be further discussed in the second chapter of this dissertation. Furthermore, Chapter 3 details a study that utilizes direct product monitori ng in conjunction with isotopic labeling to characterize the instantaneous reactivity of various oxide phases as they develop.

PAGE 23

CHAPTER 2 HOT PRECURSOR REACTIONS DURING TH E COLLISIONS OF GAS-PHASE OXYGEN ATOMS WITH DEUTERIUM CHEM ISORBED ON PLATINUM (100) Reprinted with permission from H.H. Kan, R. B. Shumbera, J.F. Weaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, Amer ican Institute of Physics. Introduction The elementary reactions between gas-phase oxygen atoms and species adsorbed on metal surfaces are fundamental to heterogeneous processes that occur in extreme environments such as flames, plasma and at high temperature. For exam ple, plasma-assisted catalysis is an emerging technology that relies on radica l-surface reactions to efficien tly destroy hazardous compounds present in industrial and automotive effluents. The potential for plasma-assisted catalysis to provide more energy-efficient ro utes for certain catalytic proc esses is indeed suggested by several recent studies that report synergistic enhancements in reaction rates when a heterogeneous catalyst is introduced in a plasma -activated gas mixture [28-40]. Radical-surface reactions are also central to plasma-assisted atomic layer deposition, which is a promising technology for generating thin films at low te mperature. In general, characterizing the mechanisms and kinetics governing radical-solid r eactions can aid in adva ncing technologies in which such interactions play an important role. The interactions between open-shell atoms and adsorbed species are also interesting from a scientific viewpoint because reactions can occur through mechanisms that are generally inaccessible to stable gaseous molecules. Gaseous radicals are likely to react with adsorbed species by non-thermal mechanisms, meaning that the reaction occurs before the radical fully thermalizes to the solid surface. Also, the adso rption and surface reactions of radicals often release large amounts of energy th at can stimulate desorption or reactions between adsorbed 23

PAGE 24

species. Prior research has shown that incident hydrogen atoms react efficiently with adsorbates by non-thermal mechanisms [41-69]. For example, the collisions of gaseous H atoms with Dcovered metal surfaces results in the evolution of both HD and D2 at surface temperatures as low as 80 K [69]. Through evaluation of product desorption rates, Kppers and coworkers [70] have shown that the HD production kinetics on most metal surf aces is well described within the context of a so-called hot atom mechanism which represents a process that is intermediate to the limiting Eley-Rideal and Langmuir-Hinshelwood mechanisms In the hot atom reaction mechanism, the incident H atom enters a mobile precursor stat e on the surface and reac ts with an adsorbed species before fully accommodating to the surface. The evolution of D2 has been similarly explained in terms of secondary hot atom reactio ns in which a hot D atom is generated by an interaction with a hot H atom a nd then reacts with adsorbed deuterium. While this prior work provides considerable insight for understand ing hydrogen atom-surface reactions, much less work has been devoted toward studying the su rface reactions of atomic radicals other than hydrogen, leaving unanswered questions about the gene rality of the mechanisms that have been proposed. In this study, we investigated chemical reactions induced by the collisions of gaseous oxygen atoms with deuterium-covered Pt(100), and focused on characterizing non-thermal pathways for water production. Reactions between co-adsorbed oxygen and hydr ogen on Pt(111) have been the focus of several surface science in vestigations [71-79] due largely to interest in understa nding the origin for the high activity of platinum as a catal yst for producing water from oxygen and hydrogen. Recent STM investigations [77,78] have been esp ecially significant in clarifying the mechanism 24

PAGE 25

for water formation on Pt(111) at low surface te mperature. These experiments provide evidence that water is generated at low temperature by an autocatalytic process involving the reactions, OH + H H2O (2-1) 2H2O + O 3OH + H (2-2) This catalytic cycle is initiated either by the slow reaction O + H OH or through the adsorption of small amounts of residual H2O. In support of this proposed mechanism, DFT calculations [79] predict that th e steps in the autocatalytic cycle have low energy barriers (< 25 kJ/mol) while the H + O reaction has a much larg er activation barrier of 100 kJ/mol. At higher surface temperature, the autocatalytic mechanism is less important because H2O coverages remain small, and water production occurs ma inly through successive hydrogenation of O and OH, respectively, with the generatio n of OH being rate limiting [77,78]. Several studies have also shown that non-thermal activation of one of the reactants can significantly enhance the rate of water production on metal surfaces. For example, investigators have reported facile production of H2O below 100 K when gaseous hydrogen atoms impact oxygen-covered Pt(111) [49,80]. Similar findings ha ve been reported for reactions of hydrogen atom beams with other oxygen-covered metal su rfaces, including Pd(111) [81], Cu(110) [82], Cu(100) [83], and Ni(100) [42]. In these system s, reactions mediated by hot H atoms result in enhanced rates because these intermediates can possess energies well above the chemisorption ground state and can therefore overcome activation barriers at low temperature. Prior studies have also shown that water is efficiently ge nerated during the intera ctions between gaseous oxygen atoms and adsorbed hydrogen on Ru(0001) [ 84] and Si(100) [85], indicating that nonthermal reactions occur during th e surface collisions of gaseous oxygen atoms as well. However, detailed studies of the kinetics governing oxygen atom-adsorbate reactions are scarce. Here, we 25

PAGE 26

report results of direct measurem ents of product evolution rates dur ing the collisions of gaseous oxygen atoms with D-covered Pt(100), and provide evidence that energetic hydroxyl groups play a dominant role in mediating the production of D2O during these radical-surface interactions. Experimental Methods Previous studies [86-88] provide details of the three-level UHV chamber utilized for the present experiments. The Pt(100) 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 contact with a liquid nitrogen cooled reservoi r. A type K thermocouple spot-welded to the backside of the crystal allows sample temperature measuremen ts. Resistive heating, co ntrolled using a PID controller that varies the output of a programmable DC power supply, supports maintaining or linearly ramping the sample temperature from approximately 80 K to 1250 K. Initially, sample cleaning consisted of sp uttering with 600 eV Ar+ ions at a surface temperature of 800 K, followed by annealing at 1200 K for several minut es. Subsequent cleaning involved routinely exposing the sample held at 800 K to an atomic oxygen beam for several minutes, followed by flashing the sample to 1200 K to desorb oxygen a nd carbon oxides. We considered the sample to be clean when we could no longer detect contaminants with X-ray photoelectron spectroscopy (XPS) and could obtain sharp low energy electron diffraction (LEED ) patterns consistent with the Pt(100)-hex-R0.7 surface [89]. A two-stage differentially-pumped chamber attached to the UHV chamber houses the inductively coupled RF plasma s ource (Oxford Scientific Instrument s) utilized to generate beams containing oxygen atoms for this study. This sy stem produces gaseous oxygen atoms by partially dissociating pure O2 (BOC gases 99.999%) continuously suppl ied to a small discharge chamber at the end of the plasma source. The plasma discharge chamber, fabricated from high-purity quartz to minimize atom recombination, terminat es at a 1 mm thick quartz plate with a small 26

PAGE 27

0.50 mm diameter orifice. Species exit the disc harge chamber through this orifice, forming a beam that is directed toward the sample surf ace held in the UHV chamber. In the first pumping stage, the beam passes between opposite ly-charged parallel plates ( kV cm-1) that deflect ions and electrons. After flowi ng through a conical skimmer ( = 3 mm) that separates the first and second pumping stages, the specie s travel down a collimating qua rtz tube and then enter the UHV chamber. A 1200 L/s diffusion pump (Varian VHS 4) evacuates the first pumping stage of the beam chamber while a 70 L/s turbomolecular pump and a titanium sublimation pump inserted into a liquid nitrogen cooled cryoshi eld evacuate the second pum ping stage. Finally, a mechanical shutter located in the first pumping stage enables control over beam introduction into the main UHV chamber. Threshold-ionization mass spectrometry provi des clear evidence that oxygen atoms are present in the plasma-activated beams. Additionally, a comparis on of the direct and background 16 amu and 32 amu components, measured with and without the plasma power enabled, shows that the beams employed in this study contain approximately 20% oxygen atoms in a balance of O2 and negligible quantities of impurities. When positioned approximately 50 mm from the end of the quartz tube with a 45 rotation with respect to the t ube axis, the Pt(100) sample experiences nearly uniform beam impingement across the surface during atomic oxygen exposures. The combination of this sample pos itioning and beam composition corresponds to an atomic oxygen flux with a lower bound of ~2.6 1013 cm-2 s-1 (~0.02 ML/s) at the sample surface, where 1 ML is defined as the atomic density of 1.28 1015 cm-2 of bulk-terminated Pt(100)-(1) [90]. A typical experiment consiste d of first preparing an atomic deuterium layer by exposing the clean Pt(100)-hex-R0.7 surface held at 150 K to doses of D2 (Sigma-Aldrich 99.96%) from 27

PAGE 28

a calibrated molecular beam [ 87]. Immediately following a D2 exposure, the sample was repositioned and exposed to the O atom beam while continually monitoring masses 4 and 20 amu using a quadrupole mass spectrometer (QMS) to acquire a direct produ ct monitoring (DPM) spectrum. Since the random fluxes of desorbing species are measured in this experiment, the QMS signals are directly proportion al to the rates at which the various speci es desorb during the O atom beam exposure. After the DPM experiment the sample is reposit ioned to directly face the QMS for a post abstraction TPD measurement. To estimate product desorption yields and adsorbate coverages, we scaled the areas under the DPM and post-abstraction TPD curves with integrated TPD signals obtained from pure adlaye rs of the species of interest, assuming values for the absolute coverages as reported in the liter ature for oxygen [91], water [92], and hydrogen [93]. The coverage calibrations are discussed in more detail below. Results Thermal Desorption of D2O and D2 TPD measurements were conducted with pure D2O and D layers on Pt(100) to aid in quantifying product desorption rates measured during O atom exposures to the D-covered surface, and also to confirm the thermal desorption behavior of th e various reaction products that evolve during the oxygen atom-surface collisions. Figure 2-1a shows D2O TPD spectra collected at a ramp rate of 1 K/s from D2O adsorbed on Pt(100) at 90 K. The TPD spectra agree well with previous reports of water TPD from Pt(100) [92], and also closely resemble H2O TPD obtained from Pt(111) [49], which suggests similar bonding environments for water on the two Pt surfaces. The TPD spectra ar e characteristic of D2O first populating a monolayer state, and then adsorbing into a multilayer once the monolayer reaches saturation. Desorption from the monolayer yields a TPD peak that shifts from about 162 to 169 K as the coverage increases to saturation, while desorption from the multilayer st ate produces a peak at 156 K. In prior work, 28

PAGE 29

Kizhakevariam and Stuve [92] es timate that a monolayer of H2O saturates at a coverage of 0.67 ML on the reconstructed Pt(100) surface. Following this estimate, the D2O coverages shown in Figure 2-1a were computed from the integrat ed signals by assuming th at the maximum area under the TPD peak for the monolayer state corr esponds to a coverage of 0.67 ML. It is worth noting that a monolayer of wa ter on Pt(111) also saturates at a coverage of 0.67 ML. Figure 2-1b shows D2 TPD spectra obtained after dissociatively adsorbing varying amounts of D2 on Pt(100)-hex-R0.7 held at 150 K. The D2 TPD spectra that we obtained are in excellent agreement with TPD data reported previously in a detailed study of hydrogen dissociation and adlayer growth on Pt(100) by Pasteur et al. [93] From this and other prior studies, it has been established that the dissociatio n of hydrogen on Pt(100)-hex-R0.7 partially lifts the surface reconstruction, re sulting in a heterogeneous su rface that consists of (1) domains with a local hydrogen coverage of ~ 0.63 ML and hex-R0.7 domains with much lower hydrogen content. Hydrogen also adsorbs on struct ural peculiarities resu lting from the surface reconstruction. The surface hete rogeneity produces multiple D2 desorption features which appear in two groupings, labeled as and The 1, 2, and 3 features populate initially and appear at temperatures of approximately 336, 360, and 372 K respectively, while the 1, 2, and 3 features, located at 192, 206, and 228 K respectively, intensify only after the features are nearly saturated at ~0.55 ML. The sharper 3 peak arises from D atoms adsorbed in fourfold hollow sites on (1) domains, and the 1 and 2 features have been attributed to an adsorbed state associated with a surface structure intermed iate to the hex and (1) structures. The 1 peak originates from D atoms adsorbed on hex-R0.7 domains, while the 3 feature has been attributed to a defect structure w ithin the (1) phase. The 2 peak is attributed to a kinetic isotope effect, 29

PAGE 30

and is not observed in H2 TPD spectra [93]. Notice that the 1 peak populates only above D coverages of about 0.95 ML so this state constitutes a small fracti on of the adsorbed deuterium. Prior measurements using nuclear microanalysi s indicate that the atomic D layer saturates at a coverage of 1.2 ML [90]. We therefore use this value to de termine the absolute D coverages shown in Figure 2-1b. For simplicity, we refe r to 1.15 ML of D atom s on Pt(100) as a Dsaturated surface throughout this paper since this is the highest initial D coverage that we could generate with the O atom s ource operating in the background. Reactions of Gaseous O Atoms with D-Saturated Pt(100) at 90 K Figure 2-2 shows normalized D2O and D2 desorption rate curves obtained while directing a beam of oxygen atoms onto D-satu rated Pt(100) held at 90 K. Th e zero of time in this figure corresponds to opening the beam shutter and allowing the oxygen atoms to impinge upon the sample surface. As seen in the figure, the collisions of gaseous oxygen atoms with the Dsaturated surface cause the prompt desorption of D2O and D2, followed by decay of the D2O and D2 desorption rates to constant levels after about 60 sec. The D2O rate curve decays more slowly and is broader than the D2 curve. The D2 and D2O signals drop abruptly at approximately 120 sec and return to their initial values when the shutter is interposed into the beam path to terminate the dose. Desorption of D2 and D2O is not observed when a pure O2 beam impinges on the Dsaturated surface at 90 K, indicating that the deso rption of these species at low temperature is stimulated by the incident oxygen atoms. Furthermore, the thermal desorption of D2 and D2O is negligible at 90 K (Figure 2-1). We theref ore conclude that the desorption of D2 and D2O in this experiment results from reactions involving interm ediates such as O, OD and D radicals that are not thermally accommodated to the surface, i.e., hot precursors. 30

PAGE 31

The production of D2 and D2O must occur through multiple reaction steps when gaseous O atoms impact D-covered Pt(100). A likely pathway for D2O production is through the consecutive reactions, O* + D(a) OD*, followed by OD* + D(a) D2O*, where an asterisk denotes a hot precursor. An estimate for the exot hermicity of the first reaction may be obtained by assuming that the oxygen atom does not dissipate energy to th e surface prior to reacting. In this case, the first reaction step is estimated to be exothermic by as much as 340 kJ/mol, and could hence produce highly energetic OD groups. In this calculation, we used the bond energy of a hydroxyl radical of H(O-D) = 428 kJ/mol [94], and, as estimates, the heats of adsorption of D and OD on Pt(111) of H(D-Pt) = 250 kJ/mol and H(OD-Pt) = 162 kJ/mol [49] since these enthalpies are unknown for the Pt(100) surface. Likewise, if the OD groups do not dissipate energy prior to reacting, the second reaction coul d be exothermic by as much as 288 kJ/mol ( H(DO-D) = 498 kJ/mol, H(D2O-Pt(111)) = 40 kJ/mol), and could generate energetic D2O molecules. In terms of thermochemical changes, the production of D2O is very favorable in these radical-surface interactio ns. The production of D2 could occur by a pathway involving first the generation of a hot D atom, followed by the reaction D* + D(a) D2 which is also exothermic and known to be facile on Pt(100) [69] and other metal surfaces [41,43,52,57,64,65]. Since the heat of adsorption for an O atom exceeds that for a D atom by approximately 100 kJ/mol on Pt surfaces, sufficient energy is released during O atom adsorption to generate hot D atoms and initiate D2 formation. Notice that sufficient energy is released during OD formation for these radicals to desorb directly from the surface. Unfortunately, it is di fficult to detect gaseous OD with measurements of the random flux since scattered O atoms produce H2O in the background, resulting in a large signal at 18 amu. Hydroxyl deso rption after its formation seems unlikely on physical grounds, 31

PAGE 32

however, because it would requir e at least 48% of the reaction energy to be channeled into OD translational motion perpendicular to surface. In s upport of this assertion, tr ajectory calculations predict that OH desorption proba bilities are less than 10% during O atom collisions with Hcovered tungsten since most of the reaction energy is channeled into O-H stretching rather than into the desorption coordinate [95]. We therefore conclude th at gaseous OD is produced in negligibly small quantities during these experiments. We estimate that 2% and 13% of the D atoms initially on the surface desorb as D2 and D2O, respectively, during the atomic oxygen exposure at 90 K. These yields are computed by scaling the integrated rate curves with TPD sp ectra obtained from estimated coverages of pure D and D2O layers. For this anal ysis, the zero of the D2 and D2O desorption rates are taken as the steady levels to which the rate curves decay be fore closing the beam shutter (between about 60 and 120 sec in Figure 2-2. This zero level is used because reactions of oxygen with deuterium adsorbed on the chamber walls cause the D2 and D2O partial pressures to remain higher than their values before the start of the measuremen t. Indeed, the total desorption yield measured during the beam exposure constitutes a relativ ely small fraction (~15 %) of the initial D coverage, which could suggest limited reac tivity between gaseous oxygen atoms and the adsorbed deuterium. However, TPD spectra obt ained after the oxygen atom exposures to the surface show that a large fraction of the D2O and OD species thermally accommodates to the surface after being generated via hot precursor reactions. Thus, the gaseous D2O that evolves during the O atom-surface reactions at 90 K repr esents only a fraction of the reaction products that form in these interactions. Post-Reaction TPD Spectra Figure 2-3 shows the yields of D2O, D2 and O2 as a function of the O atom exposure obtained from post-reaction TPD spectra The open symbols labeled as D2O (90 K) correspond 32

PAGE 33

to the yield of D2O that desorbs at 90 K during the O atom exposure, as determined by integrating the rate curve shown in Figure 2-2. Due to reactions between background H2O and Dcontaining species on the surface, HDO is produced during the post-reaction TPD measurements. Although the HDO yields obtained from the TPD spectra are less than one third of the D2O yields, the presence of HDO and H2O on the surface introduces uncertainty in the quantitative analysis of the post-reaction TPD spectra. The yields nevertheless provide insights for understanding trends in the evolution of reactio n products that form during the O atom-surface interactions. The data shows that the D2O and D2 TPD yields each reach limiting values of about 0.52 and 0.10 ML of D, respectively, by an oxygen atom exposure just above 0.50 ML, and remain at these levels as oxygen atoms begin to accumulate on the surface during prolonged beam exposure. Note that the D2O yield of 0.52 ML is an upperbound since reactions between adsorbed D atoms and background H2O produce small amounts of D2O prior to the TPD measurement. The D2O desorption yield obtained during the O atom exposure also increases steadily until the O atom exposure reaches about 0.50 ML, and thereafter more slowly approaches a constant value. The evolution of the O2 desorption yield appears to mirror the D2O and D2 yields. The O2 yield is negligible as the O atom exposure increases to about 0.25 ML, and then begins to increase slowly with O atom exposure. As the D2O and D2 yields approach their limiting values, the O2 desorption yield starts to increase more rapidly with exposure, reaching a final value of about 1.15 ML O atoms at the highest exposure employed. The yield behavior is consis tent with the post-reaction D2O originating mainly from OD and D2O species that form by hot precursor reacti ons and thermalize to the surface at 90 K. Firstly, the D2O yields obtained during and after the O atom exposure each reach limiting values at nearly the same O atom exposure. This simila rity is expected if D oxidation via hot precursor 33

PAGE 34

reactions is facile, and a branch ing exists between product desorp tion and thermalization at 90 K. Additionally, the O2 desorption yield is low during the early reaction period when D2O and D2 generation is dominant. This suggests that most of the incident oxygen atoms initially react to produce OD and D2O, and hence that the atomic oxygen coverage remains low up to an exposure of about 0.50 ML. Post-reaction D2O TPD spectra obtained as a function of the O atom exposure are shown in Figure 2-4. The TPD spectra exhi bit two main regions, namely, an intense feature at low temperature (~150 K to 230 K) and a broader, le ss intense region at higher temperature (~230 K to 360 K) that is indicative of reaction-limite d desorption initiated by reactions between coadsorbed O and D atoms. D2O desorbs mainly in the low temperature peak centered at about 162 K after a short 0.09 ML O atom exposure at 90 K. This feature overlaps well with the TPD spectrum obtained from a similar amount of pure D2O on Pt(100) (dashed curve), but the desorption intensity just above 162 K is slight ly higher in the postreaction data. The low temperature D2O peak maximum shifts toward higher temperature with increasing O atom exposure, and a distinct shoulder at a bout 210 K concurrently intensifies. Although prior studies of water formation on Pt(100) are scarce, numerous studies have been reported on the reactions betwee n oxygen and hydrogen on Pt(111) [71,73-80]. Of particular relevance to the current work is the finding that the water pr oduction yield is only 15% during TPD with a Pt(111) surface initially covered with atomic oxygen and hydrogen at saturation [80]. The water produced by the oxygen/hydrogen reactions during TPD partially desorbs at 175 K as a desorption-limited produc t, with the remainder desorbing at higher temperatures. Water can also react with atom ic oxygen on Pt(111) to generate hydroxyl groups below the water desorption temperature. The reverse disproportionation reaction, 2OH H2O + 34

PAGE 35

O, then occurs above the wate r desorption temperature, yieldi ng a peak at 210 K in the water TPD spectrum. Assuming similarities in the re activity of Pt(111) and Pt(100), these prior studies suggest that the production of water on Pt(100) also oc curs to a limited extent during TPD when only O and D atoms are initially co-adsorbed. It is therefore reasonable to conclude that D2O and OD form by hot precursor reactions and remain adsorbed on the surface prior to the TPD measurements, and that these species are respons ible for the low temperature features observed in the post-reaction TPD spectra. Specifically, D2O initially present on the surface would desorb mainly between 160 K and 170 K, while the OD gr oups can react with both adsorbed D atoms and other OD groups to generate D2O that desorbs near 160 K and ~210 K, respectively. Additionally, if the revers e disproportionation of OD groups does produce the D2O desorption feature near 210 K (dashed line in Figure 2-4), then the continuous increase in the intensity of this feature suggests that OD groups accumulate on the surface with increasing O atom exposure. Most likely, a significant frac tion of these OD groups are genera ted during the O atom exposure by a direct reaction between hot O atoms and adsorbed D. The post-reaction TPD spectra also exhibit broad D2O desorption featur es extending from about 230 to 360 K. These high temperature featur es are characteristic of reaction-limited D2O desorption that is initiated by the formation of hydroxyl groups from reac tions between adsorbed O and D atoms. Consistent with this interp retation, changes in the high temperature D2O features appear to track those of the D2 desorption features. For example, a distinct D2O desorption maximum is evident at about 355 K after the 0.2 5 ML O atom exposure, but this feature is significantly diminished after the 0.47 ML exposure (Figure 2-4). This change in the D2O TPD spectra coincides with a significant decline in the intensity of the 3 D2 desorption state at 372 K 35

PAGE 36

(not shown). We estimate that the reaction-lim ited features from about 230 K to 360 K constitute at most about 20% of the total D2O desorption yield, which supports the conclusion that the majority of the incident oxygen atoms react wi th deuterium during the 90 K beam exposure. Finally, the D2O TPD spectra continue to change w ith increasing O atom exposure above 0.50 ML, even though the D2O and D2 yields remain approximately constant and only the oxygen atom coverage increases at higher O atom exposur es (Figure 2-3). We attribute these changes to interactions between adsorbed D2O and Pt oxide domains that gr ow on the Pt(100) surface above an O atom exposure of about 0.70 ML [96]. Effective Cross Sections for D2 and D2O Production To estimate effective cross sections for D2 and D2O production induced by the collisions of gas-phase O atoms on the D-covered surface, we employ the following simple kinetic scheme, O(g) + D(a) OD* (2-3) O(g) + D(a) O(a) + D* (2-4) OD* + D(a) D2O(g) (2-5) D* + D(a) D2(g) (2-6) Here, the first two steps represent Eley-Ridea l (ER) reactions that generate hot OD and D radicals. The rates of these reactions are given by ][*D RODo OD and ][*D RDo D where is the O atom incident flux and oOD and D are cross-section-like quantities for OD* and D* generation, respectively. Invoking the steadystate approximation for the concentrations of OD* and D* then leads to the following expr essions for the rates at which D2O and D2 evolve from the surface as a function of time, ))(2exp(][ ][2 2t D R dt ODdD ODo oODoOD g (2-7) 36

PAGE 37

))(2exp(][ ][2 2t D R dt DdD ODo oDo D g (2-8) where is the initial D atom coverage and t is the time. In this case, applying the steady-state approximation is equivalent to assuming that th e reactions of hot OD a nd D radicals are much faster than the generation of these sp ecies, and hence that the rates of D2 and D2O production are equal to the generation rates of D* and OD*, respectively. Integrating the rate equations also gives the following expressions for the total D2O and D2 desorption yields, oD ][ o D OD OD total gD OD ][ )(2 ][2 (2-9) o D OD D total gD D ][ )(2 ][2 (2-10) Notice that the yields scale linearly with the init ial D atom coverage acco rding to this model. After passing the maxima, the D2O desorption rate curves obtained at 90 K are fit reasonably well by an exponential decay function w ith a time constant of 15.7 sec. Using an O atom flux of 0.018 ML/s, the total cross section D OD is then determined to be 13.8 2. Also, integration of the rate curves gives D2O and D2 desorption yields of 13% and 2% of the initial D coverage. The equations given above fo r the desorption yields are then evaluated to obtain the values of OD = 12 2 and D = 1.8 2 for the effective cross sections for D2O and D2 production. The cross section for D2 production agrees well with th e cross section reported in the literature for this reaction on Pt(100) [69]. However, the cr oss section for hot OD generation is higher than typical ER cr oss sections. This implies that other pathways for D2O production are operative in this system, and hence that the value of OD found in this analys is represents a sum of cross sections for multiple re actions that gene rate gaseous D2O. 37

PAGE 38

Variation of D2 and D2O Desorption Rates with Initial D Coverage Figure 2-5 shows D2 and D2O rate curves obtained at va rying initial D coverages for experiments conducted at a surface temperature of 90 K. As expected, the maximum desorption rates of D2 and D2O decrease as the initial D coverage decreases. The D2O desorption rate also exhibits a time delay prior to th e rate maximum that becomes l onger as the initial D coverage decreases (Figure 2-5a). As shown in several studies by Kppers and coworkers [42,49,69,70,82,83,97,98], such a dependence of the induc tion period on the initial reactant coverage is characteristic of a hot precursor reaction for which the intrinsic probability for hot precursor sticking exceeds that for reaction. Unde r these conditions, hot precursors preferentially fill sites rather than reacting with adsorbates and the rate of reaction increases toward a maximum because more hot precursors become avai lable to react as the concentration of empty sites decreases. The time delay prior to the rate maximum also increases with decreasing initial coverage because more empty sites must be filled before reaction begins to dominate over hot precursor sticking. It must be emphasized, however that this interpretation has been invoked mostly to describe abstraction kinetics in the reactions of gaseous H atoms with adsorbed D atoms, in which case the initially adsorbed species and the gas-phase reactant are chemically identical and therefore occupy the same types of surface binding sites. In the present case, hot O and OD radicals are likely intermediates in the formation of D2O, and these species can bind at different sites and experience different surf ace binding energies than D atoms. It is unclear whether or not D atoms are effective in blocking sites for O and/or OD sticking, and hence if a kinetic competition between sticking and reaction of hot O and OD radicals is responsible for the induction period observed in the D2O rate curves. In fact, reactions involving equilibrated OD groups appear to contribute to the overall production of D2O in these experiments, and, as 38

PAGE 39

discussed below, the onset of such reactions could also produce a time delay prior to the maximum D2O desorption rate. In contrast to the beha vior displayed by the D2O rates, the D2 rate curves at 90 K appear to start at a maximum for all initial D coverages (Fi gure 2-5b). This kinetic be havior occurs in the hot atom reaction scenario when the probability fo r sticking is less than that for reaction [70], which may be the case here. In prior work, HD and D2 rate curves measured during H atom collisions on D-covered Cu(100) and Pt(100) exhibit unusual beha vior, with the rates decreasing after the initial jump and then increasing to ward a maximum after the hydrogen coverage increases during the experiment [69,70]. This co mplex kinetics was attributed to the surface reconstruction that is induced during hydrogen adsorption. Specifi cally, Kammler et al. [70] show that the rate curves are well reproduced us ing a simple kinetic model in which the ratio of the sticking to reaction probabilit ies decreases below unity as the hydrogen coverage increases and causes the hex-R0.7 to (1 ) reconstruction to dominate th e surface. This model implies that the intrinsic sticking probability of D atoms on (1) domains is less than the probability for hot D atoms to react with adsorbed D atoms. While the D2 rate curves measured in the present study do not exhibit the oscillatory feature seen in prior studies with gaseous H atoms on Pt(100), the absence of an induction period at any initial D covera ge is consistent with hot D atoms having a lower probability for sticking on (1 ) domains of Pt(100) than for reacting to generate D2. It is also possible that clustering of D atoms into (1) domains of high local coverage causes reaction to be favored over sticking since locally a hot D atom will tend to encounter filled sites, even when the total D coverage is low. Influence of Surface Temperature on th e Initial D Coverage Dependence To probe the influence of the surface temper ature on the reactions induced by incident oxygen atoms, we collected D2 and D2O desorption rate curves while exposing the D-covered 39

PAGE 40

surface to the oxygen atom beam at surface temperat ures ranging from 90 K to 150 K. Increasing the surface temperature over this range cau ses only a slight narrowing in the D2 rate curves (not shown), whereas the D2O rate curves exhibit more pronoun ced changes. Figure 2-6 shows D2O rate curves obtained at surface temperatures of 90 K and 150 K for initial D coverages of 0.44, 0.65 and 1.15 ML. At each initial D coverage, the D2O desorption rates measured at 150 K reach their maxima almost immediately upon initiating th e beam exposure, in cont rast to the behavior observed at 90 K. The D2O desorption rates measured at 90 K initially track those measured at 150 K, but they continue to rise after the rate s measured at 150 K begin to decay from their maxima. Furthermore, after about 5 seconds into the exposure the D2O desorption rates remain higher at 90 K than 150 K until the rates decay to steady levels. The temperature sensitivity of the D2O desorption rate curves may be explained by considering that reactions involving adsorbed OD groups contribute to the overall production of D2O during the experiments. The desorption of D2O in the early reaction period is consistent with a hot precursor mechanism in which energe tic OD groups react with adsorbed D atoms to generate D2O. However, the post-reaction TPD spectra show that a fraction of the energetic OD groups also accommodate to the surface, causing the concentration of adsorbed OD groups to initially increase during the O atom beam exposure. The accumulation of OD on the surface could open up additional hot precursor pathways for D2O production. For example, reactions such as D* + OD(a) D2O* and OD* + OD(a) D2O* + O(a) are viable based on bonding energetics, and analogous reactions have been reported to occur during the in teractions of gaseous H or OH radicals with Pt(111) [49,75]. Since the concentration of adsorbed OD groups initially increases during the O atom beam exposure, hot precursor reactions involving eq uilibrated OD groups would occur at increasing 40

PAGE 41

rates later into the reaction period, producing gaseous D2O in parallel with reactions between hot OD groups and adsorbed D atoms. In this case, the total D2O production rate would be a sum of the rates of parallel hot precu rsor reactions, and may achieve a maximum value at specific coverages of both D and OD. If it takes longer to achieve the optimal D and OD coverages at lower then the time delay prior to the ra te maximum will increase with decreasing giving rise to behavior such as that displayed in the rate curves shown in Figures 2-5a and 2-6. Furthermore, hot precursor channels involving ad sorbed OD groups will be effectively inhibited at higher surface temperature due to the consum ption of OD groups by thermal reactions, and the rate of D2O production in the late reaction period s hould consequently decrease with increasing surface temperature, which is consistent with th e observed behavior. Recent studies indeed show that reactions between adsorbed OD groups and D atoms are facile below the D2O desorption temperature on Pt(111) [77,78]. Thus, the time delay observed in the D2O rate curves measured at 90 K appears to be attributab le to the onset of hot precursor reactions involving adsorbed OD groups, rather than to a kineti c competition between hot precurs or sticking and reaction wherein sticking is intrinsically more probable. oD ][oD ][Finally, the D2O and D2 desorption yields are shown in Figure 2-7 as a function of the initial D coverage for measurements conducted at 90 K and 150 K, where the yields are expressed in units of ML of D atoms. We find that the D2O and D2 desorption yields each scale approximately with for both surface temperatures, and the D2O yield measured at 90 K is about 1.7 times greater than that at 150 K for all initial Biener et al. [49] also observed that H2O desorption yields are higher at 80 K than 100 K during hydrogen atom collisions with Ocovered Pt(111), which suggests th at thermal and hot precursor pa thways compete for adsorbed OH groups in those reactions as well. We note also that the depe ndence of the desorption yields 3.1][oDoD ][ 41

PAGE 42

on deviates from that predicted by the simple reaction scheme pres ented above, therefore supporting the conclusion that multiple reaction path ways contribute to the overall production of gaseous D2O and D2 in these experiments. oD ] [Discussion The production of D2O and D2 during the collisions of gaseous oxygen atoms with Dcovered Pt(100) appears to occur through multip le, hot precursor reaction steps in which hydroxyl groups serve as key intermediates. The meas ured rate behavior is also consistent with parallel pathways for the production of D2O, which results in a large effective cross section for D2O desorption. We propose that the following constitutes the set of hot precursor reactions that are predominant during these radical-surface interactions, Hot D2O via D(a), O* + D(a) OD* (2-11) OD* + D(a) D2O* (2-12) Hot D atom generation and D2 production, O* + D(a) O(a) + D* (2-13) D* + D(a) D2(g) (2-14) Hot D2O production via OD(a), OD* OD(a) (2-15) D* + OD(a) D2O* (2-16) D* + OD(a) D2(g) + O(a) (2-17) OD* + OD(a) D2O* + O(a) (2-18) Thermalization and desorption of hot D2O, 42

PAGE 43

D2O* D2O(g,a) (2-19) The first two reaction steps illustra te the most direct route to D2O* which is expected to dominate at early times when the D coverage is high, and OD groups have not yet accumulated on the surface. Notice that reaction 2-11 implies that the oxygen atom react s as a hot atom (i.e. it is partially accommodated) rather than directly from the gas-ph ase. While this hypothesis seems reasonable, it is diffi cult to confirm unambiguously from the data. In the proposed reaction scheme, the D2 production step is shown to involve a hot D atom that is generated through a pro cess whereby a hot oxygen atom tran sfers energy to an adsorbed D atom. This reaction is plausible given that the binding energy of an O atom on Pt surfaces is higher than that of a D atom [4 9] and hence oxygen atom adsorpti on releases more energy than is needed to generate hot D atoms. Prior studies have shown convincingly that hot D atoms form during the collisions of gaseous H atoms with numerous D-covered metal surfaces, and react to produce D2 via reaction 2-14 [41,43,52,57, 64,65,69]. Given that the heat of H atom adsorption is lower than O atom adsorption [49], these prior ob servations suggest that hot D atoms should be readily produced during the ad sorption of oxygen atoms. Moreover, by only considering the bonding energetics, it is tempting to conclude th at gaseous oxygen atoms will be more efficient than gaseous hydrogen atoms in generating hot D at oms. However, the low rates and yields of D2 production observed in the present study suggest that hot D atom ge neration is actually inefficient during oxygen atom adsorption. In fact, higher D2 desorption rates and yields have been reported when D2 production is stimulated by the surface collisions of gaseous hydrogen atoms [69]. This difference may indicate that hot oxygen atoms have shorter surface lifetimes than do hot hydrogen atoms, and consequently that hot oxygen atoms seldom possess enough energy to generate a hot D atom when they encoun ter an adsorbed D atom. Another possibility is 43

PAGE 44

that collisions between hot O atoms and adsorbed D atoms predominantly result in OD production rather than ho t D atom generation. Reaction steps 2-16 to 2-18 depict hot precurs or reactions involving adsorbed OD groups that are expected to contribute more significantly to the observed rates later in the beam exposure when OD groups have accumulated on the surface. Reactions 2-16 and 2-17 show the production of D2O and D2, respectively, by a collision between a hot D atom and an adsorbed OD group. Biener et al. [49] have shown that both of these reactions o ccur during the in teractions of gaseous H atoms with O-covered Pt(111), and report that hydrogen abstraction from adsorbed hydroxyl, step 2-17, is the more efficient of the two. If reactions 2-16 and 2-17 were significant in the present system, however, one would expect the D2 production rate cu rves to exhibit a stronger dependence on the surface temperature than is observed since th e OD(a) concentration is temperature dependent. Thus, since the D2 rates and yields depend only weakly on the surface temperature, and considering that the D2 yield is very low, we conclude that hot D atom generation is a minor reaction channel during the O atom-surface interactions, and therefore that reactions 2-16 and 2-17 make only a small contribution to the D2O and D2 production rates. To explain the temperature depe ndence of the kinetics for D2O production, we conclude that reaction 2-18 makes a significant contribution to the observed D2O desorption rate at low temperature, and that this reaction is less im portant at higher temperature due to faster consumption of OD groups by thermal reactions. Treating the hot precursors as gas-phase species, reaction 2-18 is estimated to be exothermic by as much as 260 kJ/mol and is therefore highly favored energetically. Finally, reaction 2-19 depicts hot D2O molecules as either ther malizing to the surface or desorbing. Based on the measured yields (Figure 2-3), we estimate that at least 22% of the D2O 44

PAGE 45

desorbs at 90 K. This value compares reasonably well with prior work by Biener et al. [49] in which one third of the water that forms during H atom collisions with O-covered Pt(111) desorbs at 80 K. The high probability for ther malization could indicate that hot D2O molecules efficiently dissipate energy to the substrate and other adsorbates. Al ternatively, it could mean that most of the reaction energy is channeled into internal mo tion, and that coupling between internal modes and translation normal to the surface is slow compared with dissipation. Summary Direct reaction rate measurements and TPD show that gas-phase oxygen atoms efficiently oxidize deuterium atoms chemisorbed on Pt(100) at surface temperatures as low as 90 K to produce both OD and D2O. About 22% of the oxidized deuterium desorbs as D2O as the surface is exposed to an oxygen atom beam at 90 K, and small amounts of D2 also desorb. Using an approximate kinetic model, we estima te effective cross sections of 12 2 and 1.8 2, respectively, for the pr oduction of gaseous D2O and D2 at 90 K by reactions of gaseous oxygen atoms with chemisorbed D atoms on Pt(100) At higher surface temperature, the D2O evolution rate is found to decay more rapidly from its maximum value, causing the D2O desorption yield to decrease with an increase in the temperature at which the surface is held during exposure to the oxygen atom beam. These observations are consistent with the formation and reactions of hot precursors, mainly hydroxyl radicals, during oxygen-atom collisions with D-covered Pt(100). Specifically, we propose that hot D2O is produced by a mechanism wherei n energetic OD groups are first generated by the reaction, O* + D(a) OD*, followed by the parallel reactions, OD* + D(a) D2O* and OD* + OD(a) D2O* + O(a). The latter reaction c ontributes appreciably to D2O production only in the late reaction period wh en adsorbed OD groups have accumulated on the 45

PAGE 46

surface. Since thermal and hot precursor reactions compete for the same adsorbed reactants, we conclude that the measured D2O desorption yields and rates decrease with increasing surface temperature up to 150 K because the revers e disproportionation r eaction involving an OD* species becomes less important due to more ra pid depletion of the OD(a) concentration by thermal pathways. Overall, a key finding from this study is that gaseous oxygen atoms readily stimulate D2O formation on D-covered Pt(100) by gene rating energetic radicals that serve as reaction intermediates. The non-thermal reactions i nvestigated in this study are likely to be important in plasma-assisted applications; how ever, because they compete with thermally activated reactions, the hot precursor reactions sh ould contribute most si gnificantly in processes conducted at low temperature and high oxygen atom incident flux. 46

PAGE 47

100150200250300350400 (b)123123D2 TPD Ts = 150 K [D]o (ML)D2 desorption rate (a.u.)temperature (K)1.15 0.95 0.65 0.44 0.69 0.18(a) D2O TPD Ts = 90 KD2O desorption rate (a.u.)[D2O]o (ML) 1.36 0.46 0.21 0.12 Figure 2-1. TPD spectra of D2 O and D2 from Pt(100) (a) D2O TPD spectra obtained after exposing the clean Pt(100)-hex-R0.7 surface to D2O at 90 K, and (b) D2 TPD spectra obtained after exposing the clean Pt(100)-hex-R0.7 surface to D2 at 150 K. A linear heating rate of 1 K/s was used for all m easurements. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. W eaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, American Institute of Physics. 47

PAGE 48

020406080100120140160 O(g) + D/Pt(100) Ts= 90 K [D]0= 1.15ML D2O D2 D2, D2O desorption rate (a.u.)time (s) Figure 2-2. Normalized direct product mon itoring spectra of D2O and D2 obtained while exposing the deuterium-saturated ([ D ]o = 1.15 ML) Pt(100) surface held at 90 K to an oxygen atom beam (~0.018 ML/s). The beam shutter is opened at the zero of time. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, American Institute of Physics. 48

PAGE 49

0.00.51.01.52.02 0.0 0.2 0.4 0.6 0.8 1.0 1.2 .5 D2O (90 K) Post-Reaction TPD O(g) + D/Pt(100) Ts = 90 K O2D2OTPD yields (ML of D)O atom exposure (ML)D2 Figure 2-3. Yields of D2, D2O and O2 that evolve in TPD after exposing deuterium-saturated Pt(100) held at 90 K to an oxygen atom beam The yields are shown as a function of the atomic oxygen beam exposure, and are ex pressed in units of ML of D atoms or, equivalently, ML of O atoms for O2. The dashed line labeled D2O (90 K) represents the yield of D2O that desorbs during the O atom beam exposure at 90 K, as determined by integrating the rate curve shown in Figure 2-2. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, Americ an Institute of Physics. 49

PAGE 50

100 200 300 400 500 (x 4) Post-Reaction D2O TPD O(g) + D/Pt(100) at 90 K2.2 O ML D2O 0.52 ML(D) 0.72 O ML D2O 0.54 ML(D) 0.47 O ML D2O 0.57 ML(D) 0.25 O ML D2O 0.31 ML(D) D2O desorption rate (a.u.)0.09 O ML D2O 0.17 ML(D)temperature (K) Figure 2-4. TPD spectra of D2O obtained as a function of th e atomic oxygen exposure to the deuterium-saturated Pt(100) su rface held at 90 K. Show n by each spectrum is the corresponding O atom exposure in units of ML of O atoms (top value), and the D2O coverage in units of ML of D atoms (bo ttom value). The dotted curve shows the TPD spectrum obtained after adsorbing D2O on Pt(100) to an initial coverage equivalent to 0.17 ML of D atoms. For each TPD spect rum, the region above 230 K is magnified by 4x and superposed on the figure (dashed cu rves). Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. W eaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, American Institute of Physics. 50

PAGE 51

0.0 1.0 2.0 3.0 4.0 5.0 0204060801001200.0 0.4 0.8 1.2 (a) D2O rate/10-3 (ML/s)(b) O(g) + D/Pt(100) Ts = 90 K time (s)D2 rate/10-3 (ML/s) [D]o (ML) 1.15 0.95 0.65 0.44 0.18 Figure 2-5. Direct product monitoring spectra of (a) D2O and (b) D2 measured for different initial D atom coverages on Pt(100). The surface was held at 90 K for each measurement. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, J. Chem. Phys. 126 (2007) 134704/1. C opyright 2007, American Institute of Physics. 51

PAGE 52

01020304050607080 [D]o = 0.44 ML D2O desorption rate (a.u.)time (s) [D]o = 0.65 ML [D]o = 1.15 ML 90 K 150 K O(g) + D/Pt(100) Figure 2-6. Direct product monitoring spectra of D2O obtained for initial deuterium coverages of 0.44 ML, 0.65 ML and 1.15 ML and surface temperatures of 90 K and 150 K. The vertical axes are set to the same scale in each graph. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. W eaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, American Institute of Physics. 52

PAGE 53

0.00.20.40.60.81.01.21.4 0.00 0.05 0.10 0.15 0.20 0.25 150 K 150 K 90 K 90 K [D]o (ML) D2O D2O(g) + D/Pt(100) desorption yields (ML of D) Figure 2-7. Desorption yields of D2O (squares) and D2 (circles) obtained during surface exposure to the oxygen atom beam shown as a function of the initial D atom coverage for surface temperatures of 90 K (filled symbols) and 150 K (open symbols). The dashed lines are fits to the data illu strating that the yields vary with Reprinted with permission from H.H. Kan, R.B. Sh umbera, J.F. Weaver, J. Chem. Phys. 126 (2007) 134704/1. Copyright 2007, Amer ican Institute of Physics. 3.1][oD 53

PAGE 54

CHAPTER 3 ADSORPTION AND ABSTRACTION OF OXYGEN ATOMS ON PA LLADIUM (111): CHARACTERIZATION OF THE PRECURSOR TO PALLADIUM OXIDE FORMATION Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyright 2008, Elsevier. Introduction Palladium is an important catalyst for commerc ial oxidation processes such as the catalytic oxidation of methane in gas turbines [1-11] and the oxidation of hydrocarbons and CO in automotive exhausts [12-19]. Past research indi cates that both metallic Pd and PdO can be catalytically active toward CH4 and CO oxidation, though deba te continues over which oxygen states are more active and how conditions affect catal ytic behavior. Intermediate oxygen phases have been the focus of several recent investig ations of oxidation reac tions on Pd surfaces. For example, Zheng and Altman dem onstrated that an ordered surf ace oxide is less active than chemisorbed oxygen on Pd(100) toward the oxidati on of CO at temperatures above 300 K [19]. A more recent study of CO oxida tion on Pd(111) reports similar behavior [18]. In contrast, Gabasch et al. [10] find that so-called PdO seed s, which develop prior to bulk PdO formation, exhibit much higher activity for CH4 oxidation than other surface oxygen phases on Pd(111). This result bears similarity to another recent study [26] which demonstrates that the CO oxidation activities of several late transition-me tal surfaces undergo a dramatic increase near oxygen coverages of one monolayer (ML), just prio r to bulk oxidation. These findings signal the need for a more detailed understanding of the inte rmediate phases that develop on late transition metal surfaces, and the processes governing their formation and transformation to bulk oxides. The development of oxygen phases on Pd(111) has been studied exte nsively in ultrahigh vacuum [21-23,25,99-114], with most work focu sing on relatively low oxygen coverages. These 54

PAGE 55

studies show that oxygen atoms initially arrang e into a p(22) struct ure with a local oxygen coverage of 0.25 ML. At higher coverage an ordered two-dimensional (2D) oxide forms on Pd(111) [100,102], reportedly saturating at a coverage of 0.58 ML [99]. R ecent studies show that the 2D oxide has a Pd5O4 stoichiometry, and is incommensur ate with the Pd( 111) lattice, the structure resembling a (66) arrangement [100]. Other oxygen phases can also populate on Pd(111), but appear to be me ta-stable relative to the Pd5O4 surface oxide. These phases include a supersaturated (2) layer of chemisorbed oxyge n and several surface oxide structures [21,99]. While most prior UHV work has examined relatively low oxygen coverages on Pd(111), researchers have recently inve stigated oxidation up to coverage s at which PdO forms [22,23,25]. Of particular relevance to the present study are in situ XPS investigations of Pd(111) oxidation at millibar pressures of O2 [22,23]. This work shows that an intermediate state of oxygen forms after saturation of the 2D Pd5O4 oxide, and acts as a precursor to PdO formation. This intermediate state produces distinct characteristics in XPS spectra from Pd(111), and is highly active toward CH4 oxidation, but its propert ies are otherwise unknown. In related work, we have presented evidence th at a precursor state mediates the formation of Pt oxide particles on both Pt(111) [88] and Pt(100) [96,115] during oxidation by gaseous oxygen atoms. Our experimental data suggests that the precur sor state corresponds to oxygen atoms adsorbed on top of a 2D oxide. Evidence for the precursor state comes firstly from O2 temperature programmed desorption (TPD) spect ra which show that a low temperature desorption feature, attributed to the precursor, intensifies as th e oxygen coverage increases above about 0.5 ML, but then diminishes as a new peak due to Pt oxide particles intensifies above 1 ML. The precursor state from Pt (100) evolves in a broad peak centered at about 550 K in TPD spectra while other states desorb above 630 K [96,116]. In addition, using a relatively low 55

PAGE 56

oxygen atom flux (~0.003 ML/s), we could generate bulk Pt oxide on Pt (100) only at a surface temperature of 450 K, whereas the surface effec tively saturates upon completion of the 2D oxide at 573 K [96]. This observation indicates that the low temperat ure state must populate for bulk oxide to form. In a subsequent investigation, we showed that the kinetics of oxygen uptake on Pt(100), measured as a function of the surface temperature and the Oatom beam flux, is quantitatively reproduced by a precursor-mediate d model wherein oxygen atoms adsorbed on top of the 2D oxide act as a precursor that can either associatively de sorb or react with the 2D oxide to produce a bulk oxide particle [115]. Given that surface oxide phases grow on several late transition metal surfaces [100,115,11 7-120], it is conceivable that pr ecursor-mediated kinetics is a general characteristic of the oxidation of these materials. Additional studies are clearly needed to test this idea and to further characterize the nature of the precursor stat e(s). In this study, we investigated the adsorption a nd abstraction of oxygen atoms on Pd(111), and present evidence that oxygen atoms adsorbed on top of the 2D Pd5O4 oxide exist and act as a precursor to PdO formation. Experimental Methods Previous studies [88,96,115,116,121] provide details of the thr ee-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 c opper sample holder in thermal contact with a liquid nitrogen c ooled reservoir. A type K th ermocouple spot-w elded to the backside of the crystal allows sample temperat ure measurements. Resistive heating, controlled using a PID controller that vari es the output of a programmabl e DC power supply, supports maintaining or linearly ramping the sample temperature from 81 K to 1250 K. Initially, sample cleaning consisted of sputtering with 600 eV Ar+ ions at a surface temperature of 900 K, followed by annealing at 1100 K for several minut es. Subsequent cleani ng involved routinely 56

PAGE 57

exposing the sample held at 856 K to an atom ic oxygen beam for several minutes, followed by flashing the sample to 923 K to desorb oxygen and carbon oxides. We considered the sample to be clean when we could no l onger detect contaminants with X-ray photoelectron spectroscopy (XPS), obtained sharp low ener gy electron diffraction (LEED) pa tterns 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 s ource (Oxford Scientific Instrument s) utilized to generate beams containing oxygen atoms for this study [115,121]. This system produces gaseous oxygen atoms by partially dissociating pure O2 (BOC gases 99.999%) contin uously supplied to a small discharge chamber at the end of the plasma s ource. All oxygen atom beams used in this study were generated using an RF power of 120 W and an O2 flow rate that establishes a pressure of 3 10-6 Torr in the first pumping stage of the beam chamber. Under these conditions, we estimate that 20% of the inlet O2 dissociates in the plasma. To ensu re 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-c ollimating 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 lo wer bound of 0.02 ML/s at the sample surface, where 1 ML is defined as the atomic density of 1.53 1015 cm-2 of bulk-terminated Pd(111) [99]. Following oxygen atom exposures to the Pd(111) surface, we conducted O2 TPD by first cooling the sample to 400 K, then facing the sample toward the entrance of a quadrupole mass spectrometer ionizer at a distan ce of about 10 mm, and lastly in creasing the sample temperature at a linear rate of 1 K s-1. Cooling the sample prior to TPD avoi ds an initial transient response of 57

PAGE 58

the sample heater and ensures a linear temperature ramp over the range of O2 desorption. We performed low energy ion scatteri ng spectroscopy (LEISS) using He+ ions with an incident energy of 1 keV and a scattering angle of about 90. The hemisphe rical analyzer was operated in a retarding mode with a pass energy of 27 eV for the LEISS measurements. Finally, we conducted beam scattering experiment s to examine the kinetics of 18O abstraction from Pd(111). In these measurements, we first adsorbed 0.25 ML of 18O atoms by exposing the Pd(111) surface to 20 L of 18O2 with the sample held at 300 K. We then exposed the surface to an 16O atom beam while monitoring masses 32, 34 and 36 amu with a mass spectrometer. Since we monitor the random fluxes of desorbing products, these expe riments provide a direct measure of the 18O abstraction rate as a f unction of the time of the 16O beam exposure. After terminating a beam exposure, we reposition the sample to dire ctly face the mass spect rometer for TPD. Results and Discussion Oxygen Coverage Calibration and Dissolution into the Bulk Before presenting O2 TPD data, we clarify our pro cedure for quantifying oxygen surface coverages and our attempt to minimize complications to this ca libration due to oxygen dissolution into the Pd(111) s ubsurface region. The oxygen coverage scale was calculated by comparing integrated O2 TPD signal areas to the areas obtained after saturating the surface at 300 K with molecular oxygen and assuming the at omic oxygen coverage is 0.25 ML [102]. As mentioned above, we cleaned the Pd(111) samp le at the start of each day using a large 16O atom dose with the sample held at 856 K, followed by flashing to 923 K. In addition to removing carbon impurities, the intention of this procedur e is to saturate the subsurface region with 16O atoms and thereby minimize loss of chemisorbed oxygen atoms due to bulk dissolution during subsequent TPD experiments [ 103]. Limiting the sample te mperature to 923 K during TPD should also help to maintain approximate ly the same subsurface oxygen concentration 58

PAGE 59

throughout the day [103]. This proce dure appears to be effective as we find that the integrated TPD peak areas are highly reproducible fr om repeated saturation exposures to 16O2. Interestingly, TPD experiments with chemisorbed 18O atoms reveal that oxygen exchange with the subsurface does occur durin g heating. In particular, afte r preparing a layer of 0.25 ML 18O atoms by dosing with 18O2 at 300 K, we observe all isotopic O2 combinations desorbing in the feature associated with chemisorbed oxygen, and no other desorption features below 923 K. Since our experiments with pure 16O layers indicate that the total concentration of chemisorbed oxygen remains constant when the subsurface is saturated, we conclude that the oxygen exchange only alters the isot opic composition of the chemisorbe d layer while maintaining the total coverage at 0.25 ML, at least near temper atures for appreciable recombinative desorption. Under this assumption, we estimate that 0.18 ML of 18O atoms desorb during TPD, with the remaining 0.07 ML exchanging with subsurface 16O atoms. We also find that the amount of atomic 18O that exchanges with the subsurface, as estimated from the 18O desorption yield, changes negligibly when the total coverage is increased by adding 16O atoms (as much as 3 ML) to a surface held at 500 K and initially with 0.25 ML of 18O. This observation suggests that oxygen dissolution into the metal subsurface occu rs mainly at low coverage when oxygen atoms are chemisorbed on Pd(111), while the presence of oxide phases, which form above 0.25 ML, suppresses bulk dissolution at least at 500 K. Thermal Desorption of O2 Figure 3-1 shows a series of O2 TPD spectra collected after exposing Pd(111) to gas-phase oxygen atoms at a substrate temperature of 500 K. The TPD spectra agree well with those reported in prior studies of Pd(111) oxidation by O2 [99] and NO2 [102] up to the oxygen atom coverages that are acce ssible with these oxidants in UHV. As is well known, oxygen atoms chemisorb into a p(2) structure up to 0.25 ML from which recombinative desorption produces 59

PAGE 60

a broad feature in the TPD spectra between 650 and 875 K. Beyond 0.25 ML, we observe a sharp desorption peak that shifts from 700 to 710 K as the coverage increase s to approximately 0.7 ML. Other researchers have shown that this deso rption feature originates from the ordered 2D Pd5O4 oxide that decomposes thr ough an autocatalytic process involving quasi-equilibrium between O atoms in the 2D oxide and the p(2) phase [99,101]. As the oxygen coverage increases above 0.7 ML the desorption rate initially increases on the leading edge of the sharp 2D oxide peak, and then develops into a separate feature centered at 620 K as the coverage reaches 1.2 ML. Concurren tly, the sharp peak appe ars to shift toward higher temperature and intensify with increasing coverage. A shoul der is also evident at about 690 K on the sharp peak obtained from Pd(111) with 1.2 ML oxygen coverage. Further increasing the oxygen coverage caus es the feature at 620 K to di minish while the sharp peak continues to intensify and shift to higher temperature. Using XPS, we co nfirmed that bulk-like PdO forms on Pd(111) above an oxygen coverage of about 1.5 ML, and th erefore attribute the intense desorption peak to the autocatalytic d ecomposition of PdO domains. Since formation of PdO domains consumes the 2D oxide, one would expect the desorption peak arising from 2D oxide decomposition to diminish with coverage. The shoulder observed at 690 K and 1.2 ML initial oxygen coverage is consistent with deco mposition of a small fraction of 2D oxide domains present at this coverage. The O2 desorption behavior observed here for Pd(111) is qualitatively similar to that reported recently for both oxidized Pt(111) [88] and Pt(100) [96,116] in that a low temperature feature intensifies with increasing coverage and then diminishes as bulk oxide begins to grow. As discussed in the Introduction, we have shown th at this low temperature state is distinct from the surface and bulk oxides on Pt(100), and serves as a necessary precursor to the formation of 60

PAGE 61

bulk-like Pt oxide [115]. Indeed, the present results imply that a similar oxidation mechanism is operative for Pd(111) in whic h a weakly-bound precursor state initiates the transition from surface to bulk oxide growth. Recent in situ XPS investigations also report that a distinct oxygen state populates on Pd(111) at coverages intermedia te to the 2D oxide and bulk PdO [22,23]. The Pd core level spectra reported in that work reveal that formation of the intermediate state is characterized by an increase in the relative amount of Pd atoms that are four-fold versus two-fold coordinated with oxygen atoms. A possible interpretation of this spectral change is that formation of the intermediate state involves oxygen atoms prefer entially adsorbing onto two-fold coordinated sites of the 2D oxi de, while preserving the struct ure of the underlying 2D oxide. Such a configuration has a lower concentration of two-fold coordinated Pd atoms than the 2D oxide, and is consistent with the idea that the precursor state to initial PdO formation corresponds to oxygen atoms adsorbed on top of the 2D oxide of Pd(111). LEISS Measurements To further characterize the precu rsor state, we collected He+ LEISS spectra of the clean surface, the 2D oxide (0.72 ML), the precursor-covered surface (1.5 ML), and PdO (2.7 ML) on Pd(111). For these experiments, oxygen-covered surfaces were prepared at 100 K to suppress PdO formation and thereby maximize the concentra tion of the precursor state. The surfaces were then heated to 500 K to desorb any molecularly chemisorbed O2 that may be present at low temperature. The O2 TPD spectra obtained af ter adsorbing O atoms at 100 K (not shown) versus 500 K show similar peak evolution with coverage. However, the O2 TPD peak at 620 K can be made more intense by oxidizing at 100 K, with this peak reaching a maximum intensity at a total coverage of 1.5 ML rather th an 1.2 ML as observed for oxidation at 500 K (Figure 3-1). Most likely, the rate of conversion of the precursor to PdO is lower at 100 K than 500 K, resulting in higher concentrations of oxyge n in the precursor state. 61

PAGE 62

Figure 3-2a shows normalized LEI SS spectra as a function of the final ion energy for the representative oxygen-covered surf aces. The spectra exhibit only two features centered at 620 and 903 eV due to the incident ions scattering from surface O and Pd atoms, respectively. Note that the oxygen ISS peaks are broader than the Pd peaks due to improved mass resolution for He+ scattering from the lighter elements [122]. For the surfaces studied, the highest O/Pd intensity ratio is seen in the LEISS spectrum obtained from the precursor-covered surface with an oxygen coverage of 1.5 ML. The higher O/Pd signal is indeed consistent w ith oxygen atoms being adsorbed on top of the 2D oxide, and therefore bei ng accessible for collisions with ions incident from the vacuum. Interestingly, the LEISS spectra also reveal that the O/Pd ratio is about twice as high for Pd(111) covered with PdO compared with the 2D oxide, suggesting that the bulk and surface oxides have different surface compositions. To evaluate the stability of the vari ous oxygen phases toward reduction by ion bombardment, we collected LEISS spectra every five minutes while continuously subjecting the surfaces to 1 keV He+ ions. Figure 3-2b shows the evoluti on of the O/Pd signal ratio as a function of sputtering time for Pd(111) initia lly covered predominantly with chemisorbed oxygen atoms (0.22 ML), the 2D oxide (0.72 ML), the precursor state (1.5 ML) and PdO (2.7 ML). This plot reveals that the precursor-cove red surface is more easily reduced by incident He+ ions than the other oxygen phase s on Pd(111). In fact, the O/Pd signal ratio obtained from the precursor-covered surface decays ra pidly in the first fifteen minutes until reaching a value close to that obtained from the 2D oxide, and remains steady thereafter. The higher sputtering rate from the precursor-covered surface provides furthe r support for the conclusion that the precursor state does correspond to oxygen atoms adsorbed re latively weakly on top of the 2D oxide. 62

PAGE 63

Figure 3-3 depicts the general mechanism fo r Pd(111) oxidation that we propose based on the present data as well as our prior results wi th Pt surfaces [115]. We note that surface oxygen phases other than those depicted in Figure 3-3 have been identified, or at least suggested to exist on Pd(111), including a p(2) pha se [100] and a variety of surf ace oxide structures [21,99,102]. As such, we emphasize that the intention of Figur e 3-3 is to illustrate the key stages in the evolution of oxygen phases that l ead to PdO formation on Pd(111) as suggested by our data, and point out that pathways to some of these phases likely pass th rough other structures. Figure 3-3 shows that oxygen atoms initially populate a chemisorbed layer that begins to transform to a 2D oxide after the oxygen coverage surpasses a cr itical value (~0.25 ML). Oxygen atoms then adsorb on top of the 2D oxide a nd accumulate in the so-called pr ecursor state. Beyond a critical local coverage of precursor oxygen atoms, r eaction between the precursor oxygen and the 2D oxide produces PdO domains, depicted in Figure 3-3 as three-dimensi onal particles [115]. Direct Abstraction of 18O/Pd(111) by Gaseous 16O Atoms We conducted reactive scattering experiments to assess the reactivity of the various oxygen phases toward abstraction by gase ous oxygen atoms. In these experiments, we first adsorbed 0.25 ML 18O atoms by exposing the Pd(111) surface to 18O2 with the sample held at 500 K. Subsequently, we exposed the surface to an 16O atom beam with the sample temperature fixed, while monitoring the desorption rate of 18O16O using a mass spectrometer. This experiment provides a direct measure of the isothermal rate of 18O abstraction by incident 16O atoms as a function of the beam exposure time. We find that the collisions of gaseous 16O atoms with 18O-covered Pd(111) result in the prompt desorption of 18O16O at surface temperatures well be low that for thermally-activated recombination (Figure 3-4a). Su ch behavior suggests that 18O abstraction by gaseous oxygen atoms occurs by a non-thermal mechanism such as the Eley-Rideal or hot atom mechanism 63

PAGE 64

[70,121,123]. Further evidence that abstraction is non-thermal is the relative insens itivity of the initial abstraction rate on th e surface temperature. Although omitte d from Figure 3-4a for clarity, we find that the initial abstrac tion rate depends only weakly on the surface temperature down to at least 100 K. After the initial maximum, the abstraction rate d ecays rapidly but then shows an unexpected increase at longer exposur e times. The abstraction kinetics at later times also depends markedly on the surface temperature, with the intensity of the second maximum increasing with increasing surface temperature (Figure 3-4a). After passing through the second maximum, the abstraction rate decays to a steady value. As elaborated below, we attribute the surface temperature dependence observed at later times to the onset of a thermally-activated process that influences the concentration of 18O atoms available for non-thermal abstraction by gaseous 16O atoms. TPD measurements performed after vari ous beam exposures show that the 18O abstraction yields are very low for the c onditions examined. In these measurements, we discontinued the beam exposures at different times and then he ated the sample while monitoring masses 32, 34 and 36 amu to determine the 16O and 18O coverages as a function of the beam exposure. From this data, we are unable to detect a decrease in the 18O coverage with beam exposure, but rather observe that the 18O coverage remains constant within the error limits of the coverage estimations ( 0.02 ML). In contrast, the 16O coverage increases steadily during the beam exposure. For example, the 16O coverage is 0.5 ML after a 30 sec beam exposure at a surface temperature of 500 K, and reaches 2.8 ML after 540 sec. Since the 18O coverage remains approximately constant as the 16O coverage increases, the changes in the 18O abstraction rate with beam exposure time cannot be attributed to depletion of the 18O coverage. Instead, the 18O abstraction kinetics must refl ect how the oxygen phase evolution influences the reactivity of 64

PAGE 65

surface oxygen atoms toward abstraction by ga seous oxygen atoms. Thus, the abstraction measurements effectively provide an in situ means to probe the oxygen phases as they evolve on Pd(111). Figure 3-4b shows the 18O abstraction rate at 500 K plotted as a function of the total oxygen atom coverage where the coverages we re determined from post-abstraction TPD measurements. We find that the ab straction rate decreases to a minimum as the total coverage increases from 0.25 to 0.70 ML, coinciding exact ly with the development of the 2D oxide (Figure 3-1). The abstraction ra te then increases as the coverage increases beyond 0.7 ML, reaching a maximum at a total oxygen coverage of 1.3 ML. This coverage range corresponds to population of the weakly bound precursor state ob served in TPD (Figure 3-1). Finally, the abstraction rate decreases again as the total cove rage increases from 1.3 to 3.1 ML. As discussed above, the surface oxygen reacts to form PdO over this coverage range, consuming the precursor state in the process. These obs ervations demonstrate that ga seous oxygen atoms can abstract oxygen atoms chemisorbed on the metal surface a nd on top of the 2D oxide, but that oxygen abstraction from the surface and bulk oxides occurs at much lower rates. Physically, this seems reasonable considering that chemisorbed oxygen atoms are more accessible for collision with species incident from th e gas-phase. The data further implies th at the rate of ab straction of the precursor oxygen atoms increases with surface temperature, whereas abstraction of oxygen atoms chemisorbed on the metal surface is essent ially temperature independent (Figure 3-4a). This result is surprising given that the initial abst raction seems to occur by a non-thermal mechanism. In the abstraction experiments, 18O atoms are initially chemisorbed on the Pd(111) surface and are hence available for direct abstraction by gaseous 16O atoms. According to the mechanism 65

PAGE 66

shown in Figure 3-3, the 18O atoms incorporate into the 2D oxide as the total oxygen coverage increases, and are rendered much less reactive towa rd direct abstraction as a result (Figure 3-4b). Additional surface exposure to the 16O beam then results in the adsorption of 16O into the precursor state, followed by PdO formation. Notice that Figure 3-3 suggests that only 16O atoms populate the precursor state during the beam exposures, while the 18O atoms would be present in the 2D oxide. This seems contradict ory to our observations of higher 18O abstraction rates during population of the precursor state, at least within the context of the proposed mechanism, unless we assume that oxygen atoms exchange between th e 2D oxide and the precursor state during the abstraction measurements. Such an exchange pr ocess is likely to be thermally-activated and hence its occurrence co uld explain why the 18O abstraction rate is temperature dependent at coverages corresponding to population of the prec ursor state. Specifically we propose that an increase in the rate of oxygen exchange between the 2D oxide and the precursor state with increasing surface temperature result s in higher concentrations of 18O atoms in the precursor state and consequently higher rates of non-thermal 18O abstraction. Kinetic Model of 18O Abstraction and Exchange We developed and evaluated a kinetic model to test the idea that oxygen atoms exchange between the 2D oxide and the prec ursor state, causing the measured 18O abstraction rates to be temperature dependent in the prec ursor coverage range. Figure 3-5 depicts the steps included in the model. In developing this model, we consid ered that the exchange process is analogous to surface diffusion via concerted substitution, a mechanism proposed originally by Kellogg and Feibelman [124,125]. A requirement fo r concerted substitution is that an adsorbed atom occupies a site next to a vacant site. As illustrated in Fi gure 5, the mechanism involves an adsorbed atom replacing an atom in the surface layer, while the surface-layer atom simultaneously moves to fill 66

PAGE 67

the initially vacant site above the surface layer. A succession of such steps results in net transport of adsorbed atoms along the surface. The processes shown in Figure 3-5 are represented by the following reaction equations, 16O(g) + S 16O(a) (3-1) adr16O(a) + 18O(2D) + S S + 16O(2D) + 18O(a) (3-2) er16 er1816O(g) + 18O(a) 16O18O(g) + S (3-3) abr1816O(g) + 16O(a) 16O2(g) + S (3-4) abr16Equation 3-1 represents the adsorption of 16O atoms on top of the 2D oxide, i.e., adsorption into the precursor state. Equation 3-2 depicts the reversible exchange of 16O and 18O between the 2D oxide and the adsorbed state, O(2D) and O(a) respectively, where the symbol S represent an empty site on top of the 2D oxide. Finally, Equati ons 3-3 and 3-4 represent the direct abstraction by an Eley-Rideal process of 18O and 16O atoms from the precursor state by 16O atoms incident from the gas-phase. Although abstraction may involve hot atom intermediates, we describe the abstraction kinetics usin g the Eley-Rideal rate equation to simplify the analysis. Symbols used to designate the rates of each step are shown next to Equa tions 3-1 through 3-4, where 16re and 18re represent the rates for forward and reverse exch ange, respectively, for Equation 3-2 as it is written. For this kinetic scheme, the time evolution of the 16O(a) and 18O(a) coverages are given by the following equations, max 161816 16][ O rrrr dt dabeead (3-5) max 181816 18][ O rrr dt dabee (3-6) 67

PAGE 68

where 16 and 18 represent the fractional coverages of 16O(a) and 18O(a) species, and [ O ]max is the maximum coverage of oxygen atoms in the pr ecursor state. We assume that each of the reaction steps is elementary and use the followi ng equations to describe the reaction rates, D adSr2 (3-7) D eeOO kr2 18 max 161816 16][][)1( (3-8) D eeOO kr2 16 max 181816 18][][)1( (3-9) max 16 16][ O rab (3-10) max 18 18][ O rab (3-11) where represents the incident oxygen atom flux, represents the initial adsorption probability of a gas phase oxygen atom on the 2D oxide, represents the rate coefficient for oxygen exchange, and represent the concentrations of 16O and 18O within the 2D oxide, and DS2ekDO2 16][DO2 18][ represents an effective cross section for oxygen abstraction from the precursor state by gas phase oxygen atoms. The term in the rate expression for oxygen exchange is a necessary requirement for diffusion by concerted substitution since diffusion can only occur in the presence of empty surface sites. Notice also that Equation 3-7 excludes any dependence of the oxygen atom adsorption rate on the oxygen coverage. Our prior work shows that this is a very good approxi mation for atomic oxygen adsorpti on into the precursor state on Pt(100) up to relatively high coverages [115]. Fu rthermore, we assume that both the atomic oxygen adsorption probability, S2D, and the abstraction cross section, )1(1816 are independent of the surface temperature. Thus, only the ra te coefficient for oxygen exchange ke depends explicitly on the surface temperature in this model. 68

PAGE 69

To solve the species balance equations anal ytically, we make simplifying approximations that are consistent with the obs erved rate behavior. Firstly, since the rate coefficients for the forward and reverse reactions shown in Equation 32 are identical, neglecti ng isotope effects, the distribution of 16O and 18O between the 2D oxide and the prec ursor state should be independent of the surface temperature if the exchange reaction reaches equilibrium. Thus, in our experiments, oxygen exchange appears to be far from equilibrium, given that we observe a relatively strong dependence of the 18O abstraction rate on the surf ace temperature (Figure 3-4a). In this case, it is reasonable to neglec t the reverse exchange reaction, i.e., 16re >> 18re, in the analysis, and assume that initi al conditions prevail during abst raction of the precursor oxygen atoms such that 16 >> 18 and [18O ]2D ~ constant. We further assume that the rate of oxygen atom adsorption is much great er than the rates of exchange and abstraction such that the rate of change of the 16O(a) coverage is well approximated by the 16O adsorption rate. The rapid increase in the 16O coverage justifies the assumption that adsorption is faster than abstracti on. Finally, we assume that the rate of change of the 18O(a) coverage is equal to the rate of forward exchange (Equation 3-2). This is a consequence of neglecting reverse exchange, and also requires that forward exchange is faster than 18O abstraction for the time scale of inte rest. This seems reasonable under the approximation that the 18O(a) coverage is small. With thes e approximations, the species balance equations simplify to the following, max 2 max 16][][ O S O r dt dD ad (3-15) D e eO k O r dt d2 181616 max 16 18][)1( ][ (3-16) 69

PAGE 70

Integration of the species balances gives the following equations for the 16O(a) and 18O(a) coverages as a function of time, t O SD max 2 16][ (3-17) ) 3][2 (][ ][3 max 2 2 2 18 max 2 18t O S t O O S kD D D e (3-18) Substituting Equation 3-18 into Equation 3-11 results in the following equation for the rate of 18O(a) abstraction as a function of time, )()() 3][2 (][3 max 2 2 2 18 2 2 18tfTk t O S t OSkrse D D D eab (3-19) Consistent with the experimental data (Figure 34a), Equation 3-19, with reasonable values for S2D and [ O ]max, predicts that the rate of 18O abstraction is an increasi ng function of time over the measured time scale for abstraction from th e precursor state (~30 sec). The model further predicts that the 18O abstraction rate scales directly with the rate co efficient for exchange. To emphasize that only the exchange rate coeffi cient depends on the surface temperature, we lumped all the other parameters into the function f ( t ) shown on the right hand side of Equation 319. Assuming that the Arrhenius equatio n accurately describes the function ke( Ts), the model predicts that a plot of ln(18rab) versus Ts -1 should be linear. However, in order to construct such a plot, we must identify the time at which oxygen at oms first begin to populat e the precursor state for each measured rate curve shown in Figure 3-4a. This could introduce uncertainty in the analysis since it is unclear that the zero of tim e for the model corresponds to the minimum in the measured rate curve. To circumvent this uncerta inty, we consider an alternate analysis based on 70

PAGE 71

the observation that the abstracti on rate as a function of time exhi bits an inflection point in both the experimental data and the model. This inflec tion point represents the point of steepest ascent for the abstraction rate in the region where oxygen atoms populate the pr ecursor state and is therefore easily identified in the data, with the slope obtained by means of a linear fit. Furthermore, according to the model, the inflection point occurs at a time that is independent of the surface temperature, DS O t2 max inf][2 We can therefore avoid the n eed to define a time axis for precursor population by evaluating each rate curve at the inflection point. In particular, we evaluate the slope of the measur ed rate curve at the inflection point since this slope is easily determined from the data. From the model, the derivative of the 18O abstraction rate with respect to time is given by the following equation, ) ][ (][2 max 2 2 18 2 2 18t O S tOSk dt rdD D D e ab (3-20) Expressing ke in an Arrhenius equation and subs tituting into Equation 3-20 yields, ) ][ )(exp(][2 max 2 2 18 2 2 18t O S t RT E OS dt rdD e D D e ab (3-21) where e and Ee are the pre-expon ential factor and activation energy fo r oxygen exchange between the 2D oxide and the precursor state. Finally, writing Equation 3-21 for an Arrhenius construction gives, s e abRT E ttgtt dt rd )(ln))(ln(inf inf 18 (3-22) where the time derivative of 18rab is evaluated at th e inflection point, and g ( t=tinf) is independent of the surface temperature. 71

PAGE 72

Figure 3-6 shows the experimenta lly determined slopes at the inflection points plotted in the Arrhenius form given in Equatio n 3-22. As seen in the figure, a linear fit to the data in this plot is excellent ( R2 = 0.994), suggesting that the model correctly describes the surface temperature dependence of the 18O abstraction rate. From the slope of this plot, we estimate an activation energy of Ee = 15.2 0.4 kJ/mol for oxygen exch ange between the 2D oxide and the precursor state. Assuming that the exchange reac tion does result in surface transport, the small value of Ee suggests that oxygen atoms diffuse rapidl y on the surface of the 2D oxide via the concerted substitution mechanism. For comparison, the activation energy for oxygen diffusion by site hopping on clean Pd(111) is reported to be 59.8 kJ/mol [ 126]. According to Feibelman, the barrier for diffusion by concerted substitution is generally smaller than that for site hopping [124,125]. While the barrier fo r oxygen site hopping on the 2D oxide is unknown, we can determine desorption activation energies from th e TPD data and then estimate ratios of the diffusion (or exchange) to desorption barriers. Using a prefactor of 1013 s-1, Redhead analysis of the TPD data (Figure 3-1) pred icts that the activation energies for desorption are 170 and 215 kJ/mol for oxygen atoms chemisorbed on the 2D oxide and the bare Pd(111) surface, respectively. With these values, we estimate that the barrier for oxygen site hopping on clean Pd(111) [126] is about 28% of the desorption barr ier, while the barrier for diffusion by concerted substitution on the 2D oxide is on ly about 9% of the desorption ba rrier. If the 2D oxide and clean Pd(111) surfaces have similar corrugation, this co mparison suggests that the barrier for concerted substitution is smaller than that for site hopping on the 2D oxide. DFT calculations of oxygen atom diffusion on the 2D oxide should be helpful in testing this suggestion. Summary We investigated the adsorp tion and abstraction of oxygen atoms on Pd(111) using oxygen atom beams in UHV, and present evidence that a precursor state mediates the initial formation of 72

PAGE 73

PdO on this surface. The O2 TPD spectra show that oxygen at oms in the precursor state are weakly bound compared with oxygen in the 2D oxide and PdO, and that population of the precursor state increases upon saturation of the 2D oxide, and then diminishes above about 1.2 ML as PdO domains begin to grow on the surface at 500 K. We attribute the precursor state to oxygen atoms chemisorbed on top of the 2D Pd5O4 oxide. Consistent with this interpretation, LEISS spectra indicate that the precurso r-covered surface has a higher surface oxygen concentration than both the 2D oxide and th e PdO that forms, and that incident He+ ions sputter oxygen more readily from the precursor state th an from other surface oxygen phases on Pd(111). Considering similar findings for Pt(111) [88] an d Pt(100) [96,115], the results of the present study support the idea th at precursor-mediated kinetics is a ge neral characteristic in the oxidation of late transition metal surfaces. We also find that the collisions of gaseous 16O atoms with 18O chemisorbed on Pd(111) causes the prompt desorption of 16O18O at surface temperatures from 100 to 550 K. The insensitivity of the init ial abstraction rate to the surface temperature is consistent with a nonthermal mechanism for 18O abstraction. The direct rate measurements show that abstraction of 18O atoms chemisorbed on the metal surface and in th e precursor state is facile compared with abstraction from the 2D oxide and PdO, but the total 18O abstraction yield is very low, falling within the error limits of oxygen coverage estimates. The rate of 18O abstraction from the precursor state exhibits an unexpected increase with increasing surface temperature. We show that this temperature dependence is well desc ribed by a kinetic model which assumes that abstraction occurs by a non-ther mal Eley-Rideal mechanism and that oxygen atoms exchange between the 2D oxide and the precursor state by a thermally-activated process analogous to surface diffusion by concerted substitution. For th e experimental conditions studied, the model 73

PAGE 74

predicts that oxygen exchange causes the conc entration of 18O atoms in the precursor state to increase during the abstraction measurements, resulting in higher 18O abstraction rates at higher surface temperature. From an alysis of the measured 18O abstraction rates, we estimate an activation energy of 15.2 kJ/mol for oxygen excha nge between the 2D oxide and the precursor state. This low barrier may i ndicate that chemisorbed oxygen at oms diffuse rapidly on the 2D oxide surface by the concerted substitution mech anism. Given their relatively weak binding and high mobility on the 2D oxide, it is conceivable that oxygen atoms adsorbed in the precursor state exhibit unique reactivity towa rd adsorbed molecules such as CO. This possibility warrants further investigation as it could help to explain the r eactive behavior of Pd catalysts in varying oxidative environments. 74

PAGE 75

500550600650700750800850900 0.0 0.5 1.0 1.5 2.0 2.5 2 4 6 8 precursor PdO 2D oxide O2 TPD O(g) + Pd(111) Ts=500 K o (ML) 0.08 0.22 2.0 0.30 0.37 0.45 0.72 0.94 1.2O2 desorption rate (10-2 ML/s)Temperature (K)PdO O2 TPD O(g) + Pd(111) Ts=500 K o (ML) 1.2 2.8 Figure 3-1. O2 TPD spectra (heating rate = 1 K/s) obtained after expos ing Pd(111) held at 500 K to an oxygen atom beam. Notice that the vertical axis in the top panel is expanded by a factor of 3 relative to the bottom panel. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyri ght 2008, Elsevier. 75

PAGE 76

05101520 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11 20030040050060080090010001100 0.0 0.2 0.4 0.6 0.8 1.0 1.2 (b)He+ + O/Pd(111) tscan = 2 min O/Pd LEISS Intensity RatioTime (min) o ML 0.22 0.72 1.5 2.7(a)Pd O (x 8) He+ ISS O/Pd(111) Eo = 1000 eVN(E) (a.u.)Efinal (eV) o (ML) clean 0.72 1.5 2.7 Figure 3-2. LEISS spectra obtai ned after adsorption of oxygen on Pd(111) a) Normalized He+ LEISS spectra obtained from clean and oxygencovered Pd(111). The oxygen coverages are representative of the 2D oxide (0.72 ML), the precursor state (1.5 ML) and PdO (2.7 ML). For each measurement the He+ incident energy was 1 keV and the scattering angle was 90. b) Ratio of O/Pd LEISS signal in tensities as a function of 1 keV He+ exposure time for the initial oxygen cove rages listed in the figure. Each data point required two minutes to collect. Re printed with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyri ght 2008, Elsevier. 76

PAGE 77

a) Chemisorbed b) 2D oxide Pd Pdd) 3D oxide c) Precursor Pd PdO Pd Figure 3-3. Schematic of the key stages/oxygen phases in the proposed m echanism for the initial formation of PdO on Pd(111), including a) chemisorbed oxygen atoms, b) 2D oxide, c) oxygen atoms adsorbed on the 2D oxide, d) formation of th ree-dimensional PdO domains. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyright 2008, Elsevier. 77

PAGE 78

0.00.51.01.52.02.53.03.5 0 50 100 150 200 (b)PdO precursor 16O(g) + 18O/Pd(111) Ts = 500 K2D oxide d[18O16O]/dt (a.u.)[18O] + [16O] (ML) (a)16O(g) + 18O/Pd(111) [18O]o = 0.25 ML = 0.02 ML/sd[18O16O]/dt (a.u.)Time (s)Ts (K) 300 350 400 450 500 550 Figure 3-4. Reactive scatteri ng measurements of isotopic oxy gen abstraction by gas-phase oxygen atoms on Pd(111) a) Isothermal rate of 18O16O desorption as a function of 16O beam exposure time for surface temp eratures from 300 to 550 K. For each experiment, clean Pd(111) was ini tially covered w ith 0.25 ML of 18O atoms, and then exposed to an 16O atom beam at an incident flux of ~0.02 ML/s. The zero of time corresponds to opening the beam shu tter. b) Isothermal rate of 18O16O desorption as a function of total oxygen coverage for a su rface temperature of 500 K where the total oxygen atom coverage was determined from post-abstraction TPD measurements. Reprinted with permission from H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyri ght 2008, Elsevier. 78

PAGE 79

r e r ab 16 O(a) 2D oxide Pd Pd r ad Figure 3-5. Schematic of the elementary steps co nsidered in the kinetic model for abstraction of 18O atoms (dark circles) adsorbed on the 2D oxide by 16O atoms (light circles) incident from the gas-phase. Exchange of oxygen isotopes between the 2D oxide and the precursor state increases the concentration of 18O(a) species that are available for abstraction by gaseous 16O atoms. Reprinted with perm ission from H.H. Kan, R.B. Shumbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyri ght 2008, Elsevier. 0.00200.00210.00220.00230.00240.00250.00260.00270.0028 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 ) ln(18dt rdabEe = 15.2 0.4 kJ/mol R2=0.994 1/Ts (K-1) Figure 3-6. Arrhenius plot of the maximum sl ope obtained in the dir ect rate measurements (Figure 3-4) during the period over which the abstracti on rate increases as oxygen atoms populate the precursor state. The error bars represent uncerta inty in the linear fits used to estimate the maximum slopes of the abstraction rate curves. Reprinted with permission from H.H. Kan, R.B. Sh umbera, J.F. Weaver, Surf. Sci. 602 (2008) 1337. Copyright 2008, Elsevier. 79

PAGE 80

CHAPTER 4 A PALLADIUM OXIDE (101) THIN FI LM GROWN ON PALLADIUM (111) IN ULTRAHIGH VACUUM Reprinted with permission from H.H. Kan, J.F. Weaver, Surf. Sci. 602 (2008) L53. Copyright 2008, Elsevier. Introduction Interest in the catalytic properties of late transition-metal oxides has grown significantly in recent years due to increasing demands for more efficient energy conversion processes and the concomitant need for catalysts that oper ate effectively under oxygen-rich conditions. Considerable effort has been devoted toward ch aracterizing the catalytic properties of palladium oxide (PdO) in particular due to the favorable performance of Pd in catalytic methane oxidation in lean gas turbines [1-11]. Unfortunately, howev er, experimental challenges in oxidizing late transition metals in ultrahigh vacuum ( UHV) have significantly limited fundamental investigations of the surface properties of PdO a nd other oxides. Despite this limitation, recent studies have provided im portant insights for understanding the growth and properties of PdO and other high-concentration oxygen phases on single crystal Pd surfaces [22-25,127,128]. Prior work has focused particularly on the properties of ordered two-dimensional (2D) oxides that form on Pd as well as Rh surfaces [21,99,100,102,117,118,128-130]. As far as we know, however, polycrystalline PdO develops under the oxidation conditions used in most previous studies [24]. Indeed, th e ability to generate well-defined PdO surfaces in UHV would greatly improve capabilities for interrogating the deta iled properties of PdO surfaces. In this Communication, we present experi mental data demonstrating that a high quality PdO(101) thin film can be grown on Pd(111) in UHV by oxidizing at 500 K using an oxygen atom beam. 80

PAGE 81

Experimental Methods Previous studies [86,121] provide details of the three-level UHV cham ber utilized for the present experiments. Briefly, th e 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 contact with a liquid nitrogen cooled reservoir. Sample cleaning consisted initially of Ar+ sputtering/annealing cycles, whil e subsequent cleaning involved routinely exposing the sample to an atomic oxygen beam, followed by flashing to 923 K. As discussed recently [131], our sample cleaning procedure minimizes loss of oxygen from the subsurface during subsequent oxygen uptake experiments. We oxidized the Pd(111) sample using oxygen atom beams generated in a differentiallypumped chamber connected to the main UHV analysis chamber [86, 121], and using dosing conditions described recently [131]. The sample geometry during dosing is selected to ensure uniform impingement of the oxygen atom beam across the Pd(111) surf ace. X-ray photoelectron spectra (XPS) were obtained using Mg K X-rays, and operating the he mispherical analyzer in retarding mode at a pass ener gy of 27 eV. We obtained low energy electron diffraction (LEED) images using a four-grid optics and a primary electron en ergy of 63 eV. Results and Discussion In this study, we oxidized Pd(111) at 500 K by exposing the sample to an oxygen atom beam for 10 minutes at an O atom incident flux of ~0.02 ML/s, where 1 ML is defined as the atomic density of 1.53 1015 cm-2 of bulk-terminated Pd( 111) [99]. According to O2 TPD, this ~15 ML O atom exposure produces a layer with 3.3 ML of oxygen atoms that desorb as O2 in a single sharp peak at 760 K [131]. The XPS Pd 3d5/2 spectra shown in Figur e 4-1 demonstrate that the oxygen atom exposure produces a PdO covered su rface. Specifically, in addition to the peak at 334.9 eV arising from metallic Pd, the Pd 3d5/2 spectrum obtained from the oxygen-exposed 81

PAGE 82

sample exhibits an intense peak at higher bindin g energy (Figure 4-1), in a range close to that reported elsewhere for PdO [22,23]. Figure 4-2 shows LEED images captured from both clean Pd(111) and the 3.3 ML PdO layer on Pd(111). The PdO layer was grown at 500 K, and then annealed at 675 K prior to the LEED experiment. The LEED image obtained from the PdO layer shows a distin ct pattern that is characterized by trios of spots and single spots in hexagonal arrangeme nts just within the original locations of the Pd(111) spots in addition to singl e spots arranged in a smaller hexagonal pattern. As indicated in Figure 4-2b, this LEED pattern consists of three equivalent rectangular unit cells, each rotated by 120 relative to one another. Since the LEED pattern obtained from the oxidized surface is unique compared with LEED patte rns reported for other oxygen phases on Pd(111), particularly the 2D Pd5O4 oxide [102], we conclude that the PdO layer consists of crystalline domains. These domains appear to be relativel y large considering that the diffraction spots associated with the PdO are sim ilar in size to the those obtained from Pd(111). Notice also that the diffraction spots from the Pd(111) substr ate are no longer visible in the LEED image acquired from the PdO layer. The LEED results therefore indicate that the 3.3 ML PdO layer forms a well-ordered thin film that completely covers the Pd(111) surf ace, and consequently hinders elastic scattering of electrons from the underlying metallic Pd. Although not presented here, we have obtained LEED images after preparing lower oxygen coverages on Pd(111) using an oxygen at om beam, and observe similar diffraction patterns as reported by other inve stigators. In particular, we observe the characteristic LEED pattern of the ordered 2D Pd5O4 oxide [102] at coverages above approximately 0.30 ML, with this pattern fading only graduall y as the oxygen coverage increases to about 1.3 ML. Notably, this behavior is consistent w ith recent evidence that oxygen atom s adsorb on top of the 2D oxide 82

PAGE 83

over this coverage range, befo re reacting appreciably with th e 2D oxide to form PdO nuclei [131]. Increasing the coverage fr om 2 to 2.5 ML causes the diffr action spots from the 2D oxide and Pd(111) substrate to fade signi ficantly as the spots characteristic of the PdO film begin to appear. Finally, only the diffracti on spots from the PdO film are evident above 2.8 ML, and the quality of this pattern (Figure 4-2) improves as the c overage increases to 3.3 ML at 500 K. Interestingly, we find that the oxidation of Pd(111) by atomic oxygen occurs at higher rates at 650 K compared with 500 K, but the resulting PdO layer appears to be rougher and disordered. The morphological differences that we observe suggest that PdO clusters gr ow at relatively high rates perpendicular to the surface during oxidation at 650 K, whereas the ratio of perpendicular to lateral growth rates is smaller at 500 K, resulting in smoother PdO films. Such a difference in oxide growth is consistent with early oxidation models discussed by Cabrera and Mott [132] and Fehlner [133]. These models pred ict that oxidation at low temp erature tends to produce smooth oxide films since ion transport th rough the growing film is governed by electric field gradients, whereas thermal diffusion contro ls the oxide growth kinetics at high temperatures, resulting generally in rough, polycrystalline films. Studies of the Pd(111) oxidation kinetics are currently underway in our laboratory to test these ideas. To help identify the specific structure and orie ntation of the 3.3 ML PdO film grown at 500 K, we examined LEED patterns simulated for seve ral low-index lattice planes of bulk PdO using the LEEDpat2 software. Bulk crystalline PdO ha s a tetragonal unit cell with dimensions of a=b=3.043 and c=5.336 [134,135]. In the ideal PdO crystal structure (Figure 4-3a), Pd atoms occupy the corner sites a nd center site of th e unit cell, while oxygen atoms reside in interstitial sites and form a pl anar structure around the body centered Pd atom [134,135]. Each oxygen atom is surrounded by four Pd atoms in a quasi-tetrahedral stru cture. In the LEED 83

PAGE 84

simulations, we treated the PdO as an incommensu rate layer on Pd(111) a nd are thereby able to observe the positions of the PdO diffraction spots relative to those of the Pd(111) substrate. As detailed below, we find that the simulated LEED pattern of bulk-terminat ed PdO(101) (Figure 43b) agrees very well with the pattern observe d experimentally (Figur e 4-2b), whereas LEED patterns for other low-index PdO la ttice planes differ significantly from the experimental image. It is worth noting that the 2D oxide on Pd(100) also arranges into th e PdO(101) structure, although only as a single layer [130] This similarity with the pres ent results is in teresting as it may indicate a general preference to ward forming the PdO(101) facet. The PdO(101) plane is highlighted in Figure 4-3a, and an illust ration of its un it cell is also shown superposed on the Pd(111) lattice in Fi gure 4-3c. The bulk-terminated PdO(101) surface has a rectangular unit cell with dimensions of a1 = 3.043 and a2 = 6.143 where a1 and a2 are parallel to the [010] and [-101] directions of the PdO crysta l, respectively. The unit cell dimensions may also be expressed as a1 = 1.106x and a2 = 2.233x where x = 2.75 is the lattice constant of Pd(111). Figure 4-3a shows that Pd atoms reside at the co rners and center of the PdO(101) unit cell, while oxygen atoms are located at the (0.500, 0.189) and (0.500, 0.811) positions within the unit cell, lying ~0.661 above and below the plane formed by the Pd atoms. We generated the LEED pattern shown in Fi gure 4-3b by orienting the PdO(101) unit cell with its shorter side pa rallel to the [-110] direc tion of the Pd(111) substrate (Figure 4-3c), and by displaying three rotational domains in the simula ted image. The resulting pattern reproduces the experimental LEED pattern very well, including the hexagonal arra ngement of single spots near the (0,0) position and the hexagonal arrangements of trios and single spots that are closer to the original Pd(111) spots. The simulated pattern also has three spots locate d just outside of each Pd(111) spot. A similar arrangement can be seen most clearly at the bottom of the experimental 84

PAGE 85

LEED pattern (Figure 4-2b). The excellent agreem ent between the simulated and experimental LEED patterns supports concluding that a high quality PdO(101) th in film forms on top of the Pd(111) substrate at the oxidation conditions em ployed, and is oriented such that the a1 lattice vector is parallel to the [-110] direction of the underlying Pd(111). It is interesting that the LEED pattern obtained from the PdO film matches the bulkterminated PdO(101) structure so closely considering that PdO(101) and Pd(111) have significant lattice mismatch, correspondi ng to +10.6% and +28.9% in the a1 and a2 directions, respectively. Commensurability between the PdO(101) layer and Pd(111) can be achieved, however, by minimally straining the PdO layer. Specifically, an expansion of the PdO(101) lattice by 0.46% and 3.4% in the a1 and a2 directions, respectively, produces commensurability between PdO(101) and Pd(111) over the smallest possible length scale without discernibly altering the LEED pattern. In this case, the commensurate PdO(101) unit cell has dimensions of 9a1 3a2, where a1 and a2 equal 1.11x and 2.31x, respectively. Forcing the commensurate PdO unit cell to be smaller requires much larger changes in the PdO lattice vectors, and produces clear differences between the simulate d and experimental LEED patterns. To conclude this Communication, we discuss the likely termination of the PdO(101) thin film. The three ideal terminations of PdO(101) are the polar O and Pd terminated surfaces, and the non-polar, stoichiometric PdO(101)-PdO su rface [134]. Assuming that maintaining charge neutrality is a dominant factor in determini ng the surface stability, th en the stoichiometric surface should be the most stable termination of PdO(101). Indeed, recent quantum chemical computations predict that the stoichiometr ic (101) termination of bulk PdO is more thermodynamically stable than the polar (101 ) terminations over a wide range of oxygen chemical potential [134]. In fact out of the eleven low-index surfaces examined, Rogal et al. 85

PAGE 86

[134] predict that the stoichiometric PdO(101) surface is the second most stable, with PdO(100)PdO being only slightly more favorable. As may be seen in Figure 4-4, the stoich iometric PdO(101) surface consists of closepacked rows of Pd atoms aligne d parallel to the [010] direction, where each row contains Pd atoms of a given coordination, alternating between three-fold or four-fold coordinated with oxygen atoms. Thus, half of the surface Pd atoms of stoichiometric PdO(101), which corresponds to a Pd atom density of 0.35 ML, ar e coordinatively unsaturat ed (cus), and hence likely to be more active toward binding adsorbed molecules that the four-fold coordinated Pd atoms. Consistent with this suggestion, we have recently determined from TPD a saturation coverage of 0.27 ML, on an O-atom basis, for O2 molecularly chemisorbed on the PdO thin film at 100 K, and a saturation coverage of 0.35 ML for CO [136]. The close agreement between these saturation coverages and the cus-Pd site density is compelling evidence that the PdO(101) thin film is terminated by the stoichiometric surface. Before summarizing, we note that low energy ion scattering spectra (LEISS) obtained from the PdO thin film exhibit an Oto-Pd peak ratio that is about twice that observed in LEISS spectra obtained from the Pd5O4 surface oxide [131]. While this might be taken as evidence that the PdO(101) surface is oxygen-termin ated, we believe that the O2 and CO TPD results [136] provide more direct and convinc ing evidence that the surface of the PdO( 101) thin film is stoichiometric. Most likely, shadowing and/or blocking has a significant influence on the scattered He+ intensities from PdO(101) at the co nditions used in the LEISS measurements. Summary We have demonstrated that a high quality PdO(101) thin film can be grown on Pd(111) in UHV by oxidizing the metal at 500 K using an oxygen atom beam, followed by annealing to 675 K. The PdO film aligns with the [-110] directions of the Pd(111) substrate, and appears to have a 86

PAGE 87

stoichiometric surface termination in which half of the Pd atoms (~0.35 ML) are coordinatively unsaturated. Considering the likeli hood that electric field gradient s play a dominant role in the oxide growth kinetics [133,137], the ability to efficiently oxidize at lo w substrate temperature may be a key factor for generating highly crysta lline PdO films on Pd surfaces. The capability of generating well-defined PdO(101) surfaces on Pd(111) should en able detailed UHV studies and hence provide opportunities to gain new understandi ng of the surface reactivity of PdO as well as the mechanisms for Pd(111) oxidation. 87

PAGE 88

339338337336335334333332 Pd (334.9eV) PdO (336.4eV)XPS Pd 3d5/2 Mg KO(g) + Pd(111) Ts=500K N(E) (a.u.)Binding Energy (eV) Pd(111) PdO/Pd(111) (3.3 ML) Figure 4-1. XPS Pd 3d5/2 spectra obtained from clean Pd(111 ) and a 3.3 ML PdO layer grown on Pd(111) at 500 K using an oxygen atom beam Reprinted with permission from H.H. Kan, J.F. Weaver, Surf. Sci. 602 (20 08) L53. Copyright 2008, Elsevier. Figure 4-2. LEED images obtaine d from (a) clean Pd(111) and (b) a 3.3 ML PdO layer grown on Pd(111) at 500 K, and then annealed to 675 K. The LEED images were obtained using a primary electron ener gy of 63 eV. A rectangular un it cell is shown in Figure 2b) and the open circle is located at the original positi on of one of the diffraction spots from the Pd(111) substrate. Reprin ted with permission from H.H. Kan, J.F. Weaver, Surf. Sci. 602 (2008) L53. Copyright 2008, Elsevier. 88

PAGE 89

a) a b c (101) b) c) Figure 4-3. LEED simulation resu lts and identification of PdO(101) (a) Schematic of the bulk PdO crystal structure with th e (101) plane highlighted. (b ) Simulated LEED pattern of PdO(101) as an incommensurate layer on Pd(111). Diffraction s pots corresponding to different rotational domains are color coded and the spots from th e (111) substrate are shown as open circles. The software pack age LEEDpat2 was used to generate the image. (c) Representation of the real space PdO(101) unit cell superposed on the Pd(111) unit cell. The a1 direction of the PdO(101) laye r is parallel w ith the [-110] direction of the Pd(111) subs trate. Reprinted with perm ission from H.H. Kan, J.F. Weaver, Surf. Sci. 602 (2008) L53. Copyright 2008, Elsevier. 89

PAGE 90

90 PdO(101)-PdO 3-fold Pd 3-fold O 4-fold O 4-fold Pd Figure 4-4. Model representations of the stoichiometric PdO(101) surface shown threedimensionally and in a top view. Pd and O atoms are represented as blue and orange spheres, respectively. The (101) unit cell is shown in the top view. The software package Chemcraft was used to generate these images. Reprinted with permission from H.H. Kan, J.F. Weaver, Surf. Sci. 602 (2008) L53. Copyr ight 2008, Elsevier.

PAGE 91

CHAPTER 5 ADSORPTION OF WATER ON A PALLADIUM OXI DE (101) THIN FILM: EVIDENCE OF AN ADSORBED HO-H2O COMPLEX Reproduced in part with permi ssion from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. C opyright 2009 American Chemical Society. Introduction Palladium oxide (PdO) is an excellent catalyst for the complete oxidation of CH4 [111,138,139] and CO [12-19] under oxygen-rich cond itions. Indeed, it is widely accepted that PdO formation is responsible for the exceptional performance of supported Pd catalysts toward the oxidation of CH4 in lean gas turbines. Several studies demonstrate that water can significantly affect the catalytic activity of PdO in oxidation reactions. In particular, reports show that water inhibits th e lean oxidation of CH4 [3,140-142], whereas wate r promotes CO oxidation on PdO [143,144]. For example, Oh and Hoflund have recently shown that CO oxidation is highly facile on hydrated PdO at 373 K, and that similar reaction rates are achieved on anhydrous PdO only above 473 K [1 44]. Unfortunately, the fundame ntal understanding of PdO surface chemistry has been limited, in part, by e xperimental difficulties in preparing well-defined PdO surfaces for model studies. In the present st udy, we investigated the adsorption of water on a PdO(101) thin film grown in ultrahigh vac uum (UHV), and focused on characterizing the chemisorbed states of water that develop on this surface. The adsorption of water on transition-metal (TM) oxide surfaces has been studied extensively in UHV [145-155]. In contrast to their weak physisorption on metal surfaces, water molecules tend to bind strongly on TM oxides by forming chemical bonds with coordinatively unsaturated (cus) metal cations of the surface. Strong water-oxide interactions can also cause water molecules to dissociate [146,147,149-154], typically through hydrogen transfer to oxygen anions of the surface [156]. While water molecu les dissociate extensively on some TM oxide 91

PAGE 92

surfaces, dissociation is limited to defect site s on others, which raises fundamental questions about the conditions and surface properties favori ng dissociative versus molecular chemisorption of water on TM oxides. Interestingly, recent studies show that intermolecular hydrogen bonding can facilitate water dissociation on TiO2 surfaces, resulting in adsorbed OH-H2O complexes [147,157]. These findings highlight the importa nce of both adsorbate-surface and adsorbateadsorbate interactions in determining the m echanisms for water chemisorption on TM oxide surfaces, and the resulting structure of the chemisorbed water layer. The stoichiometric PdO(101) surface is characterized by a rectangular unit cell and consists of four types of ions present in equal concentrations, na mely, threefold and fourfold Pd and O atoms, where the threefold atoms are c oordinatively unsaturated (cus) (Figure 1) [134,158]. The surface atoms arrange into rows that run parallel to the shorter latti ce direction of the unit cell, and each row contains only one t ype of surface atom. Thus, this surface structure places the more reactive cus sites in close proximity to one another, which may lead to enhanced bonding with adsorbed molecules. Furthermore, because Pd(II) is a large, electron-rich (4 d8) cation, the bonding of adsorbates on PdO is likely to involve consid erable covalent character that may differ from the bonding of many molecules on early TM oxides. Indeed, we have recently found that PdO(101) has a high affinity for binding O2 in molecular form, and that the majority of molecularly chemisorbed O2 species experience stronger bonding on PdO(101) than on Pd(111) [136]. In the same study, we observed that CO oxidation is facile on PdO(101) under UHV conditions. In contrast, early TM oxides are generally inactive for CO oxidation due to the strong metal-oxygen bonds that ch aracterize these materials. As far as we know, among the oxide surfaces that have b een investigated, only the RuO2(110) surface has a similarly high activity for CO oxidatio n as PdO(101), but RuO2(110) does not bind O2 molecules as strongly as 92

PAGE 93

PdO(101) [159-161]. These initial resu lts suggest that PdO(101) is li kely to interact strongly with other adsorbates as well. Since PdO(101) appears to be a uniquely reactive oxide-surface, further studies with PdO(101) may help to clarify key factors that influe nce the surface reactivity of TM oxides. In this study, we investigated the adsorption of water on PdO(101) using TPD measurements and DFT calculations, and present evidence that water dissociation is facile on PdO(101) at low coverage, but th at only molecular adsorption occurs, into both first and second layer states, as the water coverages increases be yond 50% of the cus-Pd site density. The DFT calculations predict that dissociati on is unfavorable for isolated H2O molecules on PdO(101), but is highly facile for water dimers and selectively produces HO-H2O complexes adsorbed along the cus-Pd rows. We also find that uptake into the first-layer chemisorbed states effectively ceases prior to saturation, and resumes only afte r the second-layer state appreciably populates. DFT suggests that strong orient ation-dependent interactions be tween adsorbed species create unfavorable sites along the cus-Pd rows that hinder ad sorption into the firs t layer when more than about 75% of the cus-Pd site s are occupied with adsorbed H2O and OH species. Experimental Methods Previous studies [86,121] provide details of the three-level UHV cham ber 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 contact with a liquid nitrogen cooled reservoir. A type K thermocouple sp ot-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 supply, supports maintaining or linearly ramping the sample temperature from 81 K to 1250 K. Initially, sample cleaning consisted of sputtering with 600 eV Ar+ ions at a surface temperature of 900 K, followed by 93

PAGE 94

annealing at 1100 K for several minutes. Subseq uent 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 previo usly [131], we limited the sample temperature to 923 K to maintain oxyge n-saturation in the subsurface reservoir, and thereby ensure reproducibility 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 Xray photoelectron spectroscopy (XPS), obtained sharp low ener gy 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 s ource (Oxford Scientific Instrument s) utilized to generate beams containing oxygen atoms for this study [121]. This system produces gaseous oxygen atoms by partially dissociating pure O2 (BOC gases 99.999%) continuously supplied to a small discharge chamber at the end of the plasma source. A ll oxygen atom beams used in this study were generated using an RF power of 120 W and an O2 flow rate that establis hes a pressure of 3 10-6 Torr in the first pumping stage of the beam cham ber. Under these conditio ns, we estimate that 20% of the inlet O2 dissociates in the plas ma. 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-c ollimating 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 lo wer bound of 0.02 ML/s at the sample surface, where 1 ML is defined as the atomic density of 1.53 1015 cm-2 of bulk-terminated Pd(111) [99]. 94

PAGE 95

We produce a PdO(101) thin film on Pd(111) by exposing the metal sample at 500 K to an ~15 ML dose of gaseous oxygen atoms supplied in a beam. This procedure generates a highquality PdO(101) film that has a stoichiometric surface termination, contains ~3.5 ML of oxygen atoms and is about 13 thick [158]. The structure of the PdO(101) surface is discussed in detail below. Filtered and deionized H2O was further purified by no less than three freeze-pump-thaw cycles and the resulting hi gh purity was confirmed using mass spectrometry. The H2O was delivered to the sample from a calibrated beam doser at an incident flux of approximately 7x10-3 ML/s and a substrate temperature of 85 K. We set the sample-to-doser distance to about 50 mm to ensure uniform impingement of the water ac ross the sample surface. After adsorbing water, we conducted TPD experiments by positioning the sample in front of the QMS and then heating at a constant rate of 1 K s-1 until the sample temperature r eached 923 K. The PdO thin films completely decompose when heated to 923 K so it was necessary to prepare fresh PdO films for each water adsorption experiment. We estimate d water coverages by scaling the integrated desorption spectra by an integrated TPD spectrum collected from a monola yer of water adsorbed on Pd(111) at 85 K, and assuming that the m onolayer on Pd(111) satu rates at 0.67 ML as reported previously [162]. Computational Methods All the DFT calculations in this pape r were performed using the Vienna ab initio simulation package (VASP) [163166]. We use the projector augmented wave (PAW) [167,168] pseudopotentials provided in the VASP database. Ca lculations have been done using the PerdewBurke-Ernzerhof PBE exchange-c orrelation functional [169]. Both the PBE and Perdew-Wang PW91 versions of the GGA functional have been shown to be sufficiently accurate to capture adsorption behavior of water on the surface [170-176]. We have tested several of our most stable minima using the PW91 functional and the values are reported in Table 1. The PW91 functional 95

PAGE 96

gives adsorption energies that are 3-4 kJ/mol higher than th e PBE values, but differences between adsorption sites are less than 1 kJ/mol. This level of accuracy is mo re than sufficient to conclude that the choice of the exchange-correlation functional will not affect the conclusions discussed in this paper. All the DFT values discu ssed in the rest of the paper will be from the PBE functional. A plane wave expansion with a cutoff of 400 eV was used and the total energy calculations are done using the conjugate-gradient method for electronic relaxations, accelerated using Fermi-level smearing with a Gaussian widt h of 0.1 eV. The positions of the atoms are relaxed until the forces on all unconstrained atom s are less than 0.03 eV/ While experimentally the PdO(101) film is grown on the Pd(111) surface, the oxide film is suff iciently thick (13 ) that we assume the Pd(111) substrate may be i gnored in our DFT calculations. Inclusion of the Pd(111) substrate would require large supercells due to the lattice mismatch between the oxide film and Pd(111) surface. We repr esent the PdO(101) film with f our layers (see Figure 1), which corresponds to a thickness of appr oximately 9 The bottom layer is fixed but all other atoms in the oxide film and the H2O adsorbate are allowed to relax. We use a vacuum spacing of 20 which is sufficient to minimize any spurious peri odic interactions in the surface normal direction. The PdO(101) film is obtained from relaxed PdO bulk structure, but the film is strained ( a = 3.057 b = 6.352 ) to match the reported experimental structure [158]. A 4 k-point mesh was used for the (1) unit cell and for larger surface unit cells the k-point mesh is scaled accordingly. We have performed tests to ensure that additional slab thickness and finer k-point meshes do not impact the results of the calcu lations. We use the climbing nudged elastic band (NEB) method [177-179] to find mi nimum energy pathways (MEP) and identify transition states (TS) for H2O dissociation. 96

PAGE 97

Experimental Results and Discussion Structure of the PdO(101) Thin Film on Pd(111) Figure 1 depicts the structure of the stoichio metric PdO(101) surface that we examined in this study. Bulk crystalline PdO has a tetragonal un it cell and consists of square planar units of Pd atoms fourfold coordinated with oxygen at oms. 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 [010] and [-101] directions of the PdO cr ystal, respectively. The PdO(101)-PdO surface consists of alternating rows of threefold or fourfold coordinated Pd or O atoms that run parallel to the a direction shown in Figure 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 molecu les. The threefold O atoms are also more accessible to the gas-phase than the fourfold O atoms. The areal density of each type of coordinatively-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 that the film contains ~3.5 ML of oxygen atoms, we estimate that the PdO(101) film on Pd(111) consists of about five layers and has a total thickness of ~13 In a prior study [158], we found that the PdO(101) structure aligns with th e close-packed directions of the Pd(111) substrate, and would expand by 0.46% 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 Water Desorption from Pd(111) and PdO(101) Figure 2 shows H2O TPD spectra obtained from Pd(111) and the PdO(101) film after adsorbing water at 85 K to generate coverages of 1.2 and 1.3 ML, respectively. The water TPD spectrum obtained from PdO(101) e xhibits three rate maxima loca ted at temperatures of 149, 197 97

PAGE 98

and 350 K. As discussed below, the broad, asymmetric feature between 300 and 350 K is actually composed of two di stinct peaks, labeled as 1 and 2. Water TPD spectra obtained from other oxides, including TiO2(110) and TiO2(100) [152,153], -Cr2O3(001) [149], and RuO2(110) [148], are qualitatively similar to that reported here for Pd O(101). In these systems, the desorption peaks which appear at increasing temperatures in the TPD spectra have been identified with a multilayer, a physisorbed or hydrogen-bonded second layer and chemisorbed states where the chemisorbed water evolves in two, closely-spaced peaks near 300 K. For TiO2(100) and -Cr2O3(001) the highest temperature feat ure arises from dissociatively chemisorbed H2O, whereas molecularly chemisorbed water produces the high temperature TPD peaks from RuO2(110). Since chemisorbed states of water should be more prevalent under realistic reaction conditions, we focused th e present study on determining whether the 1 and 2 states correspond to molecularly or dissociativ ely chemisorbed forms of water or both. As marked in Figure 2, the TPD results show that the 1 state and the 1 + 2 state each saturate at a coverage of about 0.35 ML, which corresponds to the surface density of each type of coordinatively-distinct Pd and O atom of PdO(101) (Figure 1). This finding is consistent with the expectation that chemisorbed H2O molecules ( 1 + 2) bind on the cus-Pd surface sites, and is similar to previous results of molecularly chemisorbed O2 and CO on PdO(101) [136]. It further suggests that H2O molecules in the 1 state arrange into a well-defined structure due to specific bonding interactions with the chemisorbed H2O and/or particular surface sites, such as the cus-O anions or four-fold Pd ca tions. In this case, the 1 state is better describe d as a second monolayer state rather than a second layer, which is sim ilar to the structure of the water layer on TiO2(110) [152]. 98

PAGE 99

A direct comparison of wate r desorption from PdO(101) an d Pd(111) is helpful for clarifying distinct characteristics of water adsorb ed on the oxide surface, particularly given that water adsorption on Pd(111) has been studied extensively and is well understood [162,180-183]. Briefly, water first adsorbs in a physisorbe d monolayer on Pd(111) that develops a twodimensional hydrogen bonding network which st abilizes the water layer [162]. The H2O monolayer on Pd(111) saturates at a coverage of 0.67 ML, and yields a sharp peak at 165 K in the TPD spectrum shown in Figure 2. Since the water monolayer on PdO(101) desorbs in peaks above 165 K, it is clear that H2O molecules access more strongly bound states on the oxide compared with the metal surface. It is particularly inte resting that the 1 peak from PdO(101) appears about 30 K higher than th e monolayer peak from Pd(111) since this observation provides clear evidence that water molecules in the 1 state experience stronger bonding interactions than water in the physisorbed monolayer on Pd(111). On Pd(111), a multilayer develops above 0.67 ML and gives rise to the sharp peak at 149 K in the TPD spectrum (Figure 2). The nearly pe rfect overlap of the multilayer peaks obtained from PdO(101) and Pd(111), as se en in Figure 2, demonstrate that the substrates and monolayer structures have little influence on the multilayer binding at a total coverage of about 1.2 ML, which is nearly equal to twice the monolayer coverages on both surfaces. Considering the significant differences in the H2O monolayer states on Pd(111) vers us PdO(101), it is interesting that the multilayers become so similar before th e equivalent of even one monolayer adsorbs on top of the first water layer. As shown below, the initial development of the multilayer feature differs considerably on PdO(101) compared w ith Pd(111), indicating that the monolayer structures do influence the initial bo nding of the multilayer on these surfaces. 99

PAGE 100

TPD of Water on PdO(101) as a Function of Coverage Figure 3 shows a series of TPD spectra obtained from PdO(101) as a function of the initial H2O coverage up to 1.1 ML. Water initi ally desorbs in a single peak ( 1) at 354 K that shifts slightly to lower temperature as the coverage increases from 0.12 to 0.18 ML. In addition, the trailing edge of the 1 peak is steeper than the leading e dge, making the peak appear asymmetric. The desorption rate reaches a st eady, non-zero base line after the 1 peak that persists up to about 700 K, at which point the rate abru ptly returns to the in itial baseline. This abrupt drop coincides with the PdO film beginning to thermally decom pose [131]. For comparison, the desorp tion rate returns immediately to the ini tial baseline after water desorb s from Pd(111). The slow H2O desorption process is probably associated with water chemisorbed on defects and/or water that penetrates into the PdO film. A bout 0.03 ML of water desorbs in the high temperature tail above 440 K. Also, we note that H2 desorption was immeasurable du ring TPD for all initial water coverages studied, and that less than 0.015 ML of 18O atoms incorporate into the substrate upon heating H2 18O-covered PdO(101) surfaces. These findings indicate that H2O dissociation, if it occurs, is reversible on the PdO(101) surface. Indeed, the 1 peak exhibits characteristics that are consistent with both first and second-order desorp tion, which makes it cha llenging to assign this feature to molecularly or dissociatively chemisorbed water without more detailed analysis or measurements. As the water coverage increases above 0.18 ML, the 1 peak intensifies only marginally and a shoulder, labeled as 2, emerges on the low temperature side of the peak. The shoulder continues to grow with increasing coverage, until the 1 + 2 state eventually saturates at high total coverages. Notice that the 2 feature first becomes evid ent at ~0.18 ML, which is approximately equal to one half of the cus-Pd density. Since the 1 peak appears to saturate at 100

PAGE 101

0.175 ML, we attempted to isolate the 2 peak by subtracting the TPD spectrum obtained at 0.175 ML from those obtained at higher coverage. As seen in Figure 4, the resulting difference spectra exhibit a distinct 2 peak with a maximum intensity at 318 K. Integration of the difference spectra from 240 to 400 K gives coverages ranging from 0.04 to 0.13 ML in the 2 state for the spectra shown in Figure 4b. The 2 peak maximum remains fixed with increasing total coverage, which is indicative of first-orde r desorption and hence a molecularl y chemisorbed state of water. However, the 2 peak is very broad and nearly symmetric about the maximum. These characteristics are inconsistent with a simple first order desorp tion process, and suggest that multiple adsorbate configurations contribute to the 2 desorption feature. Prior studies demonstrate that a two-peak structure near 300 K also develops in the water TPD spectra obtained from other oxides, but that this structure can have multiple origins. For example, the low and high temperature peaks have been identified, respectively, as molecularly and dissociatively chemisorbed states of water on both TiO2(100) [153] and -Cr2O3(001) [149] surfaces. However, Lobo and Conrad [148] re port that water chemisorption occurs only molecularly on RuO2(110), and that the emergence of a s econd high-temperature peak is caused by repulsive interactions among chemisorbed wa ter molecules. These differences among prior interpretations as well as the ambiguous behavior of the 1 peak, in part, motivated us to perform the TPD analysis and DFT calculations presented below. Figure 3 shows that the sharp 1 peak first appears at 0. 25 ML and intensifies with increasing total coverage, while the peak temperat ure remains fixed at 197 K. The invariance of the 1 peak temperature with increasing coverage is characteristic of first-order desorption, and suggests that intermolecular interactions onl y weakly influence the evolution of the 1 state. This observation provides further support for assigning the 1 state to H2O molecules that interact 101

PAGE 102

directly with Pd and O atoms of the PdO surf ace. The desorption rate remains at a small but steady value at temperatures between the 1 and 2 peak maxima, particularly at higher coverages of the 1 state. Tait et al. have observed a simila r plateau region between the monolayer and multilayer TPD peaks obtained from physisorbed al kane layers, and attribute this region to compression of the alkane monolayer [184,185]. Mo re specifically, those researchers suggest that increasing the coverage in the plateau region causes the chemical poten tial in the monolayer to increase until it reaches the value of the less stable second layer. The steady desorption rate seen in the water TPD spectra between about 225 and 275 K may have a similar origin. As shown below, the binding energy of water in the 2 state decreases significantly with increasing coverage in this state, and may decrease to su ch an extent that the chemical potential of the 2 state begins to match, and even exceed that in the 1 state. A sharp desorption peak for the water multilayer ( 2) begins to appear at a total coverage of 0.71 ML, which is near saturation of the 1 state. As the coverage increases, the 2 peak initially shifts to lower temperatures and then remains at a fixed temperature (Figure 3), coinciding with the multilayer peak obtained from Pd(111) (Figure 2). This desorption behavior reveals that the water multilayer initially expe riences stronger binding to the water monolayer on PdO(101) compared with that on Pd(111). However, intermolecular hydrogen bonding appears to dominate the binding within the multilayer on PdO(101) befo re even one layer of water molecules adsorbs into the multilayer. These initia l results are intriguing and certainly indicate distinct influences of the water monolayers on Pd(111) and PdO(101) on the initial development of the multilayers. We postpone a more detailed examination of the multilayer on PdO(101) for future work. 102

PAGE 103

Uptake of Water into Different Adsorbed States Figure 5 shows uptake curves for water adsorbin g into different states on PdO(101) as a function of the total coverage. We generated this plot by calculating ar eas under the TPD curves over temperature ranges from 85 to 170 K, 170 to 250 K and 250 to 440 K as a function of the total coverage, and equati ng these areas to the approximate coverages in the 2, 1 and 1 + 2 states, respectively. The approximately 0.03 ML of water that desorbs above 440 K is included in the total coverage scale shown on the x -axis. In Figure 5, we pres ent coverages in units of MLcus, where one MLcus equals the cus-Pd site density of 0.35 ML. This coverage scale helps to clarify relationships between the state populations and binding on specific surface sites, and will be used throughout the remainder of this paper. Figure 5 shows that water adsorb s exclusively in the 1 and 2 states as the coverage in itially increases beyond ~0.25 MLcus. Once the total coverage reaches about 0.75 MLcus, water begins to adsorb into the 1 state and uptake into the 1 and 2 states concurrently slows down, plateauing at a 1 + 2 coverage of about 0.75 MLcus. Since the 1 state saturates at a to tal coverage of 0.50 MLcus (Figure 3), the data reveals that uptake into the 2 state temporarily sta lls after about 0.25 MLcus of water adsorbs into this state. Water uptake occurs predominantly into the 1 state at total covera ges from 0.75 to 1.75 MLcus. About 0.75 MLcus of water molecules populate the 1 state at a total coverage of 1.75 MLcus, at which point uptake into the 2 state slowly resumes and the multilayer begins to populate as well. The total coverage in the 1 + 2 states reaches 0.90 MLcus at a total coverage of about 2.5 MLcus, at which point a significant quantity of multilayer water is present on the surface. The temporary halt in the uptake of water into the 2 state suggests that th e binding environment in this chemisorbed state becomes unfavorable at a total coverage ( 1 + 2) near ~0.75 MLcus, perhaps due to crowding along the rows of cus-Pd sites. Fu rthermore, the conti nued uptake into the 2 state only 103

PAGE 104

after the 1 state is highly populat ed suggests that the 1 state facilitates adsorption into the 2 state at high coverage. This behavior is interesting as it may indicate that the 1 state assists slow restructuring within the chemisorbed layer, which enables water molecule s to continue to access more strongly-bound chemisorbed states. Measurements of the water uptake on PdO(101) as a function of surface temperature would help to clarify the role of the second layer 1 state in mediating water chemisorption into the 2 state. Analysis of Water TPD Spectra We analyzed the H2O TPD spectra to estimate kinetic parameters for desorption, and thereby obtain further insights into the nature of the various adsorbed states. We specifically used the inversion-optimization procedure developed by Tait et al. [184,185] to analyze the 1, 2 and 1 TPD peaks. In this analysis, one assumes that the desorption rate rd can be described by the Polanyi-Wigner equation, RTEn deTr/)(),( (5-1) where n and E represent the adsorbate coverage, the de sorption order, and the pre-factor and the activation energy for desorption, respectively. The activation en ergy is treated as a coveragedependent quantity, while the pre-factor for desorp tion is treated as a cons tant. As pointed out by Tait et al. [184,185], this choice is somewh at arbitrary since the coupling between and E should allow the coverage dependen ce of the rate coefficient to be captured by either variable. In the first step of the analysis, one determines a coverage-dependent activat ion energy by inverting a TPD spectrum obtained at a give n initial coverage. Inversion of the Polanyi-Wigner equation gives the following expression for E ( ), )/),(ln()( n dTrRTE (5-2) 104

PAGE 105

The coverage as a function of temperature is obtained by numerically ev aluating the integral, T d TdT r dT dT d '' (5-3) at each point along a TPD spectrum, where represents the heating rate in this expression. The inversion is performed for an assumed desorption order and different values of the pre-factor and then the optimum pre-factor is identified as the value which minimizes the error between experimental and simulated TPD curves obt ained for different in itial coverages. We first analyzed the 1 peak using the inversion-optimization procedure, and considered cases of both first and second order desorption. We performed the analysis over the temperature range from 250 to 440 K to avoid the broad pl ateau region that extends to 700 K. For TPD spectra obtained at total initial coverages of 0.34 and 0.50 MLcus, we find that about 0.26 and 0.41 MLcus of water desorbs between 250 and 440 K so these values are taken as the initial coverages in the 1 peak analysis. For the analysis, we i nverted the TPD spectrum obtained at an initial coverage of 0.41 MLcus (Figure 6a) and determined th e optimum pre-factor based on spectra simulated for an initial coverage of 0.26 MLcus. As may be seen in Figure 6b, the 1 peak at 0.26 MLcus can be accurately reproduced using either first or second-orde r rate equations. We find that the error between the e xperimental and simulated curves is minimized for pre-factors of 1014 s-1 and 1014.2 s-1 ML-1 for n = 1 and 2, respectively. In Figure 6a, the E () curves for n = 1 and 2 are represented by sixth-order polynomials that were determined by fitting to the inverted TPD spectrum obtained from an initial coverage of 0.41 MLcus. For both curves, the activa tion energy increases weakly with decreasing coverage below 0.41 MLcus, but then increases sharply as the coverage decreases below about 0.07 MLcus. This abrupt increase is indicativ e of desorption from defect si tes [184,185] where the binding is 105

PAGE 106

stronger than on the sites containe d within the ideal Pd O(101) unit cell. The la rge increase in the activation energy at low coverage causes the desorption rate to decr ease rapidly above the 1 peak temperature. We therefore conclude that the steep trailing edge of the 1 peak results from an increasing contribution of water desorbing from defect sites, and is not necessarily indicative of first order desorption. At coverages above 0.07 MLcus, the activation energy exhibits a slightly stronger coverage dependence for the first order model compar ed with the second order model. For n = 1, E increases continuously from about 100 to 103 kJ/mol with decreasing coverage from 0.41 to 0.07 MLcus, whereas E remains nearly constant at ~99 kJ/mol over this sa me coverage range for the n = 2 model (Figure 6a). Within the context of the mean-field approximation, the behavior for n = 1 would imply that water molecules tend to encoun ter slightly less favorab le configurations as the coverage increases in the 1 state, whereas the nearly cove rage-independent activation energy determined for n = 2 suggests that the adsorbed water sp ecies bind in essentially the same local environment regardless of the coverage. Since eith er of these possibilitie s is plausible, it is difficult to unambiguously determin e the desorption order of the 1 state based on the inversion analysis alone. As shown below, DFT predicts that a specific dissociat ed state of water is strongly favored at low coverage This finding suggests that the n = 2 model represents a more appropriate description of the 1 TPD peak. We also used the inversion-optim ization procedure to analyze the 2 TPD peaks that were determined by deconvoluting TPD spectra obtained from total, initial coverages above 0.50 MLcus (Figure 4). Figu re 7a shows the E () curve that results from inverting the TPD spectrum obtained from an estimated 0.29 MLcus of water desorbing in the 2 peak, and assuming n = 1 and a pre-factor of 1013 s-1 for the inversion. The solid curve in Figure 7a represents a fit to the 106

PAGE 107

experimental data, and is given in units of kJ/mol by the equation, E (2) = 94.6 44.12 + 7.4exp(-2/0.04), which is the functional form suggested by Tait et al.[184,185]. For convenience, we used different functions to fit the inverted 1 and 2 peaks. As seen in Figure 7a, the activation energy is highly de pendent on the coverage in the 2 state, decreasing from about 93 to 82 kJ/mol as the coverage increases from 0.07 to 0.26 MLcus. The activation energy begins to increase more sharply with decr easing coverage below about 0.06 MLcus, which likely reflects small contributions from the 1 peak that persist after the peak deconvolution. The strong coverage-dependence of the ac tivation energy is responsible for the extreme width of the 2 peak (Figure 7b), and implies that water molecules en counter a multitude of local environments above 0.50 MLcus that become increasingly less favorable as the water coverage increases. As shown in Figure 4, the measured 2 peak temperature is invariant with changes in coverage. However, the E () equation obtained for a fi xed pre-factor causes the 2 peak temperature to shift significantly in spectra simulated at different init ial coverages. For example, using the E () relation shown in Figure 7a, we predict a 2 peak temperature of 335 K for an initial coverage of 0.11 MLcus, which is about 17 K higher than that observed experi mentally. The difference between the simulated and experimental 2 peaks may indicate that the pre-factor decreases with increasing coverage and compensates the decreasing activation ener gy. However, it seems more likely that the discrepancy reflects a failure of the mean-fie ld approximation to represent the coveragedependent desorption kinetics at total wate r coverages between about 0.50 and 1 MLcus. The inadequacy of the mean field approximation suggests that water adsorbs in different environments above 0.50 MLcus, with the binding en ergies spanning a broad range. If water chemisorbs on the cus-Pd sites, then population of the 2 feature would produce high local coverages (> 0.50 MLcus) of water along the cus-Pd rows. At such high coverages, intermolecular 107

PAGE 108

interactions can be expected to have a strong influence on th e water binding energies. The inversion analysis indi cates that, on average, adsorbates in the 2 state experience repulsive interactions with neighboring adsorbates, which causes the bi nding energy to decrease with increasing coverage. This behavior likely arises because the H2O-surface interactions are dominant over the adsorbate-adso rbate interactions. In this case the adsorbate rearrangements needed to achieve favorable adsorbate-adsorbate interactions are hindered because they would disrupt the adsorbate-surface interactions and cause a net decrease in the adsorbate binding energy. Thus, maintaining a strong interaction with the substrate likely ca uses the water species adsorbed in the 2 state to adopt unfavorable configurations relative to one another. Finally, we analyzed the 1 TPD peak using the inversio n-optimization procedure, and find that the 1 peak is well described as a first or der desorption proce ss with a coverageindependent activation energy. The analys is predicts an optimum fit for the 1 peak for a prefactor of 1013.7 s-1 and a desorption activation energy that remains constant at about 54.4 kJ/mol for most of the coverage range. The activati on energy vs. coverage cu rve derived from the 1 peak exhibits a sharp in crease below about 0.06 MLcus (not shown), which is similar to those seen in the inverted 1 and 2 peaks. This increase is associat ed with the small plateau region located between the 1 and 2 peak maxima. Overall, the analysis provides evidence that the 1 state corresponds to a molecularly adsorbed state of water that e xperiences negligible intermolecular interac tions. The nearly coverage-independe nt activation en ergy supports the conclusion that the second layer water is predomin antly stabilized by interactions with the PdO substrate rather than through intermolecular hydrogen bonding. 108

PAGE 109

Computational Results and Discussion We used DFT to investigate the adsorption of H2O into molecular and dissociated states on PdO(101) as function of the wate r coverage. To initially determ ine the favored adsorption site and geometry for isolated H2O in molecular and dissociated states, we modeled the surface using a (2) unit cell wher e the longer side coinci des with the [010] direc tion that parallels the cus-Pd rows (Figure 1). The most stable adsorp tion configurations dete rmined on the (2) unit cell were then evaluated us ing a (4) unit cell, whic h corresponds to 0.25 MLcus coverage. The values discussed in the rest of the paper will be in reference to the (4) unit cell. Calculations using a 2 unit cell predict energies that are with in 1 kJ/mol of those determined using the 4 unit cell, which confirms the expectation that adso rbates have negligible interaction between cusPd rows. Because adsorbate-adsorbate interact ions between cus-Pd rows make a only small contribution to the adsorption energies of wa ter on PdO(101), the 4 unit cell is a reasonable model for studying water adsorption on the PdO(101) surface and the va lues discussed in the rest of the paper are from calculati ons using the (4) unit cell. We initially examined several binding sites and orientations for an H2O molecule on the PdO(101) surface at a coverage of 0.25 MLcus. The calculations at a coverage of 0.25 MLcus predict that adsorption of an H2O molecule into its most favorable configuration is exothermic by 79.2 kJ/mol, where the adsorption energy is referenced to the bare PdO(101) surface and a gasphase H2O molecule. The predicted adsorption ener gy is much higher than binding energies typical of water physisorption, and is therefor e indicative of a relativ ely strong, localized bonding interaction between H2O and the PdO(101) surface (i.e., ch emisorption). In the preferred configuration (Figure 8a), the H2O molecule binds with its oxyge n atom located directly on top of a cus-Pd atom, and tilts so that the molecula r plane is nearly parallel to the surface. The molecule adopts an in-plane orientation that allows both hydrogen atoms to interact with oxygen 109

PAGE 110

anions of the PdO surface. One of the hydrogen at oms points directly toward a three-fold oxygen anion, while the other hydrogen atom lies in close proximity to a four-fold oxygen anion located on the opposite side of the cus-Pd atom. This on-top binding configuration is consistent with a donor-acceptor interaction whereby a lone pair that is main ly localized on the O atom of H2O overlaps with an empty 4 d -orbital of the Pd cation, thus forming an H2O-Pd bond. The in-plane orientation of the H2O molecule provides further stabilization through hydrogen bonds with the oxygen anions of the surface. We explored adso rption on several binding sites and find that water binding directly on the cus-Pd site is favored by at least 40 kJ/mol over other sites. The binding energy is also quite sensitive to the mo lecular orientation due to hydrogen bonding with surface oxygen atoms. These DFT results establish that H2O molecules chemisorb strongly on the cus-Pd sites, and that hydrogen bonding intera ctions with the surface anions strengthen the interaction of H2O with the PdO surface To examine dissociated states of wa ter on PdO(101), we assumed that an H2O molecule dissociates by transferring an H atom to a ne ighboring three-fold oxygen anion of the surface. This reaction produces an OH/OxH pair where the O and Ox symbols distinguish the oxygen atoms that originate from the adsorbed water molecule versus the PdO surface, respectively. For 0.25 MLcus coverage, the calculations predict an adsorption ener gy of 68.2 kJ/mol for an OH/OxH pair in its preferred configuration, which is nearly 11 kJ/mol endothermic compared with the H2O chemisorption state (Table 1). As seen in Figure 8b, the hydroxyl O-H bond lies nearly within the surface plane, and rotates away from the cusPd row so that the OH group can hydrogen bond with a fourfold oxygen anion of the surface. The hydrogen atom of the OxH group is also oriented slightly toward the ne ighboring OH group, which is indicative of hydrogen bonding. However, strong bonding with the surface may restrict the OxH group from achieving 110

PAGE 111

an optimal orientation for H-bondi ng. We calculate a di ssociation barrier of 20.8 kJ/mol for the isolated H2O molecule, and hence a barrier of 9.7 kJ/mol for the OH/OxH pair to transform to chemisorbed H2O. Although the dissociation barrier is relatively small, the computed energetics indicate that isolated H2O molecules prefer to chemisorb in tact rather than dissociating. Adsorbate Pair Formation Since hydrogen bonding is generally important in determining the properties of adsorbed water, we examined the energetics associated with various clusters of H2O and OH on PdO(101). Table 1 summarizes the energies predicted for th e different clusters we considered, and Figure 9 illustrates the specific configurations of adsorbate pairs investigated at 0.5 MLcus coverage. Firstly, we predict a moderate driving force for pairing of H2O molecules, whereas OH-OH pair formation is slightly unfavorable. The computed adsorption energies ar e 66.7 and 82.0 kJ/mol per molecule for an OH-OH versus an H2O-H2O pair, respectively, which indicates that OH pairing is slightly less stable than isolated OH groups while H2O pairing is stabilized by 5.6 kJ/mol. The key finding from the calculations is that an HO-H2O pair, where the H2O molecule acts as the proton donor (F igure 9c), is highly favored over th e other pair configurations. The adsorption energy of the HO-H2O pair is 96.1 kJ/mol per molecule. Thus, formation of the HOH2O pair from isolated H2O and OH groups stabilizes these ad sorbates by 45 kJ/mol. The total adsorption energy of the HO-H2O pair is also 28.2 kJ/mol higher than that of an H2O-H2O pair. The stabilization afforded by HO-H2O pair formation greatly fac ilitates the dissociation of H2O. We predict a barrier of only 2.4 kJ/mol for an H2O-H2O pair to transform to the HO-H2O pair through H atom transfer to an Ox site. Since the pathway for this re action is similar to that for an isolated H2O molecule, for which we find a barrier of 20.8 kJ/mol, the low barrier suggests that the HO-H2O interaction stabilizes the transition state for H2O dissociation in an H2O-H2O pair on 111

PAGE 112

PdO(101), making dissociation highly facile. Interestingly, the interaction between adsorbed H2O and OH has a strong orientation de pendence. The binding energy for an H2O-HO pair, where the OH group acts as the proton donor (Figure 9d), is only 70.3 kJ/mol per molecule significantly less than the energy for the HO-H2O pair. In fact, H2O-HO pair formation is actually destabilizing by a few kJ/mol relative to the isolated species. The DFT results indicate that the dissociation of an H2O molecule on PdO(101) is favorable when the molecule pairs with another H2O molecule, and produces the stable HO-H2O complex. The experimental TPD results are consistent with this prediction since desorption from the 1 state can be accurately repr oduced with a second-order kinetic model, and a nearly coverage-independent activation energy. The ex perimentally-determined activation energy of ~99 kJ/mol for the (assumed) second-order 1 peak (Figure 6a) is remark ably close to the binding energy of the HO-H2O complex (96.1 kJ/mol per molecule) determined with DFT. Indeed, this agreement suggests that the 1 state observed in TPD corresponds to recombinative desorption of the HO-H2O complex. However, it is important to as certain whether the r ecombination reaction HO-H2O + OxH H2O-H2O + Ox is a second-order process and leads to the prompt desorption of H2O during TPD. Simple second-order kinetics occurs when the reacting adsorbates randomly populate surface sites because, in this case the probability of finding an ad sorbate pair, and therefore the rate of recombination, is dire ctly proportional to the product of the adsorbate coverages. Random population of surface sites occurs when the adsorb ates interact only weakly, and therefore do not have a strong propensity to form pairs or remain separated. For water on PdO(101), the recombining partners are the HO-H2O complex and the H atom of the OxH group. From DFT, we find that moving the OxH group several sites away from the HO-H2O complex reduces the 112

PAGE 113

adsorption energy by only 6.7 kJ/mol. This weak interaction implies that OxH groups will populate Ox sites nearly randomly and hence that the H + OH recombination rate will approximately follow second order kinetics. In addition, since the barrier for an H2O-H2O pair to dissociate to the HO-H2O complex is very small (< 2.5 kJ/mol), the concentration of H2O-H2O pairs will also be very small at the temperat ures studied. Consequent ly, H + OH recombination during heating can be expected to result in either prompt H2O desorption, or a rapid return to the HO-H2O state when the energy is in sufficient for desorption. Before presenting the final DFT resu lts, we point out that mixed OH/H2O layers have been observed previously on other surf aces, and are found to enhance th e stability of the adsorbed layer in these cases as well. For example, studies of H2O on Pt(111) reveal th at an electrostatic interaction causes the HO(acceptor)-H2O(donor) hydrogen bond to be stronger than the hydrogen bond between H2O molecules [172,186]. In this case, charge exchange with the surface results in net positive and negative charges on the adsorbed H2O and OH species, respectively, making proton donation from H2O to OH a strongly attractive and high ly directional in teraction. Recent studies show that HO-H2O complexes also form on TiO2 surfaces. Similar to our findings on PdO(101), a study by Beck et al. [147] demonstrates that formation of HO-H2O complexes is favored over pure H2O and OH layers on TiO2(011). Ketteler et al. [157] have also shown that OH groups facilitate H2O dissociation on TiO2(110), leading to H2O-HO complexes where the OH group acts as the proton donor. Th ese findings highlight the sensitivity of the adsorbateadsorbate interactions on the loca l bonding environment of water on TiO2 surfaces. For the HOH2O complex on PdO(101), we find fro m DFT that the adsorbed H2O and OH species have net charges of +0.1e and -0.51e, respectively. Thus, a favorable electrostatic interaction appears to enhance the stability of the HO(acceptor)-H2O(donor) hydrogen bond on PdO(101), which is 113

PAGE 114

similar to the predictions of H2O and OH interactions on Pt(111) [186]. We are continuing to examine the bonding of H2O and other species on PdO(101). Becau se Pd(II) is a large, electronrich cation (4 d8), it is likely that adsorbate bonding on PdO surfaces will involve considerable covalent character that may be distinct from bonding on early tr ansition metal oxides such as TiO2. High Coverage Configurations: Trimers and Saturation The experimental TPD results provide evidence that H2O begins to populate a new chemisorbed state(s) above 0.50 MLcus, and that uptake into this state slows down above about 0.75 MLcus. We investigated several H2O and OH configurati ons at 0.75 and 1 MLcus using DFT in an attempt to clarify these e xperimental observations. As seen in Table 1, DFT predicts that the HO-H2O-H2O trimer is preferred over several other trimer configurations at 0.75 MLcus coverage. Formation of the HO-H2O-H2O trimer is nearly thermoneutral relative to an isolated H2O molecule and an HO-H2O complex. In contrast, formation of the H2O-HO-H2O complex from H2O and the HO-H2O species is destabilizing by 11. 4 kJ/mol. Both dissociation and association of species in the HO-H2O-H2O trimer is also unfavorable (Table 1). While the HOH2O-H2O trimer is energetically preferred, the HO-H2O-H2O and H2O-HO-H2O trimers differ in energy by only 11.4 kJ/mol. Thus, the less stable H2O-HO-H2O trimer can be expected to coexist with the HO-H2O-H2O trimer on PdO(101). Interestingly, the inversion analysis of the 2 TPD peak indicates that the desorption activation en ergy decreases by about 11 kJ/mol as the total coverage increases from 0.50 to 0.75 MLcus, which agrees well with the energy difference between the most stable trimers predicted by DFT. The DFT calculations indicate that desorption of an H2O molecule from the H2O-HO-H2O and HO-H2O-H2O trimers is endothermic by 67.3 and 78.6 kJ/mol, respectively. Fo r comparison, assuming first orde r desorption and a pre-factor 114

PAGE 115

of 1013 s-1, the inversion analysis predicts that the activation energy for desorption into the 2 peak varies from 82 to 93 kJ/mol with decreasing coverage. The agreement between the absolute energies determined from computation and experime nt improves if a lower pre-factor is used in the inversion analysis. The directionality of the H2O-OH interaction appears to be the key factor governing the preferred trimer arrangements. For example, the HO-H2O-H2O trimer avoids the destabilizing H2O-HO configuration, while also maintaining the favorable HO-H2O linkage. Interestingly, however, formation of the HO-H2O-H2O trimer from H2O and the HO-H2O complex is slightly destabilizing by 0.5 kJ/mol, whereas H2O-H2O pair formation is exothermic by 5.6 kJ/mol. Thus, H2O attachment to the HO-H2O complex appears to produce a weaker H2O-H2O interaction than in the corresponding pair a nd/or destabilizes the HO-H2O interaction. The net result is that HOH2O-H2O trimer formation is slightly endothermic, in contrast to expecta tions based only on the pair interactions. The directionality of the adsorb ate interactions also dictates the preferred configurations at water saturation of the cus-Pd sites. For example, DFT predicts that the 2(HO-H2O) configuration is favored at saturation, and that the HO-H2O-2H2O configuration is the second most stable. However, several c onfigurations can be present on the actual PdO(101) surface at high H2O coverages that cannot be adequately captu red with the (4) surface model. Figure 9c shows the HO-H2O complex in only one of two equiva lent orientations, where the second orientation is represented by the mirror image taken across a plane of symmetry orthogonal to the direction of the cus-Pd rows. This configuration may be represented as OH2-OH. Because the HO-H2O and OH2-OH configurations are energetically degenerate, they would be present on the surface in equal concentrations. 115

PAGE 116

As the water coverage increases, H2O molecules will more frequently encounter sites located between configurationally-degenerate di mers and trimers, where they are likely to experience an unfavorable environment due to the strong orientati onal dependence of the adsorbate-adsorbate interactions. As a result, H2O molecules will likely avoid adsorption into sites between configurationally-degenerate co mplexes. The measured uptake of water into various states on PdO(101) is consis tent with this interp retation. As seen in Figure 5, the rate of uptake into the high-temperature chemisorbed states abruptly decreases at a total coverage of about 0.75 MLcus, and water molecules begin to populate the more weakly bound 1 state instead. In fact, the water concentration in the 1 and 2 states remains essentially constant until the 1 state is near saturation, and then only gradually increases to ~0.91 MLcus as the multilayer starts to develop. We therefore conc lude that continued adsorpti on onto cus-Pd sites becomes unfavorable at coverages near 0.75 MLcus due to destabilizing inter actions at sites located between configurationally-degene rate adsorbate-complexes. The strong coverage-dependence of the 2 desorption activation energy and the plateau region leading up to the 2 peak are also consistent with th e development of multiple, low-energy adsorbate configurations on the cusPd rows at coverages above 0.50 MLcus. In particular, the significant decrease in the activation en ergy with increasing coverage in the 2 state suggests that low energy configurations become increasingly more populated as the coverage increases. The chemical potential of certain configurations on the cus-Pd rows may even approach the chemical potential in the second layer (1 peak), and contribute to the steady desorption observed between about 225 and 275 K in the TPD spectra. Thus, the desorption that occurs at temperatures between the 1 and 2 peak maxima likely represent molecu les desorbing from a distribution of unfavorable local environments, while desorption closer to the 2 peak temperature arises from 116

PAGE 117

more favorable adsorbate configurations that ex ist once the coverage decreases to values near 0.50 MLcus. The DFT calculations provide a description of water adsorption onto PdO(101) that is consistent with our experimental results. Figure 10 summarizes key steps in the adsorption of water on PdO(101) up to the formation of adsorbed trimers, but we emph asize that aggregates other than the trimers depicted in Figure 10 are likely to form at high water coverages. Firstly, DFT predicts that isolated H2O molecules prefer to remain in molecular form, but that dissociation becomes facile when H2O molecules form pairs due to the high stability of the resulting HO-H2O complex. Our experimental TPD results show that water desorbs exclusively in the 1 state between coverages of 0.25 and 0.50 MLcus, and that desorption from this state is well-described as a second-order process with a nearly constant activa tion energy, which is consistent with the existence of adsorbed HO-H2O complexes. The excellent agreement between the experimentally-determined a nd DFT-derived activation energi es for desorption (~97 kJ/mol) provide further support for concluding that HO-H2O complexes predominantly cover the PdO(101) surface below 0.50 MLcus. Although we did not examine coverages below 0.25 MLcus, it is likely that HO-H2O complexes form in this low coverage range as well since adsorbed H2O molecules should be fairly mobile. For coverages above 0.50 MLcus, DFT indicates that H2O molecules begin to chemisorb intact and that HO-H2O-H2O trimer formation is nearly thermoneutral relative to an isolated H2O molecule and an HO-H2O complex. Other trimer configurations are less favorable, but would develop as the c overage of chemisorbed water increases, which is consistent with the large d ecrease in the water binding energy as the coverage in the 2 state increases. Finally, the DFT predic tion of a strongly or ientation-dependent interaction between adsorbed species offers a pl ausible interpretation of the slow water uptake 117

PAGE 118

into the 1 and 2 chemisorbed states observed above 0.75 MLcus. The directionality of the interactions suggests that unfa vorable binding environments ar e prevalent betw een adsorbate complexes at high coverages, and that water molecules tend to avoid these configurations. Additional computational studies are needed to clarify the bonding environments present between configurationally-degenerate dime rs and trimers and the nature of the 1 state. Summary We investigated the adsorption of wate r onto a PdO(101) thin film using TPD measurements and DFT calculations. TPD spectra obtained from high water coverages exhibit sharp peaks at 149 and 197 K that arise from water desorbing from a multilayer and a second layer state, respectively. Water in the second layer state appears to inter act directly with the PdO(101) surface, likely through hydrogen bondi ng with surface oxygen anions. We also observe desorption peaks at 318 and 350 K that or iginate from chemisorbed water, with each state accommodating nearly 0.50 MLcus of water molecules. Analysis of the TPD spectra shows that the higher temperature 1 peak can be accurately described using either a first or second order desorption model, with re latively weak or negligible coverage-depe ndence of the desorption activation energy, re spectively. In contrast, th e activation energy decreases significantly with increasing coverage in the 2 state, resulting in a broad TPD peak. Our experiments also show that water first saturates the 1 state before populating the more weakly bound 2 state. However, uptake into the 2 state effectively ceases at a total coverage of about 0.75 MLcus, and resumes only after the second-layer st ate is highly populated This observation may indicate that the second-layer molecules facilitate adsorption into the molecularlychemisorbed state at high coverages. 118

PAGE 119

Our DFT calculations predict that dissoc iation is unfavorable for isolated H2O molecules, but is highly facile for water dimers and selectively produc es hydrogen-bonded HO-H2O dimers adsorbed along the cus-Pd rows, where the H2O molecule acts as the proton donor in the dimer. DFT suggests that an electrostatic interac tion significantly stabilizes the HO(acceptor)H2O(donor) complex, and also serves to destabilize the opposit ely-configured H2O-HO dimer. In agreement with the experimental re sults, DFT predicts that the HO-H2O complex forms up to a total water coverage of 0.50 MLcus, and has a binding energy of 96.1 kJ/mol per H2O molecule. Above 0.50 MLcus, DFT predicts that water preferentially chemisorbs in molecular form on the cus-Pd rows, and can bind in local aggregates with different energies Formation of the HO-H2OH2O trimer is nearly thermoneutra l relative to an isolated H2O molecule and the HO-H2O complex, whereas formation of the H2O-HO-H2O trimer is slightly endothermic. An increase in the relative coverage of less stab le adsorbate clusters can explain the large decrease in the water binding energy as the total coverage in the 2 state increases. Finally, the DFT calculations suggest that strong orientationdependent interactions create unfavorable binding sites between dimers and trimers adsorbed along the cus-Pd ro ws, and that water molecules will tend avoid these sites at water coverages above about 0.75 MLcus. Overall, the DFT and experimental results provide a consistent description of water adso rption onto the PdO(101) su rface. The results of this study, particularly the evidence for selective HO-H2O dimer formation, provide new insights for modeling the oxidation of CH4, CO and other molecules on PdO surfaces in the presence of co-adsorbed water. 119

PAGE 120

a b (a) (b) cusPd cusO Figure 5-1. (a) Top and (b) side view of the PdO(101) thin film structure. The orange and blue atoms represent O and Pd atoms respectively. Rows of threefold-coordinated (cus) Pd and O atoms are indicated. The a and b directions correspond to the [010] and [-101] crystallographic directions of PdO. Reproduced in part with permission from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 120

PAGE 121

100150200250300350400450500 0 2 4 6 8 10 ~0.35 ML ~0.35 ML2112H2O TPD Ts = 85 K Heating Rate = 1 K/sDesorption Rate (10-2 ML/s)Temperature (K) 1.2 ML H2O/Pd(111) 1.3 ML H2O/PdO(101) Figure 5-2. H2O TPD spectra obtained from Pd(111) and PdO(101) after generating initial water coverages of 1.2 and 1.3 ML, respectively, at a substrate temperature of 85 K. The TPD spectra were obtained using a constant heating rate of 1 K/s. Reproduced in part with permission from H.H. Kan, R.J. Colmye r, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 121

PAGE 122

0.0 0.1 0.2 0.3 0.4 0.5 100150200250300350400450500 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 a 121[H2O]0 (ML) 0.12 0.18 0.25 0.32 0.79 0.37b Temperature (K)2112H2O/PdO(101)/Pd(111) Ts = 85 KDesorption Rate (10-2 ML/s)[H2O]0 (ML) 0.37 0.52 0.71 0.95 1.13 Figure 5-3. H2O TPD spectra obtained from PdO(101) at initial water c overages between a) 0.12 ML and 0.37 ML and b) 0.37 ML and 1.13 ML prepared by dosing water with the sample held at 85 K. The TPD spectra were obtained using a consta nt heating rate of 1 K/s. Reproduced in part with permi ssion from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 122

PAGE 123

0.000 0.001 0.002 0.003 0.004 200225250275300325350375400 0.000 0.001 0.002 0.003 a [H2O]0=0.52 ML [H2O]0=0.175 ML "sat. 1" Difference b2 Difference Spectra2 (ML) 0.04 0.10 0.13Desorption Rate (ML/s)Temperature (K) Figure 5-4. a) Difference H2O TPD spectrum (black dotted line ) obtained by subtracting a TPD spectrum obtained af ter saturating the 1 peak with 0.175 ML of water from a TPD spectrum (red dotted line) obtained from an in itial water coverage of 0.52 ML (green solid line). The difference sp ectrum exhibits a distinct 2 peak. b) Difference TPD spectra obtained from initial water coverages of 0.21 ML, 0.32 ML, and 1.64 ML. The corresponding coverages in the 2 state are estimated as 0.04, 0.10 and 0.13 ML. Reproduced in part with permi ssion from H.H. Kan, R.J. Co lmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 Am erican Chemical Society. 123

PAGE 124

0.00.51.01.52.02.53.03.54.0 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 H2O / PdO(101) Ts = 85 K State Coverages (MLcus)Total Coverage (MLcus) 1+2 1 2 Figure 5-5. Uptake of H2O into various adsorbed states plotte d as coverage within a state versus the total H2O coverage, where coverages ar e expressed in units of MLcus. Coverages within the multilayer ( 2), second layer (1) and first layer ( 1 + 2) states were estimated as the areas under the TPD trac es over temperature ranges from 85 to 170 K, 170 to 250 K and 250 to 440 K, respectively. The plot reveals that uptake into the first layer effectively ceases at about 0.75 MLcus, and resumes only after the secondlayer state is nearly saturated. Reproduced in part with permission from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weav er, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 124

PAGE 125

0.000.050.100.150.200.250.300.35 90 95 100 105 110 115 120 125 (a)1 = 1014.2 s-1 MLcus -1 ( n = 2)1 = 1014 s-1 ( n = 1)TPD inversion1 = 0.41 MLcussat. 1 peakE (kJ/mol) (MLcus) 1 300 350 400 0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030 (b) n = 1 n = 2H2O TPD 1 peak 1 = 0.26 MLcusrate (ML/s)Temperature (K) Figure 5-6. TPD inve rsion analysis of 1 a) Desorption activation energy as a function of coverage in the 1 peak determined by inverting the TPD spectrum obtained at an initial coverage of 1 = 0.41 MLcus for both first and second order desorption models. The E (1) curves were obtained with pre-factor s that give the best agreement between experimental and simulated TPD spect ra for an initial coverage of 1 = 0.26 MLcus. The curves shown are sixth-order polynomial s that were obtained by fitting to the inverted TPD spectrum. b) Comparison of experimental and simulated TPD spectra obtained at an initial coverage of 1 = 0.26 MLcus and using both first and second order desorption models afte r determining the optimum pr e-factors and corresponding E (1) relations. Reproduced in pa rt with permission from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 125

PAGE 126

0.000.050.100.150.200.250.30 70 80 90 100 110 120 2 = 1013 s-1(a)TPD inversion2 peak2 = 0.29 MLcus E (kJ/mol)2 (MLcus) 250 300 350 400 450 0.00 0.05 0.10 0.15 0.20 0.25 (b) n = 1 E = E (2) 2 = 1013 s-1 H2O / PdO(101)2 TPD peak 2 = 0.29 MLcusrate (10-2 ML/s)Temperature (K) Figure 5-7. TPD inve rsion analysis of 2 a) Desorption activation energy as a function of coverage in the 2 peak determined by inverting the deconvoluted TPD spectrum obtained at an initial coverage of 2 = 0.29 MLcus. We used a first order desorption model and a pre-factor of 1013 s-1 in the TPD inversion. The symbols represent experimental data and the solid curve represents a function that was fit to the data. b) Comparison of experimental and simulate d TPD spectra obtained at an initial coverage of 2 = 0.29 MLcus and using the first-order pa rameters depicted in a). Reproduced in part with permi ssion from H.H. Kan, R.J. Co lmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 Am erican Chemical Society. 126

PAGE 127

(a) (b) Figure 5-8. Top and side view of the most stable DFT identified configurations for (a) H2O (79.2 KJ/mol) and (b) dissociated H2O (68.2 KJ/mol) at 0.25 MLcus coverage. Reproduced in part with permi ssion from H.H. Kan, R.J. Co lmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 Am erican Chemical Society. 127

PAGE 128

(a) (b) (c) (d) Figure 5-9. Top view of (a) H2O-H2O (82 KJ/mol) (b) OH-OH (66.7 KJ/mol) (c) HO-H2O (96.1 KJ/mol) (d) H2O-HO (70.3 KJ/mol) pairs at 0.50 MLcus coverage. Reproduced in part with permission from H.H. Kan, R.J. Colmye r, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 128

PAGE 129

-60 -50 -40 -30 -20 -10 0 10 20 30 40 *Minimum energy path H2O / PdO(101)(0.0) (-21.9) (-33.3) (-33.8) (-3.2) (-5.6) H2O-HO-H2O(ad) HO-H2O-H2O (ad) H2O (ad) + HO-H2O (ad) H2O (ad) + H2O-H2O (ad) 3 H2O (ad)Binding Energy (kJ/mol) Figure 5-10. Energy pathway for the formation of adsorbed trimers on PdO(101). The zero of energy is taken as three adsorbed H2O molecules that are isol ated from one another, where an isolated water molecule is represented as H2O(ad). The transition state for HO-H2O dimer formation is labeled with an as terisk and the numbers in parenthesis are energies in kJ/mol referenced to the zero defined in the diagram. Reproduced in part with permission from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society. 129

PAGE 130

130 Table 5-1. Adsorption energies ( Eads) of H2O and OH species on PdO(101) predicted with DFT using the GGA-PBE functional for cove rages of 0.25, 0.50, 0.75 and 1 MLcus. The values obtained using the GGA-PW91 functional is given in parentheses. The reference energy is taken as the bare PdO(101) surface and N -gaseous H2O molecules where N equals the number of adsorbates. Note that an OxH group is associated with each adsorbed OH group, but is only listed once in the table in or der to simplify the symbolic representations of the adsorbates. The favored configurations are shown in bold for each coverage. Species Eads (kJ/mol) Coverage (MLcus) Monomers OH (OxH) 68.2 (71.9) 0.25 H2O 79.2 (82.8) 0.25 Dimers HO-HO 66.7 (70.6) 0.50 H2O/H2O 82.0 (85.8) 0.50 HO-H2O 96.1 (100.2) 0.50 H2O-HO 70.3 (74.0) 0.50 Trimers 3(H2O) 80.8 (84.7) 0.75 HO-H2O-H2O 90.3 (94.3) 0.75 H2O-HO-H2O 86.5 (90.4) 0.75 HO-H2O-HO 82.8 (86.8) 0.75 Saturation 4OH 62.5 (66.5) 1 4(H2O) 81.3 (85.2) 1 2(H2O)H2O-HO 85.8 (89.8) 1 2(HO-H2O) 90.3 (94.4) 1 Reprinted with permission from H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C, 113 (2009) 1495. Copyright 2009 American Chemical Society.

PAGE 131

CHAPTER 6 PALLADIUM (111) OXIDATI ON KINETICS: TEMPERATURE AND OXYGEN PRESSURE EFFECTS ON THIN FILM GROWTH, THICKNESS, AND MORPHOLOGY Introduction Palladium is an excellent heterogeneous oxi dation catalyst for oxygen rich applications such as the oxidation of methane [1-11] and carbon monoxide [1219]. Unfortunately, a fundamental understanding of palladium catalyzed oxidation is still insufficient, requiring research and development of more efficient oxida tion catalysts to proceed rather inefficiently through a trail-and-error process. Fundamental research of pa lladium catalyzed oxidation may therefore potentially aid in the development of more efficient oxida tion catalysts as well as assist in the optimization of existing palladium catalyzed applications. However, one of the prevailing challenges with characterizing oxidation catalysts involves the identification of the actual catalytically active substrate since oxidation of th e substrate itself may create new substrates due to the formation of several stable oxide surfaces and phases, all of which may exhibit varying degrees of activity. In addition, characteriza tion of the substrate oxidation mechanism also provides useful information from an applications poi nt of view since the ki netically relevant step for catalytic oxidation may in fact be substrate oxidation if the catalyzed reaction proceeds via the Mars-van Krevelen reaction. Therefore, st udies of the oxidation mechanisms for the most abundant low miller index planes of palladium in addition to the oxygen/oxide phases that develop during oxidation will provide an indisp ensible tool for the development of nextgeneration heterogeneou s oxidation catalysts. The initial oxidation of Pd(111) up to the fo rmation of a two-dimensional surface oxide is well characterized [21-23,25,99-114]. Briefly, oxygen atoms first occupy a chemisorbed 2x2 adlayer which saturates at a to tal oxygen coverage of 0.25 ML, where 1 ML is defined as the atomic density of 1.53 1015 cm-2 of bulk-terminated Pd(111) [99] Further exposure leads to the 131

PAGE 132

formation of a metastable compressed chemisorbed oxygen adlayer [21], followed by a substrate reconstruction leading to a variet y of possible two-dimensional s urface oxides, all of which have an oxygen coverage of around 0.7 ML and ar e metastable, with the exception of the stable Pd5O4 surface oxide [128]. However, oxidation beyond the completed surface oxide is poorly understood due to difficulties with conducting high precision surface characterization techniques under (or immediately following) the molecular oxygen pressures and substrate temperatures necessary for bulk palladium oxidation. High precision surface characteriza tion of this coverage range will be invaluable to catal yst designers since it has been established that intermediate oxide coverages between the surface and bulk oxide ex hibit the highest cataly tic oxidation activities [10,26]. Recent studies have provided substantial insi ght into the initial bul k oxide formation of Pd(111) [22-25]. Palladium metal is known to have a strong affinity towa rds subsurface oxide formation [103], and utilizing ambient pre ssure synchrotron based X-ray photoelectron spectroscopy (XPS), Ketteler et al. have suggested that the subsurface oxide acts as the precursor to extrinsic bulk oxide formation [23]. Furt hermore, by utilizing hi gh pressure cells in conjunction with temperature pr ogrammed desorption (TPD) and scanning tunneling microscopy (STM), Han et al.also concluded that subsurf ace oxygen acted as the precursor to bulk oxidation and, in addition, suggested that the near subsurface oxygen reservoir required a critical concentration prior to bulk oxide formation [24,25]. Howe ver, Gabasch et al., by conducting their own ambient pressure synchrotron based XPS, concluded that the nucleation of chemically unique bulk oxide seeds that grew into bulk oxide particles represented the rate limiting step for bulk oxide formation, ther efore providing a context with which to explain the observed kinetic oxidation hysteresis [22]. 132

PAGE 133

Recently, we discovered that low temperature oxidation (< 500 K) with gas-phase oxygen atoms results in the occupation of a broad ther mal desorption feature, desorbing between 500 K and 650 K during TPD, that had a corresponding coverage identical to the expected coverage range for bulk oxide seeds, which we referred to as the precursor state [131]. According to ion scattering spectroscopy (ISS), the precursor state exhibited greater extrinsic oxygen content than either the saturated chemisorbed adlayer, the surface oxide, or th e bulk oxide thin film [131]. Furthermore, reactive scattering measurements utilizing isotopically labeled oxygen have revealed that occupation of the precursor st ate corresponded to an in crease in overall oxide reactivity to gas-phase oxygen at oms [131], which is consistent with observations by Gabasch et al. and Chen et al. regarding the enhanced reactiv ity of the bulk oxide pr ecursor state to methane and carbon monoxide [10,18,26]. Based on these result s, we suggested that the precursor state represented chemisorbed oxygen atoms on top of the surface oxide. In addition, oxidizing Pd(111) below the thermal instability temperat ure of the precursor state (<500K) led to our discovery of a well-ordered PdO(101) thin film in ultrahigh vacuum as determined from LEED, XPS, and adsorption studies [136,158,187], which Hi nojosa et al.has also confirmed with STM (in preparation). This result was especially intrig uing considering the failure of previous attempts to produce a well-ordered surface with molecular oxidants at higher s ubstrate temperatures (>650K) according to STM and LEED [24,102], behavior we also observed with our experiments presented herein. Th ese results suggest a direct correlation between the precursor state thermal stability during oxida tion and the quality of thin f ilm produced. Furthermore, there is evidence that saturatio n thin film thickness is also strongly correlated with the thermal stability of the precursor state as demonstrated by the tremendous disparity between the 3.3 ML 133

PAGE 134

saturation coverage we achieved after oxidation at 500 K [131] a nd the coverages in excess of 100 ML Gabasch et al.achieved duri ng high temperature oxidation [22]. Unfortunately, a comprehensive theory for th e initial formation of bulk oxide on Pd(111) has yet to be proposed which adequately explains all of the experimental observations presented. Before such a comprehensive theory can be pr oposed, however, the strong correlations between oxidation kinetics and the thermal st ability of the precursor state s uggests that we must first learn more about this precursor state. In this study, we will present results that highlight the nature of the precursor state and elucidate the relations hip between the thermodyna mic stability of the precursor state and its effect on the final thin film thickness, morphology, a nd growth kinetics. Experimental Methods Previous studies [86,121] provide details of the three-level UHV cham ber 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 contact with a liquid nitrogen cooled reservoir. A type K thermocouple sp ot-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 supply, supports maintaining or linearly ramping the sample temperature from 81 K to 1250 K. Initially, sample cleaning consisted of sputtering with 600 eV Ar+ ions at a surface temperature of 900 K, followed by annealing at 1100 K for several minutes. Subseq uent 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 previo usly [131], we limited the sample temperature to 923 K to maintain oxygen-saturation in the subsurface reservoir to ensure reproducibility of oxygen coverages determined from temperature programmed desorption (TPD) obtained in this study. We considered the Pd(111) sample to be clean when we 134

PAGE 135

could no longer detect contaminants with X -ray photoelectron spect roscopy (XPS), obtained sharp low energy electron diffraction (LEED) patter ns consistent with the Pd(111) surface, and did not detect CO producti on during flash desorption after oxygen adsorption. A two-stage differentially-pumped chamber attached to the UHV chamber houses the inductively coupled RF plasma s ource (Oxford Scientific Instrument s) utilized to generate beams containing oxygen atoms for this study [121]. This system produces gaseous oxygen atoms by partially dissociating pure O2 (BOC gases 99.999%) continuously supplied to a small discharge chamber at the end of the plasma source. A ll oxygen atom beams used in this study were generated using an RF power of 120 W and an O2 flow rate that establis hes a pressure of 2 10-6 torr in the first pumping stage of the beam chamber. We calculated flux based on an assumption that gas-phase oxygen atoms experi ence a unity sticking coefficient up to the saturation of the surface oxide. We controlled incident flux by manua lly adjusting the refl ected power between 1 W and 60 W, enabling us to produce gas-phase oxygen atom fluxes between 0.002 and 0.03 ML/s, where 1 ML is defined as the atomic density of 1.53 1015 cm-2 of bulk-terminated Pd(111) [99]. Under the highe st flux conditions, we estimat e that 20% of the inlet O2 dissociates in the plasma under our highest atom flux cond itions. To ensure uniform impingement of the oxygen beam across the sample surface, we posi tioned 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. We annealed all surfaces to a substrate temperature of 675 K prior to obtaining the LEED images at a substrate temperature of 500 K with a beam energy of 63 eV. As discussed in the results section, annealing allows for more accurate assessments of the frac tional coverage of the surface oxide and negligibly affects the orde ring of the bulk PdO do mains according to LEED. 135

PAGE 136

X-rays for XPS were all generated from a ma gnesium coated anode, and all XPS spectra presented are Shirley background subtracted an d normalized. All ion scattering spectroscopy (ISS) results utilized a 1000 eV helium ion beam. Results and Discussion Precursor TPD Feature Figure 6-1 shows a series of TPD spectra obtained after expo sing the Pd(111) surface to gas-phase oxygen atoms at a substrate temperatur e of 100 K. The precursor desorption feature, which we had previously attributed to recomb inative desorption of oxygen adatoms chemisorbed on the surface oxide [131], desorbs as a br oad feature between 500 and 650 K immediately preceding the autocatalytic decomposition of the surface oxide observed after 700 K. However, new results presented here have allowed us to further refine our interpretation, as we will discuss. The TPD spectra presented in Figure 6-1 suggest that the th ermal desorption spectrum of the precursor state is not consistent with an elementary desorption mechanism such as the associative desorption expected from a chemisorbed oxygen adlayer. To begin with, the precursor TPD feature exhibits a common leadi ng edge consistent with zero-order desorption kinetics which is usually a result of multilayer desorption whereby the desorbing species exhibits a constant concentration and desorption area es tablished by a phase tran sition. Since multilayer oxygen desorption above 500 K under ultrahigh vac uum conditions is highly unlikely, the common leading edge observed during TPD su ggests non-elementary desorption chemistry possibly related to domain size effects. Furtherm ore, the precursor peak temperature increases with initial precursor coverage which is al so highly inconsistent with chemisorbed oxygen adlayers due to repulsive adso rbate interactions expected between chemisorbed oxygen adatoms. 136

PAGE 137

We previously labeled the feature precursor based off of evidence that the precursor state was consumed by the surface oxide to form bulk oxide for both Pt(111) and Pt(100) oxidation [115]. However, as shown in Figure 6-1, none of the precursor state reacts with the substrate to form bulk oxide during the thermal flash as shown by the invariance of the surface oxide peak shape and intensity despite increases in the initia l precursor concentration. Since we have been able to oxidize Pd(111) with gas-phase oxygen atoms at temperatures below 100 K (Figure 613), the lack of precursor convers ion to bulk oxide cannot be attributed to insufficient thermal energy. We also observed absolutely no difference in the precursor state intensity when comparing a TPD spectrum collected immediately after the dose to a TPD spectrum collected after a 10 minute wait at 500 K prior to the sample flash. Therefore it appe ars that the precursor reacts to form bulk oxide thr ough a reaction with gas-phase oxyge n atoms rather than with the surface oxide. Let us now see how the thermal stab ility of the precursor state affects oxidation kinetics. Gas-Phase Oxygen Atom Uptake Curves Figure 6-2 shows oxygen uptake as a functi on of total oxygen atom dose, substrate temperature, and incident flux. Notice that th e data was plotted as a function of gas-phase oxygen atom dose in order to accentuate how in cident flux affects on uptake behavior. The uptake curves presented in Figure 6-2, conducted at 500 K and 650 K, illustrate the drastic effect the thermal stability of the precursor state has on oxygen uptake behavior. Notice that oxygen uptake is extremely facile at 500 K, with nearly linear uptake through both the chemisorbed and surface oxide saturation coverages. Considering the fact that oxidation past the chemisorbed phase and the surface oxide phase require substa ntial increases in relative molecular oxygen pressures [21-23,99], the facile oxidation of Pd(111) through bot h the chemisorbed phase and the 137

PAGE 138

surface oxide suggests that the rate limiting step for oxidation of both substrates is molecular oxygen dissociative adsorption. Figure 6-2a also shows that incident flux has no effect on the uptake behavior as demonstrated by the nearly perfect overlap obs erved for incident fluxes of 0.0083 ML/s and 0.0283 ML/s. The flux independence at 500 K suggests that the rate of competitive oxygen desorption scales linearly with flux. This behavi or, in conjunction with the apparent sigmoidal uptake, is a strong indicator of Langmuirian uptake since the rate of adsorption appears to be directly proportional to the fraction of the surf ace not covered by bulk oxide resulting in a (1) oxidation rate dependence, where equals the coverage of the bulk oxide. However, the Langmuirian uptake observ ed during 500 K oxidation was not observed during oxidation at 650 K as shown in Figure 6-2c. While oxyge n uptake at 650 K is initially a linear function of dose, oxidation halts upon reac hing the saturation coverage of the surface oxide at 0.7 ML which suggests that competitive de sorption of the precursor state is restricting further oxidation. However, upon closer inspection of Figure 6-2d, it is cl ear that the rate of oxidation gradually increases up to a dose of approximately 30 ML, and then decelerates to another plateau which is presumably caused by th e saturation of the bulk oxide. Accelerating oxidation is clearly not Langmuiri an in nature, which suggests that oxidation at 650 K proceeds via an autocatalytic growth mechanism. Therefore, oxidatio n within the acceleratory uptake range proceeds via particle grow th and/or nucleation, since acceleratory uptake is inconsistent with Ficks law and ther efore cannot be attributed to thin film growth. At 650 K, we observe a flux dependence on the normalized rate of oxidation as shown by the lack of overlap between the two curves as shown in Figure 62b, demonstrating that the rate of competitive desorption does not scale linearly with incident flux since higher incident fluxes 138

PAGE 139

result in a larger fraction of incident oxygen atoms oxi dizing the surface. This behavior suggests that the reaction order of oxide growth exceeds the reaction order of oxide decomposition as has also been observed during Pt(100) oxidation [115], however we have since refined this model as we shall discuss later. Finally, the apparent saturation coverage achieved at 650 K substantially exceeds the saturation coverage achieved at 500 K by ove r 300% as shown in Figure 6-2d. While the saturation coverage at 500 K was 3.3 ML, we ach ieved a saturation coverage of greater than 14 ML at a substrate temperature of 650 K. This finding suggests the existe nce of strong kinetic final state effects whereby temperature and pre ssure kinetic pathway de pendencies ultimately lead to changes in the final th in film thickness and, possibly, morphology. Overall, the results presented in Figure 6-2 demonstr ate that the thermal stability of the precu rsor state strongly correlates with both a drastic sh ift in oxidation kinetics and an increase in the oxygen saturation coverage. Low Energy Electron Diffraction LEED results provide significan t insight into the character of initial oxide phase and morphological development at 500 K and 650 K as shown in Figure 6-3. All LEED images were obtained after producing the indicated oxygen cove rage, followed by annealing the substrate to 675 K to give the surface oxide the long range order necessary for LEED imaging, followed finally by dropping the substrate temperature back to 500 K to obtain the LEED images presented. Note, however, that annealing to enable LEED imaging of the surface oxide negligibly affects the ordering of the bulk oxide. Oxidation at both 500 K and 650 K results in nearly identical ex trinsic morphological progressions as shown in Figure 6-3. Oxidation beyond the sa turation of the surface oxide coincides with the gradual appearance of th e PdO(101) LEED pattern [158]. The appearance of 139

PAGE 140

the PdO(101) LEED pattern also coincides with the gradual di sappearance of the Pd5O4 [102] and Pd(111) substrate spots which suggests that bulk oxide partic les grow at both temperatures. However, the similarities of the LEED pattern progressions provide st ark contrast to the differences in the oxygen coverages associated with each LEED patte rn: much more oxygen uptake at 650 K is required to produce the same LEED pattern at 500 K. For example, the LEED pattern obtained from an oxygen concentration of 6.1 ML produced at 650 K still exhibits substantial surface oxide contributions even thou gh the saturation coverage of 3.3 ML produced at 500 K results in a saturated Pd O(101) thin film. A similar PdO( 101) thin film LEED pattern is not observe until at least 14 ML of oxygen has been deposited at 650 K, which is consistent with a slower overall rate of extrinsi c bulk oxide formation at 650 K. Nonetheless, at both substrate temperatures, oxygen uptake termination corresponds to the formation of a completed PdO(101) surface and the disappearance of th e surface oxide substrate, demonstrating that presence of the surface oxide is necessary for continued oxygen uptake. Significantly slower growth of extrinsic bulk oxide relative to total oxygen uptake at 650 K compared to 500 K suggests that either there is concurrent occupation of subsurface dissolved oxygen, or that the bulk oxide e xhibits greater extent s of growth normal to the substrate plane when grown at temperatures above the thermal stab ility range of the precur sor state. Therefore, we conducted x-ray photoelectron spectroscopy to further characterize oxygen uptake at these two substrate temperatures. X-ray Photoelectron Spectroscopy Figure 6-4 shows Pd 3d5/2 XPS spectra collected after oxidi zing the Pd(111) substrate with gas-phase oxygen atoms at 500 K and 650 K. The a ppearance of a feature centered at BE 336.5 eV has been previously associat ed with bulk oxide [22] and sugge sts that bulk oxi de is indeed formed during oxidation at both substrate temperatures. However, bulk oxide content correlates 140

PAGE 141

well only with total oxygen content, and not w ith extrinsic bulk oxide content according to LEED, which suggests that bulk oxide particles at 650 K exhibit more growth normal to the surface plane relative to bulk oxide particles grown at 500 K. Furthermore, the spectrum representing the 2.9 ML oxide grown at 650 K has substantially less bulk oxide character than the 3.3 ML oxide generated at 500 K, demonstrating that the bulk oxide of the 2.9 ML experiences more signal attenuation that would be expected from thicker particles. The results obtained from XPS, LEED and TPD collectively suggest that an oxide grown at 650 K has significantly rougher yet th icker morphology than an oxide grow n at 500 K. Therefore, oxidation at temperature greater than the thermal stability range of the precursor state corresponds to an overall increase in oxide particle growth normal to the substrate. Identifying the role of the precursor state in the overall kinetic mechanism for bulk oxide formation requires knowledge of its chemical a nd compositional state. From prior work, we know that the precursor state has the highest extrinsic oxygen c oncentration of all the oxygen phases probed [131], prompting us to assign th e precursor state to oxygen adatoms on top the surface oxide. However, we have since obtained a dditional results which have led us to further refine our earlier interpretation of the nature of the precursor state and its role in oxidation kinetics as we shall now present. Oxygen Uptake Temperature Dependence We investigated the temperature dependence on the net rate of oxidation with the idea of obtaining the kinetic barriers for the precursor transition to bulk oxide. The results, however, were unexpected. Figure 6-5a shows total oxyge n uptake after a 4.4 ML high flux gas-phase oxygen atom dose as a function of substrate temper ature ranging from 100 K to 750 K. Note that we chose the 4.4 ML high flux dose because it re sulted in the largest difference in oxygen coverage between the 500 K and 650 K uptake curv es as shown in Figure 6-2d. We observed a 141

PAGE 142

gradual increase in the rate of bulk oxide forma tion from 100 K to 600 K, however, the rate of uptake abruptly drops at 600 K, wh ere a 0.4% increase in substrat e temperature results in over a 40% decrease in total oxygen uptak e. Note that while the discontinuity was observed within the thermal instability range of th e precursor state, the decrea se in the oxygen uptake did not correspond to desorption of the pr ecursor state as demonstrated by the lack of a precursor feature in either of the TPD spectra collected at 600 K and 602 K shown in Figure 6-5b. Since there was little change in total oxygen upt ake between 550 K and 600 K (also w ithin the thermal instability range of the precursor state), the discontinuity observed from 600 and 602 K suggests that the rate limiting precursor mediated oxidation mechan ism is not an elementary reaction between chemisorbed oxygen adatoms and the surface oxide Simply put, the precursor mediated mechanism we previously presen ted [115] could not explain the discontinuity, requiring us to revisit and refine our previ ously proposed mechanism. The discontinuity suggests a complete change in the dominant oxidation mechanism which may occur after a substrate phase transition. Inde ed, the observed behavior is consistent with a phase transition; however, a substrate phase transi tion does not appear to be responsible for the discontinuity since the disconti nuity was not reproduced at 600 K for a lower incident flux as shown in Figure 6-5c. Rather, we will present argu ments that the formation of the precursor state is the phase transition responsible for the observed discontinuity, a nd that the discontinuity does not represent a sudden decrease in the rate of precursor conversion to bulk oxide, but rather a sudden increase in the rate of precursor deco mposition. However, a substrate phase transition does occur between 500 K and 650 K which may have a significant effect on the observed oxidation kinetics or the thermal de sorption trace of the precursor st ate as we shall now discuss. 142

PAGE 143

Annealing the Surface Oxide Recently, Klikovits et al. di scovered, using STM and DFT, the existence of several metastable surface oxides only formed when oxidi zing Pd(111) at substrate temperatures below 605 K [128], in contrast to the well-known stable Pd5O4 surface oxide generated at higher substrate temperatures [100]. Indeed, by monito ring the substrate LEED pattern during TPD, we also observed a substrate reconstruction occurring between 500 K and 675 K which is consistent with observations made by Klikovits et al. Figur e 6-6 shows LEED images taken before and after annealing a 0.75 ML surface generated at 500 K to a substrate temperature of 675 K. Notice that prior to annealing, the 0.75 ML surface shows only small traces of a 2x2 pattern superimposed on a blurry 1x1 substrate, with the blurriness suggesting that the substr ate exhibits poor long range ordering. We previously demonstrated th at an oxygen coverage of 0.7 ML grown at 500 K corresponded to a minimum in nonthermal oxygen ab straction reactivity whic h contrasted starkly with the relatively hi gh reactivity of the chemisorbed adlayer [131]. Unless a compressed chemisorbed phase exhibits less reactivity th an an uncompressed chemisorbed phase, the 0.75 ML appears to be consistent with the formation of an oxyge n phase unlike the chemisorbed phase which we postulate is metastable surface oxide. During the anneal to 675 K, however, we obs erved long range ordering with the formation of a sharp LEED pattern consis tent with a well-ordered Pd5O4 surface [100,102]. Therefore, the ordering observed during temper ature programmed LEED is consis tent with the formation of metastable surface oxides that subs equently reconstruct to form Pd5O4 upon annealing. However, while metastable surface oxides are likely to have different act ivities towards molecular oxygen dissociation, we observed no difference on the rate of oxide formation on both the annealed and unannealed surface oxides (not shown), which suggests that both metastable and stable surface oxides are equally reactive to oxidation by gas-phase oxygen atom s. The sudden decrease, or 143

PAGE 144

discontinuity, in net oxidation rate observed at 600 K as shown in Figure 6-5a cannot therefore be attributed to a substrate phase transition from metastable to stable surface oxide since both substrates exhibit the same reactivity toward s gas-phase oxygen atoms. Hence, the sudden decrease in net oxidation rate must instead be a result of a sudden increase in the decomposition rate of the precursor state. Unfortunately, the substrate phase tran sition introduces some uncertainty regarding the identifi cation of the precursor state. The metastable surface oxides all exhibit grea ter oxygen concentrations than the stable Pd5O4 surface oxide [128]. Therefore we would e xpect oxygen atoms to be ejected from the lattice as a result of the metastable surf ace oxide phase transition to the stable Pd5O4 structure during annealing. If thes e ejected oxygen atoms from the latt ice desorb upon ejection, then we would expect that these oxygen atoms would deso rb during the substrate reconstruction, which happens to correspond to the same temperature ra nge as the precursor state thermal desorption feature. Subsequently, these findings raise an im portant question regarding the precursor feature: is the precursor state indeed a distinct state wh ich decomposes during TPD, or is the precursor state an artifact of the surface oxide phase transition occurring between 500 K and 650 K? Perhaps the question could be bett er rephrased: would it be possible to fill the precursor state on an annealed surface oxide, and if so, then how is the precursor st ate desorption feature affected by the substrate reconstruction? Certainly if the pr ecursor state desorption feature was a result of a substrate phase transition, then it should not be possible to occupy the precursor state on a preannealed surface oxide. We therefor e attempted to generate the precursor state on a pre-annealed surface oxide. The Precursor State: A Second Look Oxidation of the annealed surface oxide still re sults in the formation of the precursor state as shown in Figure 6-7, which s uggests that the precursor desorp tion feature corresponds to the 144

PAGE 145

decomposition of a distinct state rather than a desorption artifact from a surface oxide phase transition. Figure 6-7 shows TPD spectra obta ined after dosing gas-phase oxygen atoms at a substrate temperature of 100 K on a pre-annealed surface oxide generated at 500 K then annealed to 675 K. Analysis of the precursor state generated on an annealed surf ace oxide presents an opportunity to study the thermal de sorption spectrum of the precu rsor state without the added complication of a concurrent subs trate phase transition induced de sorption feature as we shall now discuss. Aside from a change in the leading edge behavior (which now undercuts itself with increased initial coverage) and a slight increase in peak symmetry, the precursor state shares the same key desorption trace characteristics when grown on either the a nnealed or unannealed surface oxides: namely an increasing precursor st ability with initial c overage, and a thermal desorption range between 500 K and 650 K. As previously mentioned, chemisorbed oxygen adatoms typically do not increase in stability with increased initial coverage due to destabilizing repulsive interactions (i.e. cr owding effects), and an increase in desorption temperature is inconsistent with the recombina tive desorption of oxygen adatoms. Therefore, the precursor state desorption feature does not appear to represen t the recombinative desorption of chemisorbed oxygen adatoms on top of the surface oxide, as we previously suggested [ 131]. However, ejected palladium substrate atoms may st abilize the oxygen adlayer resulti ng in the formation of oxide seed particles, as Gabasch et al. previously s uggested [22], which we believe should increase in stability with increased particle size according to the Kelvin equati on. In other words, instead of modeling the precursor state as adsorbed oxygen atoms that react with the surface oxide to nucleate bulk oxide particles, the precursor state appears to be t hose bulk oxide particles. The lack of observable transition to bulk oxide in the absence of a gas-phase oxidant is also 145

PAGE 146

consistent with this interpretation since necessary reagents for bulk oxide particle growth are incident gas-phase oxygen atoms. However, we previously observed that occupa tion of the precursor state corresponded to an extrinsic oxygen content greater than the extr insic oxygen content of both the surface oxide and the bulk oxide according to ISS [131]. The high re lative oxygen content of the precursor state suggests that the precursor stat e does not represent small partic les of bulk oxide but rather represents a chemically distinct form of oxygen-rich oxide. Note however, that due to the strong geometric dependency of ISS, the high oxygen signa l obtained from ISS may not necessarily be indicative of truly higher oxygen content. Nonetheless, the assignment of the precursor state as an oxygen-rich oxide seeds which are chemically distinct from either the surface oxide or bulk oxide is further supported by the apparent depleti on of the precursor state which coincides with bulk oxide growth as shown in Figure 6-7b. Bulk Oxide Growth If the precursor state represents small and le ss stable (relative to a flat surface) oxide particles, then the particles should continuously increase in stability as they grow according to the Kevin equation. The continuous increase in particle stability w ould then manifest itself as a gradual and continuous increase of the precursor peak temperatur e and intensity as particles continue to grow in size duri ng oxidation. Indeed, we observed a gradual transformation of the precursor peak to become the bulk oxide peak du ring Pt(100) oxidation [96] which is consistent with small bulk oxide particles of PtO2 representing the precursor to bulk oxide. However, the continuous conversion of precurs or to bulk oxide observed with Pt(100) oxidation was not observed during Pd(111) oxidation. As shown in Figure 6-7b, an oxygen covera ge 1.4 ML corresponds to an effective saturation of the precursor state, with subseque nt oxidation leading to a decrease of precursor 146

PAGE 147

peak intensity and temperature as bulk oxide gr ows. In other words, the precursor state is consumed during bulk oxide formation either duri ng the temperature flash via an Ostwald ripening process, or prior to the temperature flash during the gas-phase oxygen atom dose. With no measurable conversion of precursor to bulk oxide occurri ng during the temperature flash as shown in Figure 6-1, it seemed unlikely that Ostwald ripening can completely account for the outright concurrent consumption of the precursor state with bu lk oxide growth. However, ion scattering spectroscopy tells us a different story. Figure 6-8 shows a plot of extrinsic oxygen concentration as a function of total oxygen coverage, with oxides generated at a substrate temperature of 100 K prior to scanning at a substrate temperature of 500 K. Notice that oxidation up to approximately 1.4 ML yields a near linear increase in the extrinsic oxygen conten t. The oxygen content then reaches a maximum after 1.4 ML, and then gradually declines until a covera ge of approximately 2 ML, where it then abruptly drops to a steady ISS signal consistent with bulk oxide. Also notice that 2 ML roughly corresponds to the last coverage where precursor can be detected using TPD as shown in Figure 6-7b. The key finding here is that elevated oxygen ISS signals associ ated with the precursor state persists well after the precursor state has signifi cantly diminished according to TPD, suggesting that above 1.4 ML, precursor is consumed and transformed into bulk oxi de during the thermal flash. Taken together, these findi ngs suggest that continued grow th of the precursor TPD peak immediately following 1.4 ML is restricted due to an Ostwald ripening process, and therefore precursor particles beyond a critical size grow and ultimately convert to bulk oxide during the temperature ramp. However, determining that the precursor st ate is converted to bulk oxide via Ostwald ripening and oxidation does not explain why the precursor stat e exhibits a higher extrinsic 147

PAGE 148

oxygen concentration than bulk oxide. In fact, the ISS results suggest that the precursor state is in fact chemically distinct from the bulk oxide, a conclusion also shared by Ketteler et al. and Gabasch et al. with their identification of a ch emically distinct precursor state from ambient pressure synchrotron XPS results [22,23]. We ag ree and believe the precursor state actually is chemically distinct, yet the precursor should not necessarily be classi fied as an entirely different chemical species on the surface. Perimeter atoms of smaller particles expe rience greater positive interfacial curvature which lowers coordination and raises the chem ical potential of the particle. In regards to the XPS results, it is conceivabl e that the perimeter oxygen or palladium atoms of the bulk oxide seeds do indeed have a unique ch emical signature due to low coordination. However, identification of the precursor species as a new surface compound has as much merit as labeling water molecules at a meniscus as a different compound. Furthermore, non-fullycoordinated palladium atoms at the parameter of the particles may indeed result in a higher overall oxygen concentration of the particle s as detected by ISS since the uncoordinated palladium atoms will likely bind with any excess oxygen atoms available in order to stabilize. Considering how PdO2 is also a stable bulk oxide species under certain conditions [188], this possibility is not farfetched. Therefore, the ISS re sults are still consistent with our interpretation of the precursor state as representing small par ticles of PdO. Since particle stability is well described by the Kelvin equation, we will now briefly review the Kelvin equation and then discuss how this equation can explain our results. The Kelvin Equation and Particle Growth Model The Kelvin equation, shown on t op of Figure 6-9, relates the e ffective vapor pressure of a particle, Pp, with the vapor pressure of a flat oxide surface, P0, the molar volume, V, the surface parameter, the ideal gas constant, R, temperature, T, and particle radius, rp. Note that we define an effective oxygen pressure as the 2D lattice gas pressure on the surface oxide established by 148

PAGE 149

the kinetic competition between gas-phase oxygen adsorption, oxygen associative desorption, oxide particle growth, and oxide particle decom position. While the Kelvin equation predicts that the vapor pressure of particles reach es infinity as particle radius approaches zero, note that this prediction constitutes a continuum approximation and fa ils in the limit of particle nucleation. In other words, there must be a maximum particle vapor pressure associated with precursor oxide seeds. The Kelvin equation has been plotted as a func tion of particle radius in Figure 6-9. We predict and have labeled on the chart three di stinct pressure/temperature ranges for oxygen uptake: stable seed nucleation, unstable seed nu cleation, and subsurface oxidation. The stable seed nucleation pressure range represents oxidation under effective 2D oxygen lattice gas pressures greater than the vapor pressure of precursor oxide s eeds. In this pressure range, precursor seeds are thermodynamically stable meaning all produced seeds are expected to monotonically increase in size with dose. Precursor oxide seeds nucleated in the pressure range below the precursor seed vapor pressure are, on the other hand, thermodynamically unstable and are expected to predominantly decompose rather than grow unless stabilized by defect sites. However, kinetic theory predicts that despite the precursor stat e thermodynamically instability, any nonzero rate of oxide seed nucleation resu lts in a nonzero steady-st ate concentration of precursor oxide seeds. Exposure of the substrat e to any effective 2D lattice gas pressure exceeding the vapor pressure of a flat PdO surface results in a non-zer o rate of stable bulk oxide particle formation as previously unstable partic les become stable through growth. However, the steady state concentration of precu rsor oxide seeds is expected to increases dramatically once the chemical potential of the system matches the chemical potential of the precursor seeds. Therefore, we believe that the sudden decrease in the net rate of oxide particle nucleation 149

PAGE 150

reflected by the discontinuity observed from the constant dose and temperature dependent oxidation experiments shown in Figur e 6-5a is a result of a match of the chemical potential of the system with the chemical potential of precursor seeds. In other words, the oxygen 2D lattice gas pressure established at a subs trate temperature of 600 K and a gas-phase oxygen atom incident flux of 0.0174 ML/s apparently e quals the effective vapor pressure of bulk oxide precursor seeds. This conclusion would be further supported by evidence of a shift in the discontinuity temperature for different incident fluxes, and indeed the low flux data presented in Figure 6-5c suggests a decrease of the discontinuity temperatur e. However, additional re sults are necessary to confirm this. What happens when the effective oxygen 2D lattice gas pressure is below the vapor pressure of a flat PdO surface? Since not even the flat bulk oxide surface is thermodynamically stable, the particle growth model predicts that absolutely no bulk oxide will form (at least on the surface). However, Ficks law predicts that any non-zero concentrati on gradient will drive diffusion. While subsurface oxygen dissolution is likely occurring regardle ss of effective 2D lattice gas pressure, the selectivity towards s ubsurface diffusion over oxide formation increases with decreasing pressure. This is in fact observed when exposing the Pd(111) substrate to a very low incident gas-phase oxygen atom b eam flux as we shall discuss next. Flux Effects on TPD State Population Figure 6-10 shows TPD spectra obtained after exposing Pd(1 11) held at a substrate temperature of 650 K to gas-pha se oxygen atom incident flux es of 0.0174 ML/s and 0.002 ML/s. A direct TPD spectral evolution co mparison between oxygen uptake at high (0.0174 ML/s) and low (0.002 ML/s) flux highlights th e drastic effects of flux on relative state occupations. High flux oxygen uptak e results in the typical st ate occupation progression as observed with previous studies where a chemis orbed phase gradually becomes replaced by an 150

PAGE 151

explosive autocatalytic surface oxide peak that transitions into bulk oxi de [99]. However, low flux oxygen uptake produces an entirely different thermal desorption trace beyond the saturation of the surface oxide at 0.7 ML as shown in Fi gure 6-10a. After saturating the surface oxide, the baseline established after decomposition of the surface oxide begins to increase with increasing initial coverage which appears to be consistent with subsurface oxygen growth. While little is known regarding the nature of subsurface oxygen, what is known is that subsurface oxygen is very stable and desorbs at substrate temperatures exceeding 923 K [103]. And while it is not clear from our results if we had indeed formed subsurface oxygen, what is clear is that oxygen uptake beyond 0.7 ML does not represent bulk oxide formation since surface oxide still clearly desorbs as de monstrated by the invariance of th e autocatalytic peak even as initial oxygen coverage exceeds 0.7 ML. Since a st eady baseline desorption behavior observed in Figure 6-10a is inconsistent with desorption of any known surface stat e known to thermally desorb between 750 K and 923 K, we believe this state represents depletion of a very large subsurface oxygen reservoir. However, further studies are clearly need ed to confirm this. Nonetheless, our results demons trate that incident oxygen flux has a large effect on oxidation reaction pathways and relative stat e occupation selectivity. In addi tion, the particle growth model apparently succeeds at predicting the preferential occupation of subsurface oxygen which provides more support for the valid ity of the model. Finally, a kinetic modeling treatment of particle nucleation and growth can effectively reproduce the oxyge n uptake behavior observed in Figure 6-2 for both substrate temperat ures as we shall now discuss. Simplified Oxide Growth Model We propose that a steady state concentrati on of oxygen adatoms on top of the 2D surface oxide, which acts as the primary reagent for oxi de growth, is established from a kinetic competition between incident gas-phase oxygen at om adsorption, recombinative desorption, 151

PAGE 152

oxide particle nucleation and growth, and oxide particle de composition. The model can be expresso sed a fllows. (6-1) (6-2) (6-3) (6-4) (6-5) We obtained a kinetic expression for total oxygen uptake by consideri ng only three particle sizes, labeled p1, sa naoms. p2, and p3, and assuming atedy state c oncentration of oxyge adt (6-6) (6-7) (6-8) (6-9) (6-11) (6-10) 152

PAGE 153

Although models accounting for more particle sizes would allow for a more rigorous analysis of the spectra, the additional rigo r comes at the expense of comp ounding errors which devalues the results and therefore larger systems were not considered. The particles were assumed to grow spherically, and total oxide coverage, was set to equal the sum of the spherical cross-sectional areas. Oxide particle nucleation and growth was assumed to follow Langmuirian uptake kinetics, by setting the rates of bulk oxide particle nucleation and growth to be directly proportional to the fraction of the surface not occupied by bulk oxide, 1, in accordance with e xperimental results. Both p1 and p2 formation were treated as reversible to represent thermodynamically unstable precursor oxide par ticles, however p3 formation was treated as irreversible to represent thermodynamically stable oxide particles. All nu cleation and growth pref actors were considered equal, and activation barriers for nucleation and growth were assumed to be negligible as was suggested by our experimental observation of facile oxidation at substrate temperatures as low as 100 K. However, decomposition activation barriers we re set to increases with increasing particle radius according to the Ke lvin equation by setting E-1
PAGE 154

at higher temperatures by increasing the oxygen adatom concen tration. However, the high temperature oxidation simulation predicts that competitive decomposition of p1 allows for the selective growth of larger and more stable part icles, ultimately leading to an overall thicker oxide. No attempts were made to reproduce the temperature discontinuity primarily due to the difficulties associated with modeling a phase tr ansition using only kinetics. Also, the kinetic model fails to reproduce the behavior expected of oxidation at incident fluxes insufficient to produce oxides. Since the 3particle model still predicts that bulk oxide forms regardless of the effective oxygen adatom pressure, the assumption of irreversible p3 formation results in a failure of the kinetic model under extremely low flux conditions. Furthermore, the rate of competitive subsurface diffusion becomes non-negligible and therefore must be considered for a more accurate treatment of this pressu re range. Nonetheless, the simulations presented here reasonably reproduce our experiments and give validity to the m odel presented. Controlling Oxide Thin Film Thickness The particle growth oxidation model successfully predicts that oxide film thickness and morphology depend strongly on both the ambien t temperature and oxygen pressure during oxidation. Figure 6-12 presents tw o ambient condition scenarios: 1) low temperature or high flux oxidation or 2) high temp erature or low flux oxi dation. As predicted from the model, low temperature or high oxygen pressure oxidation results in hi gher oxide seed stability, resulting in a higher concentration of oxide s eeds growing. Since our data s uggests that oxidation terminates when oxide particles agglomerate, the effectiv e oxide thickness is established by the average extent of growth normal to the su rface plane each oxide particle achieves prior to agglomeration. Oxidation at higher substrate te mperatures or lower oxygen pressures, on the other hand, results in a lower stability of oxide seeds and therefore a lower concentration of oxide seeds growing. In other words, the oxide seeds that do grow start off with a lot more room to grow. The net result is 154

PAGE 155

that oxide particles grow to larger sizes before agglomerating with neighboring oxide particles, producing thicker oxide film s. This finding is particularly exci ting since it poses the possibility of engineering oxide film thic knesses, with thin film saturation thicknesses predetermined by temperature and pressure conditions rather than dosing time due to the self-terminating nature of palladium oxidation. However, how does oxide thickness depe nd on temperature when it is grown at temperatures below the thermal instability range of the precursor oxide seeds? Figure 6-13 shows oxygen uptake at a substrate te mperature of 100 K plotted al ong with oxygen uptake conducted at 500 K. Notice that at both substrate temperat ures we observed sigmoidal Langmuirian uptake behavior. However, notice that the saturati on coverages achieved at the two substrate temperatures differ significantly. The saturation coverage of 4 ML achieved at 500 K is nearly 33% higher than the saturation coverage of 3 ML achieved at 100 K. Furthermore, the saturation coverage appears to steadily de crease with decreasing substrate temperature from 600 K to 100 K as suggested by Figure 6-5a. If increasing initial precursor oxide seed density results in a decrease in final oxide film thickness, then the decrease in saturation coverage with decreasing substrate temperature suggests th at precursor oxide seed nucl eation density increases with decreasing substrate temperature. Indeed, this behavior has been observed with TEM studies of the oxidation of copper with mol ecular oxygen, with the increased particle nucleation density attributed to the decreased diffusivity of oxygen on the subs trate causing a reduction of the effective oxygen capture zone for each particle [189-194]. Therefore, the particle growth model effectively predicts final thin film thickness regardless of the temperature range. Summary In summary, we have presented evidence that the rate limiting step for Pd(111) oxidation is dissociative adsorption on the su rface oxide, as demonstrated by the facile oxidation with gas155

PAGE 156

phase oxygen atoms. Continuing oxidation beyond th e saturation of a surface oxide results in the formation of oxide seeds, or particles, whic h form regardless of surface oxide phase, and thermally decompose in ultrahigh vacuum between 500 K and 650 K. Oxidation in the intermediate range between the surface oxide a nd completed bulk oxide thin film can be best described as a kinetic competition between mol ecular oxygen dissociative adsorption, associative desorption, oxide particle nuclea tion, and oxide particle growth, with the selectivity between oxide particle nucleation and gr owth partially determined by surface diffusivity which is a function of substrate temperature. Oxidation self-terminates upon th e completion of an oxide thin film, regardless of the substrate temperature of am bient oxygen pressure. We have also identified three distinct temperature or pressure regimes fo r oxidation: stable seed nucleation, unstable seed nucleation, and oxygen dissolution. Furthermore, th e transition from the st able seed nucleation regime to the unstable seed nucleation regime is abrupt, with the transition temperature determined by the magnitude of the incident flux. We also presented a kinetic model for particle growth which qualitatively reproduces the uptake behavi or observed experimentally. Finally, a key finding from this work is the discovery that oxide thin film thickness could effectively be controlled using substrate temperature and oxyge n ambient pressure rather than oxygen dose. Appendix: Precursor to the Oxide Precursor State Note that others have suggested that subs urface oxygen acts as a precursor to bulk oxide formation whereby bulk oxide forms when a cr itical concentration of subsurface oxide accumulates close to the surface [23-25], and have also suggested the importance of flux to bulk oxide selectivity and growth [25] This possibility appears to be supported by our results which suggest that higher concentrations of subsur face oxygen causes an overall lowering of binding energy as shown by the drop in desorption energy as the subsurface layer is populated. While we had previously assigned the precursor to bulk oxide particle nucl eation to oxygen adatoms 156

PAGE 157

chemisorbed on the surface oxide our determination that the precursor peak during TPD may instead be attributable to bulk oxide seed s suggests that adsorbed oxygen adatoms do not constitute a stable species, yet our results do not preclude their existence either. Nonetheless, identification of the precursor to the bulk oxide seed precursors is beyond the scope of this study, which does not significantly affect the accu racy of our kinetic model predictions. 400500600700800900 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 O2 (10-2 ML s-1)Temperature (K) O(g) + Pd(111) Ts = 100 K = 1 K s-1Coverages (ML) 1.07 1.25 1.35 Figure 6-1. TPD spectra obtained after exposure of Pd(111) held at 100 K to gas-phase oxygen atom beam with an incident flux of 0.0132 ML s-1 and a temperature ramp rate of 1 K s-1. 157

PAGE 158

010203040 0 1 2 3 4 5 010203040 0 1 2 3 4 5 010203040 0 1 2 3 4 5 010203040506070 0 5 10 15 20 Isothermal Ts = 650 K Flux (ML s-1) 0.0174 0.0083 Isothermal Ts=500 K Flux (ML s-1) 0.0283 0.0083 650 K 500 K Low AO Flux Coverage (ML)Dose (ML) High AO Flux 650 K 500 K a b d c Figure 6-2. Total oxygen uptake cu rves plotted as a function of total oxygen atom dose in units of ML for the different fluxes and substrat e temperatures. a) Di rect comparison of two incident fluxes of 0.0283 ML/s and 0. 0083 ML/s at a substr ate temperature of 500 K, b) Direct comparison of two incide nt fluxes of 0.0174 ML/s and 0.0083 ML/s at a substrate temperature of 650 K, c) Direct comparison of two substrate temperatures at low incident flux (0. 0083 ML/s), d) Direct comparison of two substrate temperatures at high incident flux (0.0283 ML/s and 0.0174 ML/s). 158

PAGE 159

Figure 6-3. LEED images obtaine d with a beam energy of 63 eV for various oxygen coverages, shown below the images, generated at two substrate temperatures: 500 K and 650 K. 159

PAGE 160

339338337336335334333332 Mg Anode O(g) + Pd(111) Flux=0.0174 ML/s Ts = 500 K XPS Signal (a.u.)Binding Energy (eV), (ML) 0, 0 0.75, 0.75 2.0, 2.3 Pd0Pd2+ Mg Anode O(g) + Pd(111) Flux=0.0174 ML/s Ts = 650K (ML) 0, 0 0.82, 1.3 1.1, 3.4 6.1, 21 9.8, 32 12, 44 14, 63 2.9, 11 a 3.3, 11 b Figure 6-4. XPS Pd 3d5/2 spectra collected from Pd(111) fo r various oxygen coverages generated at a substrate temperature of: a) 650 K, b) 500 K. 160

PAGE 161

100200300400500600700800 0 1 2 3 4 5 500600700800900 590595600605610615 1.2 1.8 2.4 3.0 Total Oxygen Uptake O(g) + Pd(111) Flux = 0.0174 ML/s Dose = 4.4 ML Oxygen Coverage (ML)TPD O(g) + Pd(111) Flux = 0.0174 ML s-1Dose = 4.4 ML O2 desorption rate (a.u.)T (K), (ML) 600, 2.6 a 602, 1.5 b Flux = 0.0132 ML/s Dose = 3.2 ML Flux = 0.0174 ML/s Dose = 4.4 ML Oxygen Coverage (ML)Temperature (K) c Figure 6-5. Temperature Eff ects on Oxygen Uptake a) Total oxygen uptake after exposure of Pd(111) to a gas-phase oxygen atom beam with an incident flux of 0.0174 ML/s to a total dose of 4.4 ML. b) TPD spectra obtai ned after the 4.4 ML dose at 600 K and 602 K, c) close-up view of th e discontinuity at 600 K, plotted alongside lower flux (0.0132 ML/s) dose of 3.2 ML which lacks the discontinuity at 600 K observed with the higher flux dose. 161

PAGE 162

Figure 6-6. LEED images, using a beam energy of 63 eV, obtained from a 0.75 ML coverage generated at 500 K before and after annealing to 675 K. 162

PAGE 163

0.0 0.2 0.4 0.6 0.8 1.0 450500550600650700750800 0.0 0.2 0.4 0.6 0.8 O(g) + Pd5O4/Pd(111) Ts = 100 K Oxygen (10-2 ML s-1)Coverage (ML) 1.04 1.22 1.37 Temperature (K)Coverage (ML) 1.37 a b 1.55 1.57 1.68 1.75 1.85 2.03 Figure 6-7. TPD spectra obt ained after exposing the Pd5O4 surface oxide to gas-phase oxygen atoms at a substrate temperature of 100 K s howing the: a) growth of the precursor state and b) consumption of the precursor state as bulk oxide grows. 163

PAGE 164

0.00.51.01.52.02.53.03.5 Oxygen Coverage (ML) Extrinsic Oxygen Concentration (a.u.) Figure 6-8. Extrinsic oxygen c oncentration obtained by taking th e O/Pd ISS peak ratio, plotted as a function of total oxygen concentrati on. All ISS spectra we re obtained by using helium ions with an incident energy of 1000 eV, and all oxygen covered surfaces were generated at a substrate temperatur e of 100 K, followed by obtaining the ISS spectra at 500 K to remove contaminants. 164

PAGE 165

Figure 6-9. The Kelvin equation plotted as a function of particle radius, outlining the three effective pressure ranges: 1) stable seed nucleation, 2) unstable seed nucleation and, 3) subsurface oxidation. 165

PAGE 166

500600700800900 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 TPD O(g) + Pd(111) Ts=650K Flux=0.0174 ML/sOxygen (10-2 ML/s)Temperature (K) Coverage (ML) 0.40 0.56 0.75 0.94 0.96 1.2 TPD O(g) + Pd(111) Ts=650K Flux=0.002 ML/s Coverage (ML) 0.34 0.44 0.72 1.2 a b Figure 6-10. TPD spectra obtaine d after oxidizing Pd(111) at a substrate temperature of 650 K with an incident flux of: a) 0.002 ML/s and b) 0.0174 ML/s. 166

PAGE 167

p1 p2 p3 Oxygen Coverage (a.u.)Dose (a.u.)010203040506070 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 010203040506070 1 2 3 4 5 6 7 8 9 10 Coverage (ML)Dose (ML) Temperature (K) 650 Coverage (ML) Temperature (K) 500 a b Figure 6-11. Comparison of results generated from a) simulations wi th b) experimental data for low temperature and high temp erature oxidation kinetics. 167

PAGE 168

Figure 6-12. Schematic representa tion of particle growth illustrating the eff ects of pressure and substrate temperature on final oxide film thickness, under the following two conditions: a) stable seed nucleation whic h occurs under high oxygen pressure or low substrate temperature and b) unstable seed nucleation wh ich occurs under low oxygen pressure or high substrate te mperature. Dark blue repres ents the Pd(111) substrate, the light blue represents 2D oxide, and the or ange balls represent bul k oxide particles. 168

PAGE 169

169 0 100200300400 0 1 2 3 4 5 Total Oxygen Coverage (ML)Dose (ML) Flux = 0.0283 ML/s Temperature (K) 500 100 Figure 6-13. Total oxygen uptake plotted as a f unction of total dose of gas-phase oxygen atoms with an incident flux of 0.0283 ML/s, conducted at 500 K and 100 K.

PAGE 170

CHAPTER 7 CONCLUSIONS In conclusion, these studies demonstrate the great potential of util izing gas-phase oxygen atom beams in studying oxidation catalysts unde r ultrahigh vacuum conditions. Future work includes: 1) The adsorption of all industrially relevant mol ecules (particularly hydrocarbons) on PdO(101), 2) reactive scattering measurements of gas-phase oxygen atoms in situ with reagents of interest such as hydrocarbonds present in background. These experiments will hopefully enable a direct measure of the reactivity of various oxide phases to the reag ents of interest during oxidation, 3) attempting to create a thin film ordered bulk oxide on other various low miller index surfaces of palladium, platinum, and other platinum group metals in UHV using low temperature oxidation followed by annealing, 4) obtaining scanning tunneling microscopy images of the predicted oxide phase development on Pd(111) to provide geometric evidence for the identification of th e precursor state, and the finished PdO(101) surface, 5) utilizing the isotopic oxygen abstraction method of characterizing the reactivity of oxide phases developed on other metals in UHV, and 6) implementation of other nondestructive analytical techniques such as XPS and UPS to study oxide phase developmen t and reactivity with gas-phase oxygen atom beams in situ enabling time resolved chemical analysis during oxidation. 170

PAGE 171

LIST OF REFERENCES [1] R.B. Anderson, K.C. Stein, J.J. Feenan, L. J.E. Hofer, J. Ind. Eng. Chem. 53 (1961) 809. [2] R.J. Farrauto, M.C. Hobson, T. Kennelly E.M. Waterman, Appl. Catal., A 81 (1992) 227. [3] R. Burch, F.J. Urbano, P.K. Loader, Appl. Catal., A 123 (1995) 173. [4] R.J. Farrauto, J.K. Lampert, M.C. Hobson, E.M. Waterman, Appl. Catal., B 6 (1995) 263. [5] J.G. McCarty, Catal. Today 26 (1995) 283. [6] P. Salomonsson, S. Johansson, B. Kasemo, Catal. Lett. 33 (1995) 1. [7] J.N. Carstens, S.C. Su, A.T. Bell, J. Catal. 176 (1998) 136. [8] A.K. Datye, J. Bravo, T.R. Nelson, P. Atanasova, M. Lyubovsky, L. Pfefferle, Appl. Catal., A 198 (2000) 179. [9] R.S. Monteiro, D. Zemlyanov, J.M. Stor ey, F.H. Ribeiro, J. Catal. 199 (2001) 291. [10] H. Gabasch, K. Hayek, B. Kloetzer, W. Unterberger, E. Kleimenov, D. Teschner, S. Zafeiratos, M. Haevecker, A. Knop-Gerick e, R. Schloegl, B. Aszalos-Kiss, D. Zemlyanov, J. Phys. Chem. C 111 (2007) 7957. [11] C.F. Cullis, B.M. Willatt, J. Catal. 83 (1983) 267. [12] V.A. Bondzie, P.H. Kleban, D.J. Dwyer, Surf. Sci. 465 (2000) 266. [13] C.J. Zhang, P. Hu, J. Am. Chem. Soc. 123 (2001) 1166. [14] I. Nakai, H. Kondoh, T. Shimada, A. Rest a, J.N. Andersen, T. Ohta, J. Chem. Phys. 124 (2006) 224712/1. [15] K. Nakao, S.-I. Ito, K. Tomishige, K. Kunimori, Catal. Today 111 (2006) 316. [16] E.E. Tornau, V. Petrauskas, G. Zvejnieks, Catal. Today 116 (2006) 62. [17] G. Zvejnieks, V.N. Kuzovkov, V. Petrausk as, E.E. Tornau, Appl. Surf. Sci. 252 (2006) 5395. [18] H. Gabasch, A. Knop-Gericke, R. Schloegl M. Borasio, C. Weilach, G. Rupprechter, S. Penner, B. Jenewein, K. Hayek, B. Kloe tzer, Phys. Chem. Chem. Phys. 9 (2007) 533. [19] G. Zheng, E.I. Altman, J. Phys. Chem. B 106 (2002) 1048. 171

PAGE 172

[20] R. Imbihl, R.J. Behm, R. Schlgl Phys. Chem. Chem. Phys. 9 (2007) 3459. [21] D. Zemlyanov, B. Aszalos-Kiss, E. Kleimenov, D. Te schner, S. Zafeiratos, M. Haevecker, A. Knop-Gericke, R. Schloegl, H. Gabasch, W. Unterberger, K. Hayek, B. Kloetzer, Surf. Sci. 600 (2006) 983. [22] H. Gabasch, W. Unterberger, K. Hayek, B. Kloetzer, E. Kleimenov, D. Teschner, S. Zafeiratos, M. Haevecker, A. Knop-Gericke, R. Schloegl, J. Han, F.H. Ribeiro, B. Aszalos-Kiss, T. Curtin, D. Zemlyanov, Surf. Sci. 600 (2006) 2980. [23] G. Ketteler, D.F. Ogletree, H. Bluhm, H. Liu, E.L.D. Hebenstreit, M. Salmeron, J. Am. Chem. Soc. 127 (2005) 18269. [24] J. Han, D.Y. Zemlyanov, F.H. Ribeiro, Surf. Sci. 600 (2006) 2730. [25] J. Han, D.Y. Zemlyanov, F.H. Ribeiro, Surf. Sci. 600 (2006) 2752. [26] M.S. Chen, Y. Cai, Z. Yan, K.K. Ga th, S. Axnanda, D.W. Goodman, Surf. Sci. 601 (2007) 5326. [27] R.I. Masel, Principles of Adsorp tion and Reaction on Solid Surfaces (1996). [28] Y. Wen, X. Jiang, Plasma Chem. Plasma Process. 21 (2001) 665. [29] S.L. Brock, T. Shimojo, M. Marquez, C. Ma run, S.L. Suib, H. Matsumoto, Y. Hayashi, J. Catal. 184 (1999) 123. [30] S.L. Brock, M. Marquez, S.L. Suib, Y. Hayashi, H. Matsumoto, J. Catal. 180 (1998) 225. [31] V.I. Grinevich, N.V. Kolobova, V.V. Kostrov, Hi gh Energy Chem. 31 (1997) 441. [32] M.B. Kizling, S.G. Jr s, Appl. Catal., A 147 (1996) 1. [33] K.-P. Francke, H. Miessner, R. Rudolph, Catal. Today 59 (2000) 411. [34] H. Miessner, K.-P. Francke, R. Rudolph, Appl. Cata l., B 36 (2000) 53. [35] J.F. Xia, X.X. Gao, J.Y. Kong, H.X. Hui, M. Cui, K.P. Yan, Plasma Chem. Plasma Process. 20 (2000) 225. [36] X. Chen, J. Rozak, J. Lin, S.L. Suib, Y. Hayashi, H. Matsumoto, Appl. Catal., A 219 (2001) 25. [37] U. Roland, F. Holzer, F.-D. Kopinke, Catal. Today 73 (2002) 315. [38] H. Kim, A. Ogata, S. Futamura IEEE Trans. Plasma Sci. 34 (2006) 984. 172

PAGE 173

[39] A. Ogata, K. Yamanuochi, K. Mizuno, S. Kushiyama, T. Yamamoto, Plasma Chem. Plasma Process. 19 (1999) 383. [40] F. Holzer, U. Roland, F.-D. Kopinke, Appl. Catal., B 38 (2002) 163. [41] T. Kammler, J. Lee, J. K ppers, J. Chem. Phys. 106 (1997) 7362. [42] T. Kammler, M. Scherl, J. Kppers, Surf. Sci. 382 (1997) 116. [43] S. Wehner, J. Kppers, J. Chem. Phys. 108 (1998) 3353. [44] S. Wehner, J. Kppers, J. Chem. Phys. 109 (1998) 294. [45] S. Wehner, J. Kppers, J. Chem. Phys. 111 (1999) 3209. [46] S. Wehner, J. Kppers, J. Chem. Phys. 111 (1999) 3218. [47] A. Dinger, C. Lutterloh, J. Bien er, J. Kppers, Surf. Sci. 437 (1999) 116. [48] A. Dinger, C. Lutterloh, J. Bien er, J. Kppers, Surf. Sci. 449 (2000) 1. [49] J. Biener, E. Lang, C. Lutterloh, J. Kppers, J. Chem. Phys. 116 (2002) 3063. [50] J.-Y. Kim, J. Lee, Phys. Rev. Lett. 82 (1999) 1325. [51] W.J. Mitchell, J. Xie, Y. Wang, W.H. We inberg, J. Electron. Spectrosc. Relat. Phenom. 64-65 (1993) 427. [52] T.A. Jachimowski, W.H. Wei nberg, J. Chem. Phys. 101 (1994) 10997. [53] J. Xie, W.J. Mitchell, K.J. Lyons, Y. Wa ng, W.H. Weinberg, J. Vac. Sci. Technol., A 12 (1994) 2210. [54] J. Xie, W.J. Mitchell, K.J. Lyons, W.H. Weinberg, J. Chem. Phys. 101 (1994) 9195. [55] S.Y. Li, J.A. Rodriguez, J. Hrbek, H.H. Huang, G.-Q. Xu, J. Vac. Sci. Technol., A 15 (1997) 1692. [56] C.J. Hagedorn, M.J. Weiss, W.H. Wei nberg, J. Vac. Sci. Technol., A 16 (1998) 984. [57] C.T. Rettner, Phys. Rev. Lett. 69 (1992) 383. [58] C.T. Rettner, J. Chem. Phys. 101 (1994) 1529. [59] C.T. Rettner, D.J. Auerbach, J. Lee, J. Chem. Phys. 105 (1996) 10115. 173

PAGE 174

[60] Y. Takamine, A. Namiki, J. Chem. Phys. 106 (1997) 8935. [61] S.A. Buntin, J. Chem. Phys. 108 (1998) 1601. [62] M. Xi, B.E. Bent, J. Vac. Sci. Technol., B 10 (1992) 2440. [63] S.I. Yi, W.H. Weinberg Surf. Sci. 415 (1998) 274. [64] G. Eilmsteiner, W. Walkner, A. Winkler, Surf. Sci. 352-354 (1996) 263. [65] H. Plzl, G. Strohmeier, A. Winkler, J. Chem. Phys. 110 (1999) 1154. [66] T. Kammler, S. Wehner, J. K ppers, J. Chem. Phys. 109 (1998) 4071. [67] T. Kammler, J. Kppers, J. Chem. Phys. 111 (1999) 8115. [68] J.-Y. Kim, J. Lee, J. Chem. Phys. 112 (2000) 6015. [69] T. Zecho, B. Brandner, J. Kppers, Surf. Sci. 418 (1998) L26. [70] T. Kammler, D. Kolovos-Velliani tis, J. Kuppers, Surf. Sci. 460 (2000) 91. [71] M. Nagasaka, H. Kondoh, T. Ohta, J. Chem. Phys. 122 (2005) 204704. [72] F.P. Netzer, G. Kneringe r, Surf. Sci. 51 (1975) 526. [73] K.M. Ogle, J.M. White, Surf. Sci. 169 (1986) 425. [74] K.M. Ogle, J.M. White, Surf. Sci. 139 (1984) 43. [75] K.M. Backstrand, M.A. Weibel, R.M. Mois ion, T.J. Curtiss, J. Chem. Phys. 112 (2000) 7209. [76] K.M. Backstrand, M.A. Weibel, T.D. Hain, T.J. Curtiss, Chem. Phys. Lett. 287 (1998) 475. [77] C. Sachs, M. Hildebrand, S. Volkening, J. Wintterlin, G. Ertl, J. Chem. Phys. 116 (2002) 5759. [78] C. Sachs, M. Hildebrand, S. Volkening, J. Wintterlin, G. Er tl, Science 293 (2001) 1635. [79] A. Michaelides, P. Hu, J. Am. Chem. Soc. 123 (2001) 4235. [80] A.T. Capitano, A.M. Gabelnick, J.L. Gland, Surf. Sci. 419 (1999) 104. [81] G. Pauer, A. Winkler, J. Chem. Phys. 120 (2004) 3864. 174

PAGE 175

[82] D. Kolovos-Vellianitis, J. Kppers, J. Phys. Chem. B 107 (2003) 2559. [83] D. Kolovos-Vellianitis, T. Kammler, J. Kppers, Surf. Sci. 482 (2001) 166. [84] M.J. Weiss, C.J. Hagedorn, W.H. Wei nberg, J. Vac. Sci. Technol., A 16 (1998) 3521. [85] S. Shimokawa, F. Khanom, T. Fujimoto, S. Inanaga, A. Namiki, T. Ando, Appl. Surf. Sci. 167 (2000) 94. [86] A.L. Gerrard, J.-J. Chen, J.F. Weaver, J. Phys. Chem. B 109 (2005) 8017. [87] A.L. Gerrard, J.F. Weaver, J. Chem. Phys. 123 (2005) 224703. [88] J.F. Weaver, J.-J. Chen, A.L. Gerrard, Surf. Sci. 592 (2005) 83. [89] P. Heilmann, K. Heinz, K. Mller, Surf. Sci. 83 (1979) 487. [90] P.R. Norton, J.A. Davies, D.K. Creber, C.W. Sitter, T.E. Jackman, Surf. Sci. 108 (1981) 205. [91] K. Griffiths, T.E. Jackman, J.A. Da vies, P.R. Norton, Surf. Sci. 138 (1984) 113. [92] N. Kizhakevariam, E.M. Stuve, Surf. Sci. 275 (1992) 223. [93] A.T. Pasteur, S.J. Dixon-Warren, D.A. King, J. Chem. Phys. 103 (1995) 2251. [94] D.R. Lide, CRC Handbook of Chemistry and Physics, 71th ed. (CRC, Boca Raton, 1990), pp. 9-96. [95] J. Ree, H.K. Shin, Ch em. Phys. Lett. 258 (1996) 239. [96] R.B. Shumbera, H.H. Kan, J. F. Weaver, Surf. Sci. 601 (2007) 235. [97] T. Zecho, A. Horn, J. Biener, J. Kppers, Surf. Sci. 397 (1998) 108. [98] A. Guttler, D. Kolovos-Vellianitis, T. Zecho, J. Kppers, Surf. Sci. 516 (2002) 219. [99] H. Gabasch, W. Unterberger, K. Hayek, B. Kloetzer, G. Kresse, C. Klein, M. Schmid, P. Varga, Surf. Sci. 600 (2005) 205. [100] 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) 246103/1. [101] B. Klotzer, K. Hayek, C. Konvicka, E. Lundgren, P. Varga, Surf. Sci. 482-485 (2001) 237. 175

PAGE 176

[102] G. Zheng, E.I. Altman, Surf. Sci. 462 (2000) 151. [103] F.P. Leisenberger, G. Koller, M. Sock, S. Surnev, M.G. Ramsey, F.P. Netzer, B. Klotzer, K. Hayek, Surf. Sci. 445 (2000) 380. [104] A. Steltenpohl, N. Memm el, Surf. Sci. 443 (1999) 13. [105] E.H. Voogt, A.J.M. Mens, O.L.J. Gij zeman, J.W. Geus, Surf. Sci. 373 (1997) 210. [106] D.H. Parker, B. Koel, J. Vac. Sci. Technol., A 8 (1990) 2585. [107] K.W. Kolasinski, F. Cemic, E. Hasselbrink, Chem Phys. Lett. 219 (1994) 113. [108] B.A. Banse, B.E. Ko el, Surf. Sci. 232 (1990) 275. [109] X. Guo, A. Hoffman, J.T. Yate s, Jr., J. Chem. Phys. 90 (1989) 5787. [110] R. Imbihl, J.E. Demuth, Surf. Sci. 173 (1986) 395. [111] T. Matsushima, Surf. Sci. 157 (1985) 297. [112] L. Surnev, G. Bliznakov, M. Kiskinova, Surf. Sci. 140 (1984) 249. [113] D.L. Weissman-Wenocur, M.L. Shek, P.M. Stefan, I. Lindau, W.E. Spicer, Surf. Sci. 127 (1983) 513. [114] P. Legare, Y. Holl, G. Maire, Solid State Commun. 31 (1979) 307. [115] R.B. Shumbera, H.H. Kan, J. F. Weaver, Surf. Sci. 601 (2007) 4809. [116] R.B. Shumbera, H.H. Kan, J. F. Weaver, Surf. Sci. 600 (2006) 2928. [117] E. Lundgren, A. Mikkelsen, J.N. Andersen, G. Kresse, M. Schmid, P. Varga, J. Phys.: Condens. Matter 18 (2006) R481. [118] J. Gustafson, A. Mikkelsen, M. Borg, J.N. Andersen, E. Lundgren, C. Klein, W. Hofer, M. Schmid, P. Varga, L. Kohler, G. Kresse, N. Kasper, A. Stierle, H. Dosch, Phys. Rev. B 71 (2005) 115442. [119] E. Lundgren, J. Gustafson, A. Mikkelsen, J.N. Andersen, A. Stierle, H. Dosch, M. Todorova, J. Rogal, K. Reuter, M. Sche ffler, Phys. Rev. Lett. 92 (2004) 046101. [120] M. Todorova, E. Lundgren, V. Blum, A. Mikke lsen, S. Gray, J. Gustafson, M. Borg, J. Rogal, K. Reuter, J.N. Andersen, M. Scheffler, Surf. Sci. 541 (2003) 101. [121] H.H. Kan, R.B. Shumbera, J.F. Weaver, J. Chem. Phys. 126 (2007) 134704/1. 176

PAGE 177

[122] G.B. Hoflund, Handbook of surface and interf ace analysis: Methods for problem solving, New York, Marcel Dekker, Inc., 1998. [123] J. Harris, B. Kasemo Surf. Sci. 105 (1981) L281. [124] G.L. Kellogg, P.J. Feibelman, Phys. Rev. Lett. 64 (1990) 3143. [125] P.J. Feibelman, R. Stum pf, Phys. Rev. B 59 (1999) 5892. [126] Z.-X. Wang, X.-F. Jia, F.-H. Tian, S.-G. Chen, Chin. J. Chem. 22 (2004) 152. [127] J. Wang, Y. Yun, E.I. A ltman, Surf. Sci. 601 (2007) 3497. [128] J. Klikovits, E. Napetschnig, M. Schmid, N. Seriani, O. Dubay, G. Kresse, P. Varga, Phys. Rev. B 76 (2007) 045405/1. [129] G. Zheng, E.I. Altman, Surf. Sci. 504 (2002) 253. [130] P. Kostelnik, N. Seriani, G. Kresse, A. Mikkelsen, E. L undgren, V. Blum, T. Sikola, P. Varga, M. Schmid, Surf. Sci. 601 (2007) 1574. [131] H.H. Kan, R.B. Shumbera, J. F. Weaver, Surf. Sci. 602 (2008) 1337. [132] N. Cabrera, N.F. Mott, Rep. Prog. Phys. 12 (1949) 163. [133] F.P. Fehlner, N.F. Mo tt, Oxid. Met. 2 (1970) 59. [134] J. Rogal, K. Reuter, M. Sche ffler, Phys. Rev. B 69 (2004) 075421/1. [135] D.B. Rogers, R.D. Shannon, J.L. G illson, J. Solid State Chem. 3 (1971) 314. [136] J.A. Hinojosa, Jr., H.H. Kan, J.F. Weaver, J. Phys. Chem. C 112 (2008) 8324. [137] R. Franchy, Surf. Sci. Rep. 38 (2000) 195. [138] R. Burch, F.J. Urbano, Appl. Catal., A 124 (1995) 121. [139] J.H. Lee, D.L. Trimm, Fu el Process. Technol. 42 (1995) 339. [140] R. Kikuchi, S. Maeda, K. Sasaki, S. Wennerstrom, K. E guchi, Appl. Catal., A 232 (2002) 23. [141] D.L. Mowery, R.L. McCormick, Appl. Catal., B 34 (2001) 287. [142] D.L. Mowery, M.S. Graboski, T.R. Ohno, R.L. McCormick, Appl. Catal., B 21 (1999) 157. 177

PAGE 178

[143] G. Croft, M.J. Fuller, Nature 269 (1977) 585. [144] S.-H. Oh, G.B. Hoflund, J. Catal. 245 (2007) 35. [145] N.G. Petrik, G.A. Kimmel, Phys. Rev. Lett. 99 (2007) 196103. [146] M. Abu Haija, S. Guimond, A. Uhl, H. Kuhlenbeck, H.J. Freund, Surf. Sci. 600 (2006) 1040. [147] T.J. Beck, A. Klust, M. Ba tzill, U. Diebold, C. Di Valent in, A. Tilocca, A. Selloni, Surf. Sci. 591 (2005) L267. [148] A. Lobo, H. Conrad, Surf. Sci. 523 (2003) 279. [149] M.A. Henderson, S.A. Chambers, Surf. Sci. 449 (2000) 135. [150] M. Schulze, R. Reissner M. Lorenz, U. Radke, W. Schnurnberger, Electrochim. Acta 44 (1999) 3969. [151] A.M. De Asha, J.T.S. Critchley, R.M. Nix, Surf. Sci. 405 (1998) 201. [152] M.A. Henderson, Surf. Sci. 355 (1996) 151. [153] M.A. Henderson, Surf. Sci. 319 (1994) 315. [154] M. Hendewerk, M. Salmeron, G.A. Somorjai, Surf. Sci. 172 (1986) 544. [155] M.A. Henderson, Surf. Sci. Rep. 46 (2002) 5. [156] V.E. Heinrich, P.A. Cox, The Surface Sc ience of Metal-Oxides, Cambridge University Press, Cambridge, 1994. [157] G. Ketteler, S. Yamamoto H. Bluhm, K. Andersson, D. E. Starr, D.F. Ogletree, H. Ogasawara, A. Nilsson, M. Salmeron, J. Phys. Chem. C 111 (2007) 8278. [158] H.H. Kan, J.F. Weaver, Surf. Sci. 602 (2008) L53. [159] J. Assmann, V. Narkhede, N.A. Breuer M. Muhler, A.P. Seitsonen, M. Knapp, D. Crihan, A. Farkas, G. Mellau, H. Over J. Phys.: Condens. Matter 20 (2008). [160] J. Wang, C.Y. Fan, K. Jacobi, G. Ertl, J. Phys. Chem. B 106 (2002) 3422. [161] Y.D. Kim, A.P. Seitsonen, S. Wendt, J. Wang, C. Fan, K. Jacobi, H. Over, G. Ertl, J. Phys. Chem. B 105 (2001) 3752. 178

PAGE 179

[162] M. Wolf, S. Nettesheim, J.M. White, E. Hasselbrink, G. Ertl, J. Chem. Phys. 94 (1991) 4609. [163] G. Kresse, J. Furthmuller, Comput. Mater. Sci. 6 (1996) 15. [164] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169. [165] G. Kresse, J. Hafner, J. P hys.: Condens. Matter 6 (1994) 8245. [166] G. Kresse, J. Hafner Phys. Rev. B 47 (1993) 558. [167] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758. [168] P.E. Blochl, Phys. Rev. B 50 (1994) 17953. [169] J.P. Perdew, K. Burke, M. Ern zerhof, Phys. Rev. Lett. 77 (1996) 3865. [170] Y.L. Cao, Z.X. Chen, Surf. Sci. 600 (2006) 4572. [171] V.A. Ranea, A. Michaelides, R. Ramrez, P.L. de Andres, J.A. Ve rgs, D.A. King, Phys. Rev. Lett. 92 (2004) 136104. [172] T. Schiros, L.A. Naslund, K. Andersson, J. Gyllenpalm, G.S. Karlberg, M. Odelius, H. Ogasawara, L.G.M. Pettersson, A. N ilsson, J. Phys. Chem. C 111 (2007) 15003. [173] J. Zhao, B. Li, K.D. Jordan, J. Ya ng, H. Petek, Phys. Rev. B 73 (2006) 195309. [174] S.X. Yin, X.Y. Ma, D.E. Ellis, Surf. Sci. 601 (2007) 2426. [175] J. Weissenrieder, A. Mikkelsen, J.N. Andersen, P.J. Feibelman, G. Held, Phys. Rev. Lett. 93 (2004). [176] P.J. Feibelman, Science 295 (2002) 99. [177] R.A. Olsen, G.J. Kroes, G. Henkelman, A. Arnaldsson, H. Jonsson, J. Chem. Phys. 121 (2004) 9776. [178] G. Henkelman, B.P. Uberuaga, H. Jonsson, J. Chem. Phys. 113 (2000) 9901. [179] G. Henkelman, H. Jonsson, J. Chem. Phys. 113 (2000) 9978. [180] G.A. Kimmel, N.G. Petrik, Z. Dohnalek, B.D. Kay, J. Chem. Phys. 126 (2007) 114702/1. [181] C. Clay, L. Cumm ings, A. Hodgson, Surf. Sci. 601 (2007) 562. [182] G.A. Kimmel, N.G. Petrik, Z. Dohnalek, B.D. Kay, J. Chem. Phys. 125 (2006) 044713/1. 179

PAGE 180

180 [183] J. Cerda, A. Michaelides, M.L. Bocquet, P. J. Feibelman, T. Mitsui, M. Rose, E. Fomin, M. Salmeron, Phys. Rev. Lett. 93 (2004) 116101/1. [184] S.L. Tait, Z. Dohnalek, C.T. Camp bell, B.D. Kay, J. Chem. Phys. 125 (2006). [185] S.L. Tait, Z. Dohnalek, C.T. Camp bell, B.D. Kay, J. Chem. Phys. 122 (2005). [186] G.S. Karlberg, G. Wahnstr om, Phys. Rev. Lett. 92 (2004). [187] H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C 113 (2009) 1495. [188] C.-B. Wang, T.-H. Huang, Thermochim. Acta 381 (2002) 37. [189] G. Zhou, X. Chen, D. Gallagher, J. C. Yang, Appl. Phys. Lett. 93 (2008) 123104. [190] G. Zhou, L. Wang, J.C. Yang, J. Appl. Phys. 97 (2005) 063509. [191] G. Zhou, J.C. Yang, Ap pl. Surf. Sci. 222 (2004) 357. [192] G. Zhou, J.C. Yang, Surf. Sci. 559 (2004) 100. [193] G. Zhou, J.C. Yang, Ap pl. Surf. Sci. 210 (2003) 165. [194] J.C. Yang, M. Yeadon, B. Kolasa, J.M. Gibson, Scr. Mater. 38 (1998) 1237.

PAGE 181

BIOGRAPHICAL SKETCH Heywood Kan earned his BS degree in chemical engineering from the University of California at Berkeley where he did undergradu ate research under Dr. D.V. Schaffer and Dr. D.B. Graves. Heywood also worked as a process e ngineer at the Public Utilities Commission of San Francisco during an undergraduate co-op be fore beginning his doctoral studies at the University of Florida in the Department of Ch emical Engineering under Dr Jason F. Weaver in the fall of 2004. He has authored or coauthored eleven manuscripts in pe er-reviewed journals in addition to giving four presentati ons at international conferences.