1 FIRST PRINCIPLES STUDIES OF THE OXIDATION AND REACTIVITY OF Pt(111), Pd(111), AND THIN FILM Pd OXIDES By JEFFERY MICHAEL HAWKINS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 Jeffery Michael Hawkins
3 To Shelley
4 ACKNOWLEDGMENTS First, I would like to thank Prof Aravind Asthagiri for his support and guidance th roughout my graduate career He has introduced me to and shared his passion for challenging yet rewarding work in computational surface science I am fortunate to have been one of his first students, and I hope that the skills I have acquired will continue to serve me well in my future endeavors I would also like to thank P rof Jason Weaver, with whom I shared many conversations, for his valuable insights in connecting experimental and theoretical results I also thank Profs Susan Sinnott and Tanmay Lele for serving on my committee and for their helpful suggestions for my research. Many people have helped to make my time at the University of Florida enjoyable, and I am grateful for the role, however large or small it may have been, each of you playe d I am especially indebted to Beverly Hinojosa for her positive attitude despite having endured four years of sharing office space with me. I acknowledge the University of Florida HighPerformance Computing Center for providing computational resources for performing some of the calculations reported in this dissertation. I also acknowledge financial support from the University of Florida Alumni Fellowship and ACSPRF Grant No 47596G5 I must also thank my family for their constant support, even with a thousand miles between us Without them none of this would have been possible. Finally, I offer my deepest gratitude to my fiance, Shelley, to whom this dissertation is dedicated. Her unconditional love and encouragement have been instrumental in my su ccess, and I eagerly look forward to our future together.
5 TABLE OF CONTENTS ACKNOWLEDGMENTS .................................................................................................. 4 page LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 ABSTRACT ................................................................................................................... 11 CHAPTER 1 INTRODUCTION .................................................................................................... 13 Motivation ............................................................................................................... 13 Density Functional Theory ...................................................................................... 15 Schr dinger Equation ....................................................................................... 16 KohnSham Equations ..................................................................................... 17 ExchangeCor relation Functionals ................................................................... 18 Pseudopotentials .............................................................................................. 20 2 DENSITY FUNCTIONAL THEORY STUDY OF THE INITIAL OXIDATION OF THE PLATINUM (111) SURFACE .......................................................................... 21 Introduction ............................................................................................................. 21 Calculation Details .................................................................................................. 25 Results and Discussion ........................................................................................... 26 O on 22Pt(111) .............................................................................................. 26 O on 44Pt(111) .............................................................................................. 34 Summary and Conclusions ..................................................................................... 51 3 STABILITY AND GROWTH OF PLATINUM OXIDE CHAINS ON PLATINUM (111) ....................................................................................................................... 54 4 STRUCTURE OF THE PALLADIUM OXIDE (101) THIN FILM ON PALLADIUM (111) ....................................................................................................................... 62 5 FIRST PRINCIPLES STUDIES OF THE ADSORPTION OF HYDROGEN ON THE OXIDES OF PALLADIUM (111) ...................................................................... 69 Introduction ............................................................................................................. 69 Calculation Details .................................................................................................. 71 Results and Discussion ........................................................................................... 74 H2 on PdO( 101) ................................................................................................ 74 Molecular chemisorption of H2 ................................................................... 74 Pathways for dissociative chemisorption of H2 ........................................... 81
6 H2 on Pd(111) and Pd(110) .............................................................................. 86 H2 on Pd5O4/Pd(111) ........................................................................................ 94 H2 on Pd5O4/Pd(111) versus H2 on Pd5O4 ........................................................ 98 H2 on PtOx/Pt(111) ......................................................................................... 101 6 MOLECULAR ADSORPTION OF SMALL ALKANES ON A PALLADIUM OXIDE (101) THIN FILM: EVIDENCE OF SIGMA COMPLEX FORMATION ................... 104 Introduction ........................................................................................................... 104 Computational Details ........................................................................................... 108 Results and Discussion ......................................................................................... 110 Methane Adsorption on PdO(101) .................................................................. 110 Ethane and Propane Adsorption on PdO(101) ............................................... 114 Comparison between Experiment and Computation ...................................... 117 Methane Dissociation on PdO(101) ................................................................ 118 Methane Adsorption on Modified PdO(101) ................................................... 120 Summary .............................................................................................................. 122 7 CONCLUSIONS ................................................................................................... 124 LIST OF REFERENCES ............................................................................................. 126 BIOGRAPHICAL SKETCH .......................................................................................... 133
7 LIST OF TABLES Table page 2 1 Binding energies of subsurface O and surface O configurations on 22Pt(111) ................................................................................................................ 28 2 2 Binding energies of surface and subsurface O configurations on 44Pt(111). .. 39 2 3 Selected structural data for bare Pt(111), buckled Pt atoms along oxide PtO2(0001). .................................................................................. 48 2 4 PtO2. ................................................................................................................... 50 5 1 The vibrational frequencies of H2 adsorbed atop cus Pd perpendicular and parallel to the cus Pd rows. ................................................................................ 78 5 2 Selected properties of adsorbed H2 on Pd metal and PdO(101) surfaces .......... 89 6 1 Binding energies for CH4 adsorbed on modified PdO(101) surfaces in a 2(H,H) configuration. ....................................................................................... 121
8 LIST OF FIGURES Figure page 2 1 Illustration of the surface and subsurface adsorption sites on Pt(111) ............... 27 2 2 Top views of the most stable O surface atom configuration at 0.25 ML, 0.50 ML, 0.75 ML, and 1 ML found on the 22 surface unit cell. ................................ 29 2 3 Top and side views of the most stable O subsurface atom configuration at 0.50 ML, 0.75 ML, and 1 ML found on the 22 surface unit cell. ........................ 31 2 4 Binding energy as a function of O coverage on 22Pt(111) for the most stable surface and subsurface configurations. ................................................... 32 2 5 The MEP for an O atom diffusing from surface fcc to subsurface tetraI site at 0.25 ML and 0.50 ML. ......................................................................................... 33 2 6 Top and side views of the most stable O surface and subsurface atom configuration at 0.5625 ML found on the 44 surface unit cell. .......................... 36 2 7 The MEP for an O atom diffusing from surface fcc to subsurface tetraI site at 0.25 ML, 0.50 ML, and 0.5625 ML on the 44 surface unit cell. ......................... 37 2 8 Top views of the most stable surface O atom configurations at 0.625 ML found on the 44 surface unit cell that illustrate the preference f or O atoms added post 0.50 ML to cluster between p (21) rows. ......................................... 40 2 9 Top and side views of the most stable surface O atom configurations at 2/3 ML on the 33 surface unit cell and 0.6875 ML on the 44 surface unit cell. ..... 41 2 10 Top and side views of t he most stable surface O atom configurations at 0.75 ML found on the 44 surface unit cell and the 22 surface unit cell. .................. 43 2 11 Top and side views of the most stable surface O atom configuration at 1 ML found on the 22 surface unit cell, with occupying fcc sites and occupying hcp sites in a PtO2like structure PtO2(0001). ......................... 46 2 12 Binding energy as a function of O coverage on 22and 44Pt(111) for the most stable surface and subsurface configurations. ........................................... 47 3 1 An illustration of the original and redefined surface unit cells to more easily allow the study of the growth and stability of Pt oxide chains. ............................ 55 3 2 Top and side views of Pt oxide chains at 0.525 ML and 0.725 ML. .................... 55 3 3 Added binding energy versus Pt oxide chain length. .......................................... 56
9 3 4 Ex amples of oxygen atoms added between the same p (2 1) rows as the Pt oxide chain and in the empty space between adjacent p (2 1) rows. ................. 57 3 5 STM images of O on Pt(111) at 200 mV and 1 nA: simulated, 9/16 ML O on Pt(111), and experimental, 0.5625 ML O on Pt(111). ......................................... 58 3 6 Top and side views of CO, NO, and O2 adsorbed at the ends of a Pt oxide chain at 0.650 ML O on 2Pt(111). ........................................................... 59 3 7 Top and side views of a possible mechanism for NO oxidation via coadsorbed O2 at the end of a Pt oxide chain at 0.650 ML O on 2 Pt(111). ............................................................................................................... 61 4 1 Top and side view of the PdO(101) thin film structure. ....................................... 62 4 2 Illustration of the rotation of the original PdO(101) unit cell to align the O cus Pd O along the xaxis ([ 101 ] direction). ..................................................... 64 4 3 Projected density of states for a four fold coordinated Pd atom on the PdO(101) surface before and after a clockwise rotation of the unit cell. ............. 66 4 4 Projected density of states for threefold coordinated Pd, threefold coordinated O, and four fold coordinated O atoms on a PdO(101) surface. ........ 67 5 1 Top and side views of two stable configurations of molecular hydrogen adsorbed on PdO(101) ....................................................................................... 75 5 2 Charge density difference plot for H2/PdO(101) with H2 parallel to cus Pd rows. ................................................................................................................... 78 5 3 Projected density of states of Pd d states and H states for H2 adsorbed on PdO(101). ........................................................................................................... 80 5 4 The top and side view of H adsorption on top cus O, top cus Pd and between cus Pd sites on the PdO(101) surface. ................................................. 82 5 5 The three identified final states for H2 dissociation from H2 adsorbed in a parallel orientation on the cus Pd rows. .............................................................. 83 5 6 The pathway for H2 dissociation into H atoms bonded to atop cus Pd and atop cus O. ......................................................................................................... 85 5 7 Top views of two equivalent stable configurations of H2 adsorbed atop a surface Pd on Pd(111); t op and side views of two stable configurations of H2 adsorbed atop a single Pd atom on Pd(111). ..................................................... 87 5 8 Top and side views of two stable configurations of H2 adsorbed on Pd(110). .... 88
10 5 9 Charge density difference plot for H2/Pd(111), H2/Pd/Pd(111), and H2/Pd(110) with H2 parallel to cus Pd rows. ........................................................ 90 5 10 Projected density of states of Pd d states and H states for H2 adsorbed on Pd( 111), Pd/Pd(111), and Pd(110). .................................................................... 91 5 11 Molecular H2 adsorption energy versus Pd d band center both before and after interaction with H2. ...................................................................................... 93 5 12 Top views of stable configurations of dissociated H2 on Pd5O4/Pd(111). ........... 95 5 13 Illustrations of the local geometries of PdO(101) and Pd5O4 both before and after H2 adsorption. ............................................................................................. 96 5 14 Projected density of states of Pd d states and H states for H2 adsorbed on Pd5O4/Pd(111). ................................................................................................... 97 5 15 Charge density difference plots for bare Pd5O4/Pd(111) illustrating the interaction between the Pd5O4 layer and underlying Pd(111). ............................ 99 5 16 Top views of two additional stable configurations of dissociated H2 on Pd5O4 with the underlying Pd(111) substrate removed. .............................................. 100 5 17 Charge density difference plots for H2 fixed on Pd5O4/Pd(111) both with and without the underlying Pd(111). ........................................................................ 100 5 18 Top views of three stable configurations of dissociated H2 at the end of PtOx chains/Pt(111). ................................................................................................. 101 5 19 Charge density different plots for H2/PtOx chains/Pt(111) with H2 fixed to mimic the geometry of H2 adsorbed on PdO(101). ........................................... 102 6 1 Top and side view of the 2(H,H) configuration of CH4 adsorbed on PdO(101) determined by DFT. .......................................................................................... 111 6 2 Charge dens ity difference plot of the plane defined by the upward oriented H C H bond angle for CH4 on PdO(101) in the 2(H,H) configuration; pDOS of CH4 states and Pd d states for the 2(H,H) configuration of CH4 on PdO(101) 112 6 3 Top and side view of the favored 2(H,H) configurations of C2H6 and C3H8 adsorbed on PdO(101) determined by DFT. .................................................... 115 6 4 Top and side view of the 1(2H) configuration of C3H8 adsorbed on PdO(101) determined by DFT ........................................................................................... 116 6 5 Initial, transition, and final states for the dissociation of CH4 into CH3 and OxH groups on PdO(101). .................................................................................. 119
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FIRST PRINCIPLES STUDIES OF THE OXIDATION AND REACTIVITY OF Pt(111), Pd(111), AND THIN FILM Pd OXIDES By Jeffery Michael Hawkins May 2010 Chair: Aravind Asthagiri Major: Chemical Engineering Density functional theory (DFT) calculations have been used to examine the initial stages of oxidation of the Pt(111) surface, the adsorption of hydrogen on Pd(111) and its oxides, the dissociative chemisorption of hydrogen on PdO(101), and the molecular adsorption of methane, ethane, and propane on P dO(101). In the initial stages of oxidation of Pt(111), we predict that subsurface oxygen is not the precursor to oxidation, but instead, there is a strong preference for the formation and growth of onedimensional Pt oxide chains As oxygen atoms aggregate between the c lose packed oxygen rows, they induce large buckling and charge modification of the surface Pt atoms, and the resulting oxide compound grows as a onedimensional chain running parallel to the oxygen rows of the p (2 1) structure Our results agree well with a recent scanning tunneling microscopy ( STM ) study and suggest a novel precursor mechanism to the oxidation of metal surfaces involving Pt oxide chain formation and growth on terrace s at moderate oxygen coverages Unfortunately, DFT calculations are insu fficient to fully explore the behavior of such oxide chains due to system size limitations The development of accurate charge transfer potentials will be useful in this pursuit.
12 DFT calculations predict that H2 binds relatively s trongly on both Pd(111) and PdO(101) by forming complexes on coordinatively unsaturated Pd sites The nature of the bonding between the H2 molecule and metal center is explored. A single pathway for H2 dissociation is identified that generates stable products on PdO(101) W e also determine that quantum mechanical tunneling dominates the dissociation of H2 on PdO(101) at low temperature and that differences in tunneling rates are responsible for the large kinetic isotope effect that is o bserve d experimentally. DF T calculations also predict that small alkanes bind on PdO(101) by forming dative bonds with coordi natively unsaturated Pd atoms, and t he resulting adsorbed species are analogous to alkane complexes We conclude that both the dispersion interaction and the formation of complexes contribute to the binding of small alkanes on PdO(101) and estimate that complex formation accounts for 3050% of the total binding energy for the molecules studied. The predicted weakeni ng of C H bonds resulting from comp lex formation may help to explain the high activity of PdO surfaces toward alkane activation.
13 CHAPTER 1 INTRODUCTION Motivation Understanding the oxidation behavior of platinum, palladium, and other late transition metals, as well as their oxides, is of fundamental scientific interest because of their many applications in heterogeneous catalysis These metals are used in several industrially releva nt catalytic processes, for example, in the selective oxidation of methane [1 12] in lean gas turbines or carbon monoxide  in automobile catalyt ic converters Under typical operating conditions of such processes, which is often in an oxygenrich environment, the surface of a transition metal catalyst can be in a variety of oxygencovered states which are distinctly dif ferent from dilute chemisorbed oxygen layers and thick bulk oxides Thus, the extent to which the surface is oxidized can significantly affect its catalytic performance since these various states often exhibit dramatically different chemisorptive and reac tive properties As such, pursuing a more detailed understanding, especially at the atomic level, of the changes that occur on the transition metal surface under oxidizing conditions would be useful in interpreting and optimizing catalyst behavior under r elevant reaction conditions Fundamental surface science studies in ultrahigh vacuum (UHV) of the oxidation of transition metal surfaces are complicated by the low sticking coefficient of O2, which results in the so called pressure gap  Using molecular oxygen, surface oxygen coverages are restricted to 0.25 ML, which is an insufficient coverage to probe the transition from a surface chemisorbed oxygen phase to the formation of surface oxide(s) The coverage limit can be circumvented by the use of more aggressive oxidants, such as NO2  ozone  atomic oxygen  or highpressure cells
14  Scanning tunneling microscopy (STM) has had a particularly significant impact on our understanding of oxide phases on several transition metal surfaces Recently, combined STM and DFT studies have resolved 2D oxide phases on Rh(111) [27 ,28] and Pd(111) [29,30] In fact, intermediate oxygen phases on late transition metals have been the subject of several recent studies of C O oxidation [17,20,3133] and NO oxidation [34,35] but as one might imagine, having the correct atomic level model of the various oxygen phase s is essential to obtaining meaningful results from studying processes such as CO and NO oxidation. Oxides of late transition metals can be quite effective as catalysts for promoting complete oxidation reactions In particular, palladium oxide (PdO) is sp ecial among late transition metal oxides as it is exceptionally active toward the complete oxidation of alkanes In fact, prior studies conducted at commercially relevant pressures demonstrate that the formation of PdO is responsible for the exceptional activity of supported Pd catalysts in the catalytic combustion of natural gas in excess oxygen [1 12] Its high activity provides substantial motivation for initiating more detailed investigations into the chemistry of these oxide surfaces, and i n general, clarifying their chemical properties is instrumental in understanding the catalytic behavior of transition metals in oxidizing environments. Fundamental surface science studies, particularly those utilizing first principles methods, will be advantageous in the development of next generation heterogeneous oxidation catalysts As current trends are drifting away from traditional oxidesupported platinum group metal catalysts towards alloys and composites, traditional trial anderror methods of cat alyst design are quickly becoming insufficient and ineffective in keeping
15 up with all of the possible compositional combinations Instead, first principles methods are becoming increasingly popular as an effective and efficient means for designing and dev eloping novel catalytic materials. Another issue in performing fundamental surface science studies is the so called materials gap. Catalyst powders, like any other polycrystalline material under industrially relevant conditions, exhibit multiple exposed f acets Technically, in order to fully capture the behavior of a particular catalyst, one must study an infinite number of possible surfaces Since one cannot easily perform this task, it is often advantageous to study low Miller index surfaces Fortunat ely, the fractional representation of the various exposed facets is determined by the thermodynamic stability of each Miller index plane, and assuming the metal is the catalytically active surface, examining the properties of the close packed, low Miller index planes provides considerable insight into the overall activity of a catalyst. Density Functional Theory Density functional theory (DFT) calculations have quickly become a useful tool in evaluating a variety of problems in chemistry, physics, materials science, and many branches of engineering [36,37] especially with recent advances in the accuracy of functionals, efficiency of algorithms, and computational power For example, DFT calculations have been used to elucidate the catalytic synthesis of ammonia on metal nanoparticles  the embrittlement of copper by traces of bismuth impurit ies  and probing material properties of magnesium silicate, a mineral important in planet formation, under extreme conditions  These are just a few examples of the many instances where information about a physical, macroscopic problem can be gained by exploring materi als at atomic length scales a process that is often quite challenging to
16 accomplish experimentally Computational methods, such as density functional theory calculations, are wellsuited to this task, but researchers utilizing these methods must remain mindful of the theory behind the calculations as well as both their capabilities and limitations A brief overview of the fundamentals of density functional theory is provided i n the remainder of this chapter. Schr dinger Equation At the atomic level, materials are represented by both a positively charged nucleus and negatively charged electrons Since the mass of the nucleus is orders of magnitude larger than the mass of an electron the mass of a proton or neutron is roughly 1800 times greater than an electron the nucleus is significantly slower than electrons in responding to changes in its surroundings As a result, the motion of the nucleus can be separated from that of the electrons and the respective mathematical problems can be solved separately This ability is the basis of the BornOppenheimer approximation  Thus, only electrons interacting in the nuclear potential and electronelectron interactions remain. This approximation is used to simplify the time independent Schr dinger equation (Equation 11). = (1 1) The particular definition of the Hamiltonian, depends on the nature of the system being studied. For simple systems like a particle in a box or a harmonic oscillator, the Hamiltonian takes a simple form, and the Schr dinger equation can be solved exactly Where multiple electrons are interacting with multiple nuclei, such as in a typical solid state system, the Hamiltonian takes a more complicated form, as shown in Equation 12.
17 = 22 2 + ( ) +1 2 2 (1 2) The terms in the Hamiltonian are the electronic kinetic energy, the energy of the interaction between each electron and all nuclei, and the energy of the interaction between electrons, respectively For this Hamiltonian, is the electronic wave function, and is the groundstate energy. In 1964, the modern formulation of density functional theory was born when Hohenberg and Kohn pr oved two fundamental theorems  First, they illustrated that all properties including the groundstate energy, of a quantum many body system can be considered as unique functional s of the groundstate electron density As a result, instead of requiring wave functions of 3N variables to solve the Schr dinger equation, only the electron density, a function of three spatial variables is required. Second, they showed that the electron density that minimizes the total energy is indeed the true electron density that corresponds to the solution of the Schr dinger equation. Unfortunately, the true functional form of the electron density is unknown, so approximate forms of the functional are used together with this variational principle to determine an approximate solution to the Schr dinger equation. KohnSham Equations Shortly after Hohenberg and Kohn published their arguments, Kohn and Sham provided a formulation of density functional theory  that has become the foundation for many of the current methods of tackling the full many body electron problem They determined a way to express the true electron density such that each equation only involves a single electron, which is shown in Equation 13. 22 2+ ( ) + ( ) + ( ) ( ) = ( ) (1 3)
18 O f the three potentials that appear on the left hand side of the equation, only the first potential, ( ), also appeared in the full Schr dinger equation; it defines the interaction between each electron and all nuclei The second potential, ( ) is the Hartree potential, which describes the Coulomb repulsion between one electron and the total density of all electrons, including itself The third potential, ( ) defines exchange and correlation contributions. In order to solve the KohnSham equations, an iterative process is necessary, which often begins by defining an initial, trial electron density, and then solving the KohnSham equations using this trial density to determine the singleelectron wave functions The electron density is then calculated using these wave functions If the two electron densities match, then the groundstate electron density has been obtained and can be used to calculate the total energy If the two electron densities differ, then the trial density must be changed in some way, and the process is repeated. This procedure ensures a self consistent solution to the KohnSham equations. ExchangeCorrelation Functionals The exact form of the exchangecorrelation functional ( ) is still unknown, even though its existence is guaranteed by the HohenbergKohn theorem Thus various approximations must be made in order to completely define the KohnSham equations, and subsequently, obtain an approximate solution to the Schr dinger equation. The local density approximation (LDA), shown in Equation 14, assumes that the exchangecorrelation potential is the known exchangecorrelation potential of a homogeneous electron gas at the observed electron density. ( ) = [ ( ) ] (1 4)
19 In practice, LDA has been shown to be more suited for systems with slow ly varying electron densities, such as those like a nearly free electron metal LDA overestimates bond energies, and as a result, predicts shorter bond lengths than those observe d experimentally, but it often agrees well with structural and vibrational experimental data. The generalized gradient approximation (GGA), shown in Equation 15, is similar in form to the LDA, but it assumes that the exchangecorrelation potential also depends on the gradient of the electron density. ( ) = [ ( ) ][ 1 + ( ) ] (1 5) Here, ( ) is a g radient of the electron density (a Taylor expansion, for example) and is useful in describing systems where the electron density varies more sharply, such as surfaces and small molecules In general, GGA is more accurate than LDA, but it often underestimates bond energies, and as a result, overestimates parameters such as lattice constants Because the gradient of the electron density can be included in different ways, many variations of the GGA functional exist The two most common are the Perdew Wang (PW91) functional  and the Perdew Burke Ernzerhof (PBE) functional  Many other approximations for the exchangecorrelation functional exist, each with their advantages and disadvantages, though LDA and GGA are the most widely used. The average density approximation (ADA) and weighted density approximati on (WDA) are two nonlocal density approximations, and several orbital dependent functionals exist, such as LDA+U  Additionally, newer hybrid functionals, such as B3LYP, are most accurate for computing energetics but are more computationally demandi ng than calculations utilizing LDA or GGA.
20 Pseudopotentials One means of expanding the electronic wave functions is through the use of a series of plane waves In order to represent all of the electrons associated with a particular atom, very large plane wave sets would be required. Core electrons, however, are not generally important in chemical bonding or defining material properties Thus, the use of pseudopotentials to approximate the core electrons and reduce the number of plane waves required, whil e maintaining the character of valence electrons, is advantageous The electron density of the core electrons is replaced with a smoothed density that approximates the properties of the true ionic core, such as the nuclear attraction and inner electron repulsion, and is then fixed. Calculations using this frozen core approximation are more common than all electron calculations Two classes of pseudopotentials exist: ultrasoft pseudopotentials (USPP) and projector augmented wave (PAW) pseudopotenials For many systems, these methods give very similar results; however, in systems with strong magnetic moments or large differences in electronegativity, the PAW pseudopotentials are more reliable.
21 CHAPTER 21DENSITY FUNCTIONAL T HEORY STUDY OF THE INITI AL OXIDATION OF THE PLATINUM (111) SURFACE Introduction Understanding the oxidation behavior of platinum and other late transition metals (TM) is important since these metals are used in several industrially important oxygenrich catalytic processes, such as the selective oxidation of organic compounds, oxidation of NO to NO2, and combustion exhaust remediation. An atomic level understanding of the changes to the TM surface under oxygenrich conditions would be useful in interpreting and optimizing catalys t behavior under reaction conditions Fundamental surface science studies in ultrahigh vacuum (UHV) of the oxidation of Pt surfaces are complicated by the low sticking coefficient of O2, which result s in the so called pressure gap  Using molecular oxygen, surface oxygen coverages are restricted to 0.25 ML, which is an insufficient coverage to probe the transition from a surface chemisorbed oxy gen phase to the formation of surface oxide(s) The coverage limit can be circumvented by the use of more aggressive oxidants, such as NO2  ozone  atomic oxygen  or highpressure cells  STM has had a particularly significant impact on our understanding of oxide phases on several TM surfaces Recently, combined STM and DFT studies have resolved 2D oxide phases on Rh(111) [27,28] and Pd(111) [29,30] Several groups have probed oxygen on the Pt(111) surface in UHV with temperature programmed desorption (TPD), X ray photoelectron spectroscopy (XPS), and low energy el ectron diffraction (LEED) [22,25,26] as well as by using a combination 1Reproduced with permission from J.M. Hawkins, J.F. Weaver, A. Asthagiri, Phys. Rev. B 79 (2009) 125434. Copyright 2009 The American Physical Society.
22 of DFT  and cluster expansion based Monte Carlo (MC) [49,50] At coverages below 0.25 ML, the O atoms adsorb on fcc hollow sites and form a p (22) surface structure This structure maintains maximum distance between the O atoms at 0.25 ML coverage and minimizes the O O repulsion. DFT calculations have shown that the O binding energy declines wi th increasing O surface coverage  Subsequent increases in the surface coverage abov e 0.25 ML show the development of a series of different O phases as indicated by differences in the O2 desorption TPD spectra [22,25,26] TPD data show three unique states, labeled 123, which appear as the surface coverage is increased from 0 to 0.75 ML. The aforementioned p (22) 3) appears first, and the peak associated with this state shifts downward in temperature as the coverage is increased from 0 to 0.25 ML, indicative of the O O repulsion. As the coverage is increased beyond 0.25 ML, a second peak ( 2) starts to 3 peak and grows in intensity from 0.25 to 0.5 ML. At the 2 peak starts to appear, the LEED pattern associated with the p (22) starts to fade and disappears rapidly above 0.5 ML. Unfor tunately, a clear structure cannot be resolved bas e d on the LEED for the 2 phase, but researchers have postulated two possible structur es [22,25,51] In the first possible structure, the O atoms continue to occupy fcc sites and arrange into a p (21) structure The (22) LEED pattern can match this structure if three rotationally degenerate p (21) domains co exist [22,25] In the alternative structure, additional O atoms adsorbed on the p (22) structure bind on hcp hollow sites, producing a honeycomb structure. TPD spectra obtained using a mixture of 18O and 16O atoms provide evidence that oxygen atoms adsorb on different sites above 0.25 ML  This result suggests that the honeycomb
23 structure is associated with the 2 state DFT calculations by several groups contradict this conclusion and predict t hat the p (21) structure is favored over occupation on the hcp site by 0.5 eV  A recent STM study has resolved this controversy  revealing p (21) domains of oxygen atoms below 0.5 ML. We will return to the results from that work later since it plays an important role in our approach to examining O surface structures in this chapter. Further increases in the surface coverage from 0.5 to 0.75 ML introduce yet a new desorption peak ( 1), which appears at a l ower temperature than the other two peaks The atomic 1 state was unknown until the recent STM work by Devarajan et al.  Several DFT studies have attempted to resolve the behavior of O beyond 0.25 ML and provide the missing structural information about higher coverage oxygen phases on Pt(111) The approach has been to generate all possible structures of oxygen on the Pt(111) surface and evaluate the energetics (or thermodynamics) of the various structures To make these studies practical, restrictions must be made on the possible structures for the oxygen adlayer The primary assumptions have been to restrict the structures to chemisorbed O on the surface, and in some studies, to restrict the O binding sites to the fcc hollow As noted above, DFT calculations indicate that the hcp hollow is not favored over the fcc hollow at 0.5 ML, and test calculations at higher coverages seem to support this assumption [46,49,50] While the majority of the DFT studies of O/Pt(111) focus on surface oxygen, Lgar examined structures with a mixture of surface and subsurface oxygen  Todorova and co workers had demonstrated using DFT that subsurface oxygen becomes favorable on Pd(111) at a coverage between 0.5 and 0.75 ML  Lgar predicts that subsurface oxygen
24 becomes favored on Pt(111) at a coverag e between 0.5 and 0.75 ML  As we will discuss later, we have performed similar calculations in this study that contradict the result of Lgar. A recent DFT study compared ch emisorbed O structures to a 2D PtO2(0001) oxide film rotated 30 on Pt(111) and found the oxide film to be more stable at temperatures below around 800 K [32,54] An ab initio thermodynamics study of the PtO2(0001) oxide and various surface O coverages on Pt(111) PtO2(0001) oxide surface is more stable than fcc surface O coverages at oxygen chemical potentials close to the transition between p (22) and p (21) O surface configurations on Pt(111)  The finding that a 2D oxide is favored over chemisorbed O suggests a transition from surface O to an oxide phase, but the details of the transition coverage and the structures that form in the intermediate stages have not yet been explored computationally In this study, we perform DFT calculations to probe the energetics of various oxygen structures on Pt(111), and guided by recent STM results, w e have discovered new structures that are substantially preferred to the structures that have been reported to date. Furthermore, we have probed the kinetics for O atoms to diffuse into the subsurface We show that the initiation of oxide growth on the P t(111) surface does not proceed by subsurface diffusion, but instead an unexpected clustering of the O atoms occurs beyond 0.5 ML with pronounced surface buckling that results in Pt oxide chain structures on Pt(111) Our results shed new light on the transition from a chemisorbed surface oxygen phase to 2D oxide structures on Pt(111), as they reveal that the formation of low dimensional oxide structures is favored over the development of highdensity chemisorbed layers or the occupation of subsurface sites
25 Calculation Details All of the DFT calculations in this chapter are performed using VASP (Vienna ab initio simulation package)  We use the projector augmented wave (PAW) pseudopotentials [59,60] provided in the VASP database Calculations have been done using the GGA PBE approximation for the exchangecorrelation functional  We have also tested select calculations using the PW91 functional and find the adsorption energy is about 0.1 eV/O atom higher than the PBE values Nevertheless, the differences between adsorption e nergy of configurations at the same coverage are sufficiently small that the choice of functional does not affect the results reported in this chapter. In particular, the new oxygen phases that we identify and discuss later remain the most stable phase independent of the choice of GGA functional A planewave expansion with a cutoff of 400 eV is used, and the total energy calculations are done using a block Davidson iteration method for electronic relaxations, accelerated using Methfessel Paxton Fermi level smearing with a Gaussian width of 0.2 eV  The positions of the atoms are relax ed using a limited memory BroydenFletcher GoldfarbShanno ( LBFGS ) optimization method  until the forces on all unconstrained atoms are less than 0.03 eV/ All calculations are done using a five layer slab with a 22 (44) surface unit cell, which allows us to study O coverages in increments of 0.25 (0.0625) ML. The adsorbed O atoms and top three Pt layers are permitted to fully relax while the lower two Pt layers are held fixed. A vacuum region of about 13 ensures that the slab does not interact with its periodic image in the surface normal direction. We use a 441 and 221 Monkhorst Pack mesh  for the 22 and 44 surface unit cell, respectively We have confirmed that higher energy cutoffs and finer k point meshes have a negligible effect (< 0.01 eV) on the binding energies reported. For
26 example, using a 441 k point mesh for the 44 surface unit cell alters the differenc e in binding (total) energies of the all fcc and oxide chain configurations at 0.75 ML [shown in Figures 2 10 b and 2 10a, respectively] by 8104 eV/O atom (0.01 eV) The lattice parameter for Pt is found to be 3.977 using DFT versus an experimental val ue of 3.92  All calculations reported are performed while fixing the inplane lattice parameter to 3.977 Minimum energy pathways (MEP) and barriers for chemisorbed O atoms to diffuse into the subsurface are calculated using the climbing nudged elastic band (NEB) method  Results and Discussion O on 22Pt(111) We initially examined the energetics of O atoms in various configurations occupying surface and subsurface sites (illustrated in Figure 2 1) on the 22 surface unit cell of Pt(111) from 0.25 to 1 ML. We discuss the stability of the various oxygen configurations in terms of the binding energy per O atom ( Eb), which is given by Equation 21. = ,+ 2 2 / (2 1) T he bare Pt(111) slab and an isolated triplet O2 molecule are used as reference states, and n indicates the number of O atoms adsorbed on the Pt(111) surface. By our definition of Eb, a larger positive value indicates a more stable adsorption site. We do not examine the surface free energy as a function of oxygen chemical potential (i.e., the t hermodynamics) for this system because our interest is in identifying the most stable oxygen structure as a function of surface O coverage. The differences in vibrational and configurational entropy contributions between various O configurations at the sa me
27 O coverage are not expected to be large enough to change the conclusions that are drawn in this chapter [45 48] Figure 21 Illustration of the (a) surface and (b) subsurface adsorption sites on Pt(111), where 1 = fcc, 2 = hcp, 3 = atop, 4 = tetraI, 5 = tetra II, and 6 = octa. The 22 surface unit cell is shown in (a). The three possible surface adsorption sites are the two threefold hollow sites, the fcc (hcp) site, located above a Pt atom in the third (second) layer, and an atop site located directly above a Pt atom in the first layer We only consider the three highsymmetry i nterstitial sites located between the first and second Pt layers for the subsurface incorporation of O atoms There are two fourfoldcoordinated tetrahedral sites (tetraI and tetraII) and one sixfoldcoordinated octahedral (octa) site [see Figure 2 1b] The tetra I site is located below the hcp surface site, tetraII is below a first layer Pt atom, and octa is below the fcc surface site. The existence of several surface and subsurface sites leads to a multitude of possible configurations To limit th e total number of configurations, we make several restrictions, some of which are revisited when we examine O on a 44Pt(111) system discussed later The configurations can be split into two general types: surface and subsurface In the surface configur ations, all of the O atoms reside on the surface At each coverage, we have examined surface configurations where all O atoms reside in
28 the fcc, hcp, and atop sites, respectively, and mixed configurations consisting of ( n 1) O atoms on the fcc and 1 O atom on the hcp or atop sites For the subsurface configurations, ( n 1) O atoms are on the surface fcc sites, and 1 O atom is placed in one of the three possible subsurface sites Even with these restrictions, there are several possible configurations that need to be explored at each coverage. We note that our approach is very similar to the recent study of subsurface O on Pd(111)  Lgar performed a similar study of O/Pt(111)  but did not examine as many surface/subsurface configurations as we report in this chapter. Table 21 report s the binding energy for the most important configurations we have examined on the 22Pt(111) surface Configurations such as those with O atoms all on the hcp or atop sites are considerably less stable and are not reported in Table 21 The top views of the favored surface configurations at each coverage are shown in Figure 2 2 Figure 22b shows the p (21) surface configuration which was recently observed in STM experiments of oxygen on Pt(111)  There is a modest Pt buckling due to the presence of the surface O which peaks at 0.5 ML coverage before declining to zero at 1 ML. Table 21 Binding energies of subsurface O and surface O configurations on 22Pt(111) Sites total (ML) on / sub surface 0.25 0.50 0.75 1.00 ( n ) fcc / 1.21 0.90 0.52 0.15 ( n 1) fcc + 1 hcp / 0.78 0.60 0.34 0.11 ( n 1) fcc / tetra I 0.77 0.47 0.21 a 0.48 b 0.24 ( n 1) fcc / tetra II 1.74 0.25 0.26 a 0.01 b 0.13 ( n 1) fcc / octa 1.61 0.45 0.09 0.17 a,b = tetra I, II site; fcc/octa moves to hcp/octa
29 Figure 22 Top views of the most stable O surface atom configuration at (a) 0.25 ML, (b) 0.50 ML, (c) 0.75 ML, and (d) 1 ML found on the 22 surface unit cell The O atoms are represented by red, and the gray and blue represent the down and up buckled Pt atoms, resp ectively The O atom pulls up the nearest Pt atoms, and the Pt buckling is 0.13, 0.19, and 0.04 for 0.25, 0.50, 0.75 ML O coverage. Our results for the O surface configurations match previous DFT results [45,46,68,69] Specifically, the favored O surface configuration corresponds to O atoms residing in the fcc sites at all surface O coverages as shown in Figure 2 2 Configurations generated by placing one of the O atoms in the hcp site (referred to as ( n 1) fcc + 1 hcp in Table 21 ) are found to have smaller binding energies This difference in energy of pure fcc and configurations with fcc and 1 hcp decreases with increasing cove rage, but even at 1 ML, the all fcc configuration is favored by 0.04 eV/O atom From these results, several
30 studies i n the past have concluded that the only relevant surface site is the fcc hollow With this restriction and neglecting subsurface O atoms, the most stable O configurations can be identified  and it has been reported that O atoms reside in the fcc sites at maximum separation to reduce the O O repulsion. With the incorporation of subsurface O atoms, we can expect to have a transition from the preference of surface configurations at lower O coverages to the subsurface at sufficiently high coverage. The subsurface will become favo red when the surface O O repulsions become too large and the cost of the strain associated with subsurface O is sufficiently balanced by the screening of the O O interactions Table 21 reports the values for the most stable subsurface configurations we have identified, and two general features can be observed. Firstly, for all subsurface configurations, the tetraI subsurface site is the most favored. Figure 2 3 shows the favored subsurface configurations at each coverage from 0.50 to 1 ML. As expected, there is much more significant surface Pt buckling with the presence of subsurface O atoms The surface O atoms are screened from the subsurface O atoms by the Pt atoms, which reduces the O O repulsion. The second observation is that the crossover between surface and subsurface configurations occurs at some coverage above 0.75 ML, as illustrated by Figure 24, which plots the most favored surface and subsurface configurations as a function of ML of surface O At low coverages (0.25 ML), the difference between surface and subsurface is quite large, and in fact, the subsurface O configurations are not stable. As the coverage is increased, the difference between surface and subsurface starts to drop, and by 0.75 ML coverage, the difference in binding energy is 0.04 eV/O atom.
31 Figure 23 Top and side views of the most stable O subsurface atom configuration at (a) 0.50 ML, (b) 0.75 ML, and (c) 1 ML found on the 22 surface unit cell The O atoms are represented by red, and the gray and blue represent the down and up buckled Pt atoms, respectively The surface (subsurface) O atom pulls (pushes) up the nearest Pt atoms, and the Pt buckling is 0.62, 0.84, and 0.69 for 0.50, 0.75, 1 ML O coverage.
32 Figure 24 Binding energy as a function of O coverage on 22Pt(111) for the most stable surface ( At 1 ML, the subsurface is favored by 0.09 eV/O atom, but for both the surface and subsurface, the binding energy decreases from the 0.75 ML value. We note that Lgar predicts the transition from surface to subsurface on Pt(111) to occur beyond 0.5 ML. We have examined Lgars reported subsurface structure at 0.75 ML, which is also a tetra I subsurface configuration, and find differences in the buckling between the two structures We have used Lgars structure as input, but relaxation reproduces our most favored arrangement and energy The differences in the parameters of the DFT calculations are insufficient to cause this discrepancy, so the reasons for the structural differences are unclear We note that the difference in binding energy between the
33 surface and subsurface configurations at 0.25 and 0.5 ML are in relatively good agreement with the values reported in Lgars study. We conclude the discussion of our results on the 22Pt(111) system by reporting the minimum energy pathway ( MEP ) and energy barriers to an O atom diffusing from the surface to the subsurface The barriers and paths for subsurface O at 0.25 and 0.5 ML coverage are shown in Figure 2 5 Figure 25 The MEP for an O atom diffusing from surface fcc to subsurface tetraI site at 0.25 ML ( The barrier for the fcc to hcp hop at 0.25 (0.50) ML is 0.61 (0.74) eV, and the barrier for diffusion from the hcp surface site to the tetraI subsurface site at 0.25 (0.50) ML is 2.23 (3.19) eV. The O atom hops from the fcc to the hcp site and then diffuses into the subsurface tetraI site The barrier is quite large for this process at 2.23 (3.19) eV at 0.25 (0.50) ML
34 cover age. Our value at 0.25 ML agrees with a recent calculation by Gu and Balbuena  We attempted to determine the barriers at 0.75 and 1 ML, but the NEB calculations failed to converge. This failure to converge is due to the very unfavorable transition stat e structures at these coverages Using a 22Pt(111) system causes considerable interactions between the diffusing O atom and its periodic image. A truer measure of the energy barriers for subsurface O can be obtained using larger unit cells We revisit the barriers to subsurface O in the next section when we examine O on a 44Pt(111) surface unit cell. O on 44Pt(111) To examine more complex configurations, we turn to a 44Pt(111) surface unit cell. With a 44 surface cell, we can probe increments i n coverage of 0.0625 ML. Furthermore, configurations where the O atoms cluster can be more readily explored in the larger system The need to explore other configurations is motivated by recent STM images of oxygen on Pt(111) from 0.25 to 0.75 ML. In co ntradiction with our earlier conclusions and other DFT studies [45,46,68] Devarajan and co workers observed O atom clustering and chain f ormation with pronounced surface buckling at total coverages as low as 0.4 ML  As the coverage is increased past 0.5 ML, Y shaped structures consisting of oxide chains appear within the p (21) domains The STM results reveal that the true minimum energy configurations have not yet been identified by DFT. Rather than attempting every possible combination on the 44Pt(111), we use a different approach based on the viewpoint of adding O atoms to a configuration which is in agreement with experimental results, namely, the p (21) confi guration at 0.5 ML. The STM results clearly show that the p (21) structure is the dominant phase at total coverages near 0.5 ML and that the new chain structures nucleate within the p (21)
35 domains We have confirmed that the energy of the configurations reported in the previous section for the 22Pt(111) system can be reproduced exactly on the 44Pt(111) system The first set of configurations that we explored is at 0.5625 ML coverage. At this coverage, one O atom was added to the p (21) structure and placed on the fcc, hcp, and atop sites, as well as the three subsurface sites The favored surface and subsurface configurations at 0.5625 ML are shown in Figure 2 6 with the added O atom clearly marked. For the subsurface configuration, the preferred site is still the tetra I The surface configuration is favored by 0.07 eV/O atom over the subsurface, but more interestingly, the O atom resides near an hcp site instead of an fcc site in the favored surface configuration. The additional surface O atom near the hcp site is favored over the fcc site by 0.04 eV/O atom As seen in the favored surface configuration [ Figure 2 6a], adding an oxygen atom between the p (21) oxygen rows produces two Pt atoms that are threefold coordinated with oxygen atoms, two of which originate from a close packed row of the p (21) structure Interestingly, these threefold coordinated Pt atoms undergo tremendous upward buckling (1.44 ), in good agreement with recent STM images that indicate a buckling of 1.7  The side view in Figure 2 6 a shows clearl y that the buckled Pt atoms can serve to screen the O O interactions on the surface As elaborated below, the new surface structure that results from adding an oxygen atom to the p (21) structure represents the initial building block of an oxide chain com pound that preferentially develops above oxygen coverages of 0.50 ML. Our identification of an oxide chain compound can explain prior experimental TPD results which suggest that oxygen atoms begin to bind at sites other than the fcc hollow
36 after saturati on of the p (22) structure  Recall from the introduction that Jerdev et al. suggested the presence of O atoms in hcp sites above 0.25 ML based on TPD spectra obtained after sequentially adsorbing 18O and 16O atoms on Pt(111). However, instead of a honeycomb structure as the source of the hcp binding sites for oxygen atoms  the STM and our DFT results suggest that the different adsorption site is associated with the nucleation of O atoms at high local coverages within the p (21) structure. Oxygen atoms can begin to adsor b on local domains of p (21) slightly above 0.25 ML, but STM results suggest that such adsorption becomes more prevalent around 0.4 ML where large domains of p (21) coexist with remaining domains of the p (22) structure  Figure 26 Top and side views of the most stable O (a) surface and (b) subsurface atom configuration at 0.5625 ML found on the 44 surface unit cell The p (21) O atoms are represented by red while the additional O atom is orange. The gray and blue represent the down and up buckled Pt atoms, respectively T he additional O atom on the surface (subsurface) pulls (pushes) up the nearest Pt atoms, and the Pt buckling is 1.44 (0.75) We briefly return to the MEP for subsurface oxygen that are calculated on the 44Pt(111) surface cell Figure 2 7 shows the bar riers and MEP for subsurface oxygen at
37 0.25, 0.50, and 0.5625 ML. With a 44Pt(111) cell at a coverage of 0.25 ML, we can examine the more realistic scenario of subsurface coverages of 0.0625 ML instead of 0.25 ML on the 22Pt(111). The binding energy of such a low concentration of subsurface oxygen at total coverages of 0.25 and 0.5 ML are far more stable than the values reported for the 22Pt(111) system (see Figure 212 discussed below). Figure 27 The MEP for an O atom diffusing from surface fcc to subsurface tetraI site at 0.25 ML ( The barrier for the fcc to hcp hop at 0.25 (0.50) ML is 0.59 (0.69) eV, which is comparable to that on the 22 surface unit cell The barrier for diffusion from the hcp surface site to the tetra I subsurface site at 0.25 (0.50) ML is 1.43 (1.23) eV, which is notably lower than on the 22 surface unit cell At 0.5625 ML, the barrier for the fcc to buckled hcp hop is 0.44 eV, and the barrier from buc kled hcp to tetraI is 1.48 eV The transition state between fcc and buckled hcp is shown in the inset.
38 As on the 22Pt(111) system, the oxygen atom hops from fcc to hcp to subsurface tetra I at coverages of 0.25 and 0.50 ML, but the barriers are much lower on the 44 surface cell The barrier from hcp to subsurface drops from 2.23 (3.19) to 1.43 (1.23) eV for 0.25 (0.50) ML when we use the larger surface cell While we would expect that the 44 is sufficiently large to give the true barrier for one oxy gen atom to diffuse into the subsurface at these coverages, explicitly testing this prediction through larger surface cells is beyond the scope of this chapter. At 0.5625 ML, the barrier for the additional oxygen atom located between p (21) oxygen rows to move from the fcc to hcp site, which results in the buckled O structure shown in Figure 26a, is 0.44 eV The barrier to subsurface oxygen at 0.5625 ML is 1.48 eV, which is close to the value at 0.5 ML Given that the barriers for diffusion of oxygen into the subsurface are relatively large, these results suggest that a negligible amount of oxygen atoms will occupy subsurface sites on Pt(111) up to 0.5625 ML, and as we show below, the same holds true for higher coverages up to 1 ML. We have also examined the favored surface and subsurface configurations at coverages from 0.5 to 0.75 ML on the 44Pt(111) surface using a similar procedure as that outlined for 0.5625 ML. Our calculations reveal a strong preference for oxygen atoms to aggregate and grow a onedimensional (1D) Pt oxide chain compound that exists within the p (21) structure The binding energies obtained for various configurations on the 44Pt(111) surface are reported in Table 2 2 We first considered the addition of two oxygen atoms to the p (21) structure, which results in a coverage of 0.625 ML, and compared clustered versus nonclustered arrangements of the extra oxygen atoms (see Figure 28).
39 Table 2 2 Binding energies of surface a nd subsurface O configurations on 44Pt(111). Sites total (ML) on/ subsurface 0.25 0.50 0.5625 0.6250 0.6875 0.75 Non buckled Pt surface O configurations ( n ) fcc / 1.21 0.90 0.80 0.70 0.61 0.52 ( n 1) fcc + 1 hcp / 1.09 0.82 Buckled Pt surface O configurations ( n 1) fcc + 1 hcp / 0.84 ( n 2) fcc + 2 hcp / 0.83 c 0.75 nc ( n 3) fcc + 3 hcp / 0.80 ( n 4) fcc + 4 hcp / 0.82 c 0.62 nc Subsurface O configurations ( n 1) fcc / tetra I 0.73 0.67 a 0.78 b 0.68 a 0.77 b ( n 2) fcc + 1 hcp / tetra I 0.73 c 0.70 nc ( n 2) fcc / 2 tetra I 0.69 c 0.66 nc ( n 4) fcc + 3 hcp / tetraI 0.52 c 0.50 nc ( n 4) fcc / 4 tetra I 0.45 c 0.51 nc ( n 1) fcc / tetra II 0.67 a 0.63 b 0.70 a 0.63 b ( n 4) fcc + 3 hcp / tetra II 0.44 b a,b = tetra I, II site; c,nc = clustered, nonclustered We find that the additional O atoms prefer to bind on hcp sites and aggregate between the close packed oxygen rows formed by the p (21) structure (Figure 28a) The clustered surface configuration is favored over the all fcc and nonclustered configurations by 0.13 and 0.08 eV/O atom, respectively Similarly, clustering of the subsurface O atoms is slightly favored, but the surface configuration is found to be favored over subsurface by 0.1 eV/O atom As seen in Figure 28a, the clustered
40 surface configuration produces an oxide chain containing three outwardly displaced Pt atoms Thus, extending the Pt oxide chain in this case involves buckling of only one additional Pt atom, and results in a structure with two Pt atoms each with threefold Pt O coordination and one Pt atom with fourfold Pt O coordination. Figure 28 Top views of the most stable surface O atom configurations at 0.625 ML found on the 44 surface unit cell that illustrate the preference for O atoms added post 0.50 ML (orange) to cluster between p (21) rows (red) The gray and blue represent the down and up buckled Pt atoms, respectively Clustered O (a) is favored by 0.72 eV to noncluste red O (b d), which are all energetically equivalent. In contrast, generating one of the nonclustered configurations involves buckling of two new Pt atoms, which is more energetically demanding than displacing only one additional Pt atom out of the surface plane. Enhanced Pt O bonding in the fourfoldversus threefoldcoordinated Pt species may also provide a driving force for oxygen aggregation and chain growth.
41 Figure 2 9 b shows the most favored arrangement at 0.6875 ML associated with an extension of the oxide chain parallel to the p (21) rows This configuration can be compared with the most favored arrangement reported at 2/3 ML for an all fcc surface configuration  which consists of maximally separated O atoms residing on fcc sites ( Figure 2 9a) Figure 29 Top and side views of the most stable surface O atom configurations at (a) 2/3 ML on the 33 surface unit cell and (b) 0.6875 ML on the 44 surface unit cell. In (a), O atoms are arranged in a p ( 2O configuration; in (b), O atoms forming p (21) rows are red, and those added post 0.50 ML are orange. The gray and blue represent the down and up buckled Pt atoms, respectively The striped structure (b) is favored by 0.08 eV/O atom to the hexagonal structure (a) The Pt buckling in (b) is 1.75 Despite t he slightly greater surface coverage, the 0.6875 ML configuration has a 0.11 eV/O atom greater binding energy than the all fcc 2/3 ML configuration, demonstrating that oxide chain growth is favored over the all fcc configuration up to nearly 0.70 ML. In f act, the difference in energy between oxide chain structures and all fcc surface configurations increases at higher coverages, which indicates that the oxide chain
42 configurations become increasingly favored over maximally separated all fcc configurations as the coverage increases above 0.50 ML. The final coverage that we examined on the 44Pt(111) surface is 0.75 ML, and the clustered configuration is favored over the all fcc configuration by a substantial 0.3 eV/O atom. The increasing stability of the ox ide chain with added O atoms can be more clearly observed by the energy gained from adding an O atom We can def ine this added binding energy ( En add) by Equation 22. = / 1+ 22 / (2 2) In Equation 22, n and n 1 refer to the number of O atom s in the O/Pt(111) structure. En add is a measure of the stability gained by the system upon the addition of 1 O atom, where, by our definition, a positive value for En add indicates increased stability The added binding energy for the buckled structure at 0.5625 ML is 0.33 eV, which increases to 0.73 eV for the buckled structure at 0.625 ML. This increase suggests a large stability to growing the oxide chain from the initial nucleation. At 0.6825 ML, En add drops to a value of 0.52 eV which can be attributed to interactions with the periodic image in our relatively small unit cell To more accurately probe the stability of the chain with increasing length will require larger unit cells and is discussed in Chapter 3. In contrast to the stability gained by the oxide chains with O atom addition, when we examine En add values for all fcc surface configurations, we find a destabilizing effect upon addition of O atoms beyond 0.5 ML. At 0.5625 M L, the added binding energy is 0.03 eV for t he all fcc surface configuration, and this value steadily decreases with subsequent O atom addition to 0.44 eV at 0.75 ML.
43 Figure 2 10 shows a comparison between the oxide chain configuration and the allfcc configuration at 0.75 ML, and several differences are clear Firstly, the use of a 44Pt(111) surface cell allows us to arrange the four additional oxygen atoms in a chain configuration between the oxygen rows of the p (21) structure at 0.75 ML. Figure 210 Top and side views of the most s table surface O atom configurations at 0.75 ML found on (a) the 44 surface unit cell and (b) the 22 surface unit cell. O atoms forming p (21) rows are red, and those added post 0.50 ML are orange. The gray and blue represent the down and up buckled Pt atoms, respectively The striped structure (a) is favored by 3.59 eV /unit cell to the symmetric structure (b). The Pt buckling in (a) is 1.79 At the same coverage on the 22 Pt(111) system, we are restricted to adding one oxygen atom that results in a configuration where the oxygen atoms are separated to a maximum distance The result of this more separated configuration is that the surface does not buckle (see Figure 210b) In contrast, the clustered configuration results in massive buckling (1.79 ) of the Pt atoms located between the added row of oxygen atoms and the close packed oxygen row of the p (21) This buckling and formation of
44 an oxide chain allows the oxygen atoms to be screened from one another and reduces the O O repulsion that would be normally found on oxygen surface configurations The net effect of the buckling is a novel oxide chain configuration, which involves oxygen clustering instead of the separation of the oxygen atoms Furthermore, the predicted chain structures closely r esemble the chains that have been identified by STM at coverages greater than 0.40 ML. In particular, the STM images indicate oxide chain growth along p (21) rows and a buckling of 1.7 which compares well with our value of 1.79 As mentioned above, t he chain structures do not represent a new chemisorbed state but are better described as a Pt oxide chain compound that serves as a precursor to the formation of a full 2D or 3D oxide on the Pt(111) surface, as demonstrated in prior experimental studies  The nearly 2 buckling of t he Pt atoms out of the surface suggests that chain formation may be viewed as a chemical reaction involving the cleavage of Pt Pt bonds between the first and second layer Pt atoms and the concurrent formation of PtO bonds Moreover, the Pt atoms within t he chains experience fourfold coordination with oxygen atoms, forming square planar units that closely resemble the building blocks of bulk PtO, Pt3O4 PtO2. Thus, following Devarajan et al. we refer to the 1D metal oxide compound as Pt oxide chains  However, it is important to stress that the structures we have identified in our 44Pt(111) calculations are not a full representation of the experimental STM images obtained after the chains have grown appreciably In the DFT calculations, the periodic boundary conditions and t he use of a 44Pt(111) system result in an infinitely long Pt oxide chain at 0.75 ML running parallel to the p (21) rows In contrast, the STM results show that the chains aggregate and
45 eventually arrange into an interconnected network of Y shaped struct ures near 0.75 ML, where each leg of the Y structure consists of two to three sideby side Pt oxide chains with lengths between 19 to 24 Minimizing interfacial stresses must be responsible for the narrow distribution of chain lengths and the development of the chain superstructure. In particular, lattice mismatch between the Pt oxide chains and the Pt(111) substrate causes stress to build up along the Pt oxide chains Termination of the chain length and chain branching likely relieves this stress, res ulting in the narrow distribution of chain lengths and branched superstructure observed experimentally Unfortunately, however, the interfacial stresses cannot be relieved adequately in our DFT supercells due to the characteristic large length scales of t he chain superstructures Future DFT calculations with larger supercells will be needed to quantify these strain effects Interestingly, the 0.75 ML Pt oxide chain structure has a slightly larger binding energy than at 0.6875 ML (0.82 versus 0.80 eV/O at om), which suggests that the oxide chain should prefer to grow until the optimal length is reached to minimize strain effects Based on the finding of oxide chain growth on Pt(111), we revisited the 22Pt(111) system at 1 ML. Originally, as detailed in the previous section, we examined mixed surface configurations by considering ( n 1) fcc hollow sites and 1 hcp hollow site. Following this procedure, we found that the all fcc 1 ML configuration is favored over 3 O atoms on fcc hollows and 1 O atom on an hcp hollow site If instead we allow 2 O atoms to reside on fcc hollows (the p (21) structure) and add 2 O atoms to the hcp sites in between the p (21) rows, the resulting mixed surface configuration is more stable by 0.36 eV/O atom The resulting relax ed structure is shown in Figure 2 11a, and we can
46 observe the formation of a surface oxide phase that is quite distinct from chemisorbed oxygen atoms or subsurface oxygen (see Figures 2 2 and 2 3) The configuration shows substantial buckling of both O and Pt surface atoms and can be considered a 2D oxide phase. PtO2(0001) surface ( Figure 2 11b) shows that the two surfaces have very similar geometric structures PtO2(0001) oxide film is predicted by DFT to be the stable oxide structure on Pt(111) [32,54] Figure 211 Top and side views of (a) the most stable su rface O atom configuration at 1 ML found on the 22 surface unit cell, with occupying fcc sites and occupying hcp sites in a PtO2PtO2(0001) The O atoms are represented by red, and the gray and blue represent the down and up buckled Pt atoms, respectively The Pt buckling in (a) is 1.64 The line illustrates a 1D PtO2 chain. Figure 2 12 summarizes the binding energies of the most favored surface and subsurface configurations we have identified using DFT on both the 22and 44Pt(111) surface cells
47 Figure 212 Binding energy as a function of O coverage on 22and 44Pt(111) for the most stable surface ( The binding energies for 1 ML O/22Pt(111) with occupying fcc sit es and occupying hcp sites in a PtO2like structure ( Pt(111) [ p ( 2O] ( ) are also included. Interestingly, the binding energy for the Pt oxide structures drops quite sharply from 0.75 ML and 1 ML coverages The main difference between the structures of the 0.75 ML 44Pt(111) and 1 ML 22Pt(111) is the reduction of the distance between parallel Pt oxide chains (compare Figures 210a and 211a) which indicates a repulsive interaction between Pt oxide chains for the configuratio n shown in Figure 211a. However, because chain pairing is observed with STM at high coverages  the predicted destabilization suggests that an optimum arrangement of sideby side chains
48 is not obtained using the 22 unit cell For example, the oxygen atoms in adjacent chains lie in close proximity to one another in the 1 ML structure o btained with the 22 unit cell ( F igure 211a) PtO2(0001) surface (Figure 211b) suggests that the chainchain interaction could become favorable if one of the chains is displaced along the chain direction. Larger supercells are likely to be needed to obtain an optimum configuration of chain pairs, since the chains would be able to relax both parallel and perpendicular to the chain direction. The primary changes in the s tructure of the oxide chain with increasing coverage is captured in Table 2 3 where both the Pt Pt bond distances within the oxide chains and the Pt buckling are reported. Table 2 3 Selected structural data for bare Pt(111), buckled Pt atoms along oxide PtO2(0001) Pt Pt bond lengths are between buckled Pt atoms in the first Pt layer, and Pt buckling is the displacement in the surface normal direction of Pt atoms in the first Pt layer. Structure Pt Pt bond length () Pt buckling () Bare Pt(111) 2.812 0.00 0. 5 625 ML O / Pt(111) 2.940 1.44 0.625 ML O / Pt(111) 3.000 1.77 0.6875 ML O / Pt(111) 2.916 1.75 0.75 ML O / Pt(111) 2.812 1.79 1 ML O / Pt(111) 2.812 1.64 PtO 2 (0001) 3.082 A clear trend of the Pt Pt bond lengths approaching similar bond lengths to the bulk oxide surface can be observed as the surface oxygen coverage is increased beyond 0.5 ML to 0.75 ML. This observation reinforces the expectation that strain relief plays a large role in dictating the optimal chain lengths and oxide chain/Pt(111) substrate commensurability At 0.75 and 1 ML, the oxide structure is restricted to the inplane lattice constant of Pt(111) and therefore cannot expand laterally to the more optimal
49 lattice constant for the oxide chains Even with this artificial constraint, the chain structures are substantially more stable than chemisorbed oxygen structures at coverages above 0.50 ML. The buckling of the Pt does not dramatically change with coverage, except at 1 ML, where the buckling becomes slightly less pronounced, reflecting oxideoxide interactions perpendicular to the Pt oxide chains The buckling observed in our DFT results matches the value of 1.7 reported in the STM study  We conclude the chapter by examining the atomic charges on the O and Pt atoms in the various configurations We mentioned above that the new chain structures better resemble a Pt oxide compound rather than a new chemisorbed state of oxygen on Pt(111) This classification is ev ident from the geometric structure but should also be reflected in the electronic structure. Table 24 reports the atomic charge values, obtained from a Bader analysis of the charge density  for surface O and Pt atoms in various surface configurations as a function of coverage. Table 24 also PtO2 PtO2(0001) surface Since the PtO2 PtO2 are separated by about 4.5 and interact only through weak van der Waals forces  PtO2 PtO2. An examination of the charge on the O atom shows that the O atoms in the Pt oxide chain structures have slightly more negative charge than the all fcc nonclustered configurations The O atom charge in the clustered configurations increases with increasing coverage as these structures bec ome more Pt oxidelike, and at coverages of 0.75 and 1 ML, the O atom PtO2 oxide.
50 A much larger charge difference is found on the Pt atoms For the all fcc configurations, the small Pt buckling results in charge difference s between the up and down surface Pt atoms of approximately 0.35 electrons This charge difference between up and down Pt atoms is more pronounced for configurations with more fully developed Pt oxide chains (0.75 and 1 ML), where we find a 0.5 and 0.8 el ectron difference, respectively Furthermore, the up Pt atoms of the chains are more positively charged than those found i n the chemisorbed oxygen phase. Table 24 PtO2 (given in units of electrons) For O coverages above 0.50 ML where oxide chain structures are observed, values for qPt,down outside the parentheses indicate the atomic charges of Pt atoms surrounding the upwardbuckled Pt atoms that are coordinated to 2 O atoms, whi le values inside parentheses indicate the atomic charges of bulk like Pt atoms in the first layer that are coordinated to only 1 O atom Buckled Pt configurations are denoted with *. Coverage (ML) q O q Pt ,up q Pt ,down 0.25 0.75 0.25 0.05 0.50 0.76 0.54 0.24 0.5625 0.77 0.90 0.53 (0.24) 0.625 c lustered 0.78 0.87 / 1.17 0.54 (0.24) not clustered* 0.78 0.91 0.53 (0.24) 0.6875* 0.80 1.00 0.46 (0.26) 0.75 all fcc 0.74 0.52 0.80 fcc + hcp* 0.82 1.14 0.63 (0.36) 1 all fcc 0.71 0.78 fcc + hcp 0.84 1.26 0.44 PtO 2 0.82 1.64 PtO2(0001) 0.82 1.64 The more positively charged Pt atom screens the negatively charged oxygen atom and results in oxidelike bonding in these structures Similar arguments have been recently invoked to describe oxide chains on Pt(110)  oxide formation along the step edges on Pt(332)  and oxygenand sulfur induced restructuring of Au(111) surfaces [75-
51 77] Therefore, the evidence suggests that oxide compound formation can be induced on a range of metal surfac es through charge transfer and overcomes the penalty in buckling the metal or breaking metal metal bonds Based on the work on Pt(332), Wang and co workers suggested that the Pt oxide chains forming on step edges might be a precursor to bulk oxide formati on  but our results demonstrate that oxide chains can form on Pt(111) terraces as well, which implies that bulk oxide formation does not necessarily initiate only at step edges of Pt surfaces The reactivity of the 1D Pt oxide chains could be quite distinct from chemisorbed O atoms on Pt(111) and could impact the present understanding of CO and NO oxidation on Pt surfaces. Summary and Conclusions Using DFT we have identified new Pt oxide chain structures that develop on Pt(111) at oxygen coverages beyond 0.5 ML that are more stable than previously reported configurations of chemisorbed oxygen atoms [45,46] Formation of the Pt oxide chains involves oxygen atoms binding near hcp sites and forming close packed rows that run parallel to the p (21) oxygen rows The oxygen atoms in the added rows induce strong buckling of the Pt surface atoms which allows for screening of the repulsive O O interactions and stabilizes the structure. An analysis of the atomic charges shows that the Pt atoms in the chain structures are quite distinct from Pt atoms bonded to chemisorbed oxygen atoms and are more properly termed 1D Pt oxide chains The structures we have identified reproduce many of the features of recent STM images of O/Pt(111)  Based on our newly identified structures and the STM work, we can roughly describe the evolution of oxygen phases on Pt(111) up to an oxygen coverage of 0.75 ML At low coverages ( p (22) structure
52 With additional O atoms, a p (21) structure starts to form and co exists with the p (22) phase As the coverage starts to approach 0.5 ML, it becomes increasingly dominated by the p (21) phase On the p (21) domains, we begin to observe oxygen atoms aggr egating and growing Pt oxide chains that run parallel to the p (21) oxygen rows Prior STM images show that the chains form Y shaped structures that restrict the lengths of the chains (see Ref  for more details) Our DFT results on 1 ML coverage and earlier studies of oxide films on Pt(111) [32,54] suggest that the chains eventually merge and transition into a 2D PtO2(0001) film that is rotated by 30 with respec t to Pt(111) We note that similar oxide compound formation has been reported on Au(111) [75,76] P t(110)  and along steps on Pt(332)  These results suggest that the initiation of oxide formation on metal surfaces is the development of low dimensional metal oxide compounds that induce large metal atom buckling and are st abilized by large charge transfer Such oxide structures form at even moderate oxygen coverages and could play an important role in the catalytic behavior of these metal surfaces. While we have gained a better understanding of oxygen phase evolution on Pt (111), there are still several open questions that need to be further addressed. It is important to stress that the structures that we have identified on the 44Pt(111) surface do not fully represent the true experimental picture. Experimental STM images show that at coverages approaching 0.75 ML the chains form an interconnected network of Y shaped structures with each chain 19 to 24 long. Fully reproducing the longrange chain network observed in experiment is beyond the scope of DFT due to size lim itations We are currently extending our examination of the growth of the oxide
53 chains through the use of larger surface unit cells While such surface cells cannot incorporate Y structures, they should allow us to probe chainchain interactions and pref erred chain growth directions and possibly chain lengths A combination of such studies with comparisons between DFTderived and experimental STM should assist in better understanding the formation and growth of the oxide chains More studies, both experimental and modeling, are needed to fully explore the transition of Pt oxide chains to a full 2D PtO2 oxide film on Pt(111) The findings from this study already have important implications for proposed models of NO oxidation on Pt(111) [34,35,47] where NO oxidation on chemisorbed O surface phases was examined. Our work suggests that the NO oxidation on the Pt oxide chains should also be examined, since these structures form at relatively modest oxygen coverages Potentially, the presence of the Pt oxide chains plays an important role in the observed NO reactivity at coverages around 0.5 ML [34,35,47] The results on Pt(111) also suggest that the initial stages of the oxidation of Pd(111) should be revisited. DFT studies found that subsurface O becomes favorable on Pd(111) above 0.5 ML  but 2D oxide structures form before those surface coverages are reached [29,78,79] It will be interesting to determine if oxygen atom clustering plays a similar role in bulk oxide formation on Pd(111) as on Pt(111) Finally, the present study points to the difficulty of relying solely on DFT to predict surface structures in strongly reactive systems, such as oxygen on TM surfaces Studies of the oxidation of TM surfaces show that quite complex and difficult to predict configurations often form under even moderate oxygen coverages.
54 CHAPTER 3 STABILITY AND GROWTH OF PLATINUM OXIDE CHAINS ON PLATINUM (111) As an extension of Chapter 2, where we identified new Pt oxide chain structures that develop on Pt(111) at moderate oxygen coverages we wish to study these onedimensional oxide chains in more detail. As discussed in Chapter 2, the structures that have been identified with DFT do not fully reproduce all of the features that are observed experimentally with STM Specifically, at coverages approaching 0.75 ML, the DFT models predict infinitely lon g parallel Pt oxide chains, while experimental STM images show Pt oxide chains that form an interconnected network of Y shaped structures  Unfortunately, due to size limitations, our DFT calculations will not allow us to fully explore the formation and longrange interactions of the Y structures observed experimentally We can, however, examine in greater detail the growth and relative stability of the Pt oxide chains through the use of larger surface unit cells. Instead of using 2 2 or 4 4 Pt(111) surface unit cells by redefining the surface unit cell as a long rectangle parallel to the p (21) oxygen rows [2 Pt(111)] as illustrated in Figure 3 1 we can more easily explore the effects of chain length while minimizing the overall system size. For example, in a 4 4 Pt(111) surface unit cell, the longest chain that can be studied is 5.6 whereas in a 2Pt(111) surface unit cell, the longest chain that can be studied is 22.5 which is approximately the same length as one branch of the experimentally observ ed Y structure In addition, a five layer 2 Pt(111) slab contains ~200 atoms, which is currently the upper limit for computationally feasible DFT calculations Thinner slabs, which would allow for larger surface unit cells to be considered, were tes ted but were unable to accurately reproduce the Pt oxide chain structure.
55 Figure 3 1 An illustration of the original (left) and redefined (right) surface unit cells to more easily allow the study of the growth and stability of Pt oxide chains. Figure 32 Top and side views of Pt oxide chains at (a) 0.525 ML and (b) 0.725 ML. 5.6 22.5 4 4 Pt(111) 11.2 11.2 2 10 Pt(111) 9.7 28.1 (a) (b)
56 As the Pt oxide chain grows in length along the p (2 1) rows illustrated in Figure 32, its stability increases but with a decreasing rate, as shown in Figure 3 3 Positive values of added binding energy (calculated using Equation 22) indicate increased stability Even at lengths of ~17 (Figure 32b) the Pt oxide chains continue to gai n stability with the addition of oxygen atoms The chains expand laterally by about 0.35 as they grow in length, and the upwardly buckled Pt atoms (by 1.71.8 ) at the end of the chain expand outward while those in the center of the chain structurally resemble those of an infinitely long Pt oxide chain. Unfortunately, i n the 2Pt(111) surface unit cell, Pt oxide chains longer than ~17 will begin to interact with their periodic images along the p (21) rows so even with the redefined larger surfa ce unit cell, we cannot reproduce the experimentally observed chain lengths of 1924 Figure 3 3 Added binding energy versus Pt oxide chain length. Positive values of added binding energy indicate increased stability 0 2 4 6 8 10 12 14 16 18 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Added binding energy (eV)Pt oxide chain length ()
57 Alternate structures have been explored where oxygen atoms are added in locations other than at the end of an existing Pt oxide chain. When oxygen atoms are added between the same p (21) oxygen rows as the Pt oxide chain but with some separation, such as that shown in Figure 34a, the added binding energy is between 0.10 and 0.33 eV, compared to 0.90 (0.66) eV when the oxygen atom is added to the end of a 2.8 (14.2) chain. Similarly, when oxygen atoms are added in the empty space between adjacent p (21) oxygen rows, such as that shown in Figure 34b, the added binding energy is between 0.25 eV (where the added oxygen atom creates an infinitely long chain perpendicular to the p (2 1) row) and 0.37 eV In all cases, t he most stable structure is always that which grows the Pt oxide chain by adding an oxygen atom to the end of an existing chain. Figure 34 Examples of oxygen atoms added (a) between the same p (2 1) rows as the Pt oxide chain and (b) in the empty space between adjacent p (2 1) rows. Much larger surface unit cells would be required to allow us to fully examine the behavior of the Pt oxide chains, which is beyond the current capabilities of DFT calculations The development of an accurate charge transfer potential, such as the (a) (b)
58 chargeoptim i zed many body (COMB) potential  would allow us to study larger systems in an effort to better evaluate the features observed experimentally. In addition to comparing optimized atomic structures determined from DFT calculations with STM images obtained experimentally, we can create simulated STM images from DFT using the Tersoff Hamann approach  Briefly, the simulated STM image is a threedimensional contour plot of the partial ch arge density, where the tunneling current is proportional to the local density of states of the surface near the model probe tip. The resulting image allows for qualitative, but more direct, comparisons between DFT calculations and experimental results. Figure 3 5 ST M images of O on Pt(111) at 200 mV and 1 nA, (a) simulated, 9/16 ML O on Pt(111) and (b) experimental, 0.5625 ML O on Pt(111) In (a), th e light gray areas are O atoms, and the brightest white areas are upwardly displaced Pt atoms ; (b) is taken from Ref  As an example, if we examine the atomic scale protrusions on the p (21) O phase on Pt(111), which are indicative of the initiation of oxide formation, we observe similar features on both the simulated and experimental STM images, shown in Figures 35a and 35 b, respectively The Pt buckling predicted by DFT, expressed as bright white (a) (b)
59 areas in the simulated STM image, is qualitatively similar to that estimated from experimental STM images The ability to effectively simulate STM images with DFT can be a useful tool in more directly comparing theoretical and experimental results. We have begun to explore the reactivity of the Pt oxide chains to CO and NO oxidation and compare our results to existing studies Specifically, we are inter ested in examining the adsorption of CO NO and O2 on and near the Pt oxide chains and comparing to the p (21) O, p ( 2O (previously suggested as the most stable surface oxygen structure near 0.7 ML  ) PtO2(0001) surfaces, then determining appropriate pathways to CO2 and NO2 formation. Preliminary results indicate that CO NO and O2 molecules prefer to adsorb at the ends of the Pt oxide chains, as illustrated in Figure 36, where the upwardly buckled Pt atoms are most accessible to ads orbates, by about 0.6 eV/CO, 0. 03 eV/NO, and 0.5 eV/O2, respectively, compared to the p (21) O surface The adsorption energies of these configurations are 1.14 eV/CO, 1.45 eV/NO, and 0.28 eV/O2, respectively. Figure 36. Top and side views of (a) CO, (b) NO, and (c) O2 adsorbed at the ends of a Pt oxide chain at 0.650 ML O on 2Pt(111). (a) (b) (c)
60 Since the structure of the Pt oxide chains is much different than chemisorbed oxygen phases on Pt(111), we would expect that these structures exhibit di fferent reactivity, and as such, we are interested in whether or not these molecules can react when adsorbed at the end of the Pt oxide chains. Focusing on NO oxidation, the catalytic oxidation of NO to NO2 proceeds as NO(g) + 1/2 O2(g) 2(g). Specifically, the mechanism which explains the experimental observations is NO(ad) + O(ad) 2(g) + 2 (where is an adsorption site). P revious DFT calculations indicate that conversion of NO(ad) to NO2(ad) is energetically feasible on a highly O cove red Pt(111) surface [34,47] but questions remain about the character of the surface, its interactions with NO and O2, and the function of reaction intermediates. A combined experimental and DFT study by Smeltz et al. explored NO oxidation on both the clean Pt(111) surface and a p ( 2O (2/3 ML O) co vered Pt(111) surface  They determined that independent adsorption of NO and O2 are less favorable on the O covered surface than clean Pt(111) and that O2 adsorption and dissociation are endothermic on the O covered surface. However, a peroxynitrite (OONO*) interme diate which actually enhances NO and O2 binding is moderately stable. Such an intermediate may provide an alternate pathway for NO2 formation via NO assisted O2 dissociation. W e have explored a few possibilities for forming NO2 from an NO molecule adsorbed at the end of a Pt oxide chain. Considering only chemisorbed O atoms, two possible mechanisms exist: the adsorbed NO combines with an O atom either from the adjacent p (2 1) row or from the end of the Pt oxide chain. The energetics for NO (g) + O (ad) N O2(g) indicate that the reaction with the O atom from the end of the oxide
61 chain is nearly energetically neutral ( E = +0.01 eV), but the reaction with the O atom from the p (2 1) row is mildly exothermic ( 0.18 eV) Another possibility for the formation of NO2 exists: the reaction of adsorbed NO with coadsorbed O2, as shown in Figure 37. Figure 37. Top and side views of a possible mechanism for NO oxidation via coadsorbed O2 at the end of a Pt oxide chain at 0.650 ML O on 2 Pt(111). The energetics for NO(g) + O2(g) NO2(g) + O(ad) indicate that NO can react exothermically ( E = 0.72 eV) with a coadsorbed O2 molecule at the end of an oxide chain to form NO2. The reaction kinetics are yet to be determined, but the overall energetics are nearly equal to previous studies of NO oxidation on Pt(111)  These results are indicat ive that the Pt oxide chains can indeed be reactive though much f uture work will be necessary to fully elucidate CO and NO oxidation mechanisms on the Pt oxide chains.
62 CHAPTER 4 STRUCTURE OF THE PALLADIUM OXIDE (101 ) THIN FILM ON PALLADI UM (111) Bulk crystalline PdO has a tetragonal unit cell and consists of square planar units of Pd atoms fourfold coordinated with oxygen atoms The bulk terminated PdO(101) surface is defined by a rectangular unit cell of dimensions a = 3.04 and b = 6.14 where the a and b lattice vectors coincide with the  and [ 1 01] directions of the PdO crystal, respectively Figure 41. (a) Top an d (b) side view of the PdO(101) thin film structure. The red and gray atoms represent O and Pd atoms respectively Rows of threefoldcoordinated (cus) Pd and O atoms are indicated. The a and b directions correspond to the  and [ 1 01] crystallograph ic directions of PdO. The PdO(101) PdO surface consists of alternating rows of threefold or four fold coordinated Pd or O atoms that run parallel to the a direction shown in Figure 4 1 Thus, half of the surface O and Pd atoms are coordinatively unsaturated (cus) and likely to be more active than the four fold coordinated atoms for binding adsorbed molecules 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
63 Pd atoms and 0.7 ML of O atoms Given that the film contains ~3.0 ML of oxygen atoms, we estimate that the PdO(101) film on Pd(111) consists of between four and five layers and has a total thick ness of ~12 In a prior study, Kan et al.  found that the PdO(101) structure aligns with the 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 = 3.057 and b = 6.352 Unfortunately, the geometry of the PdO(101) thin film on Pd(111) is not conducive to some simple analyses, such as of the density of states (DOS). Briefly, the density of states is a description of the number of electronic states at each energy level, often shown relative to the Fermi level, that are available to be occupied. The DOS can be calculated using wave functions in DFT. In fact, DFT is also able to provide both siteand angular momentum projected DOS, which allows us to examine the DOS projected onto the molec ular orbitals of each individual atom in the system. Thus, t he projected density of states (pDOS) is useful in analyzing, for example, chemical bonding of an adsorbate to a substrate. For the PdO(101) surface, aligning the O cus Pd O with a primary Ca rtesian axis would assist in simplifying the analysis and comparison of pDOS among different systems, such as the clean Pd(111) metal surface By aligning the system in this manner, the projected molecular orbitals are correctly defined relative to the PdO(101) surface and will match those of the clean metal surface ; that is, for example, the lobes of the dz 2 orbital are aligned along the zaxis, which is also normal to the O cus Pd O of the PdO(101) surface and to the surface Pd of the Pd(111) surface This is
64 important for later analysis, as we expect adsorbates to interact with the surface approximately along the surface normal direction and with the dz 2 molecular orbital. A simple rotation of the unit cell about the yaxis ( direction), as il lustrated in Figure 4 2, produces the desired orientation. Figure 42. Illustration of the rotation of the (a) original PdO(101) unit cell to (b) align the O cus Pd O along the xaxis ([ 1 01 ] direction). Using the Cartesian coordinates of the original PdO(101) unit cell (Figure 42a) and simple geometric calculations, we determined that a 17.4 clockwise rotation about the yaxis would provide the correct alignment of the O cus Pd O, as illustrated in Figure 42b. To implement this in the DFT calculations, we rotated the a and c lattice vectors in the xzplane about the yaxis clockwise by 17.4, ensuring that the lengths of the lattice vectors remained constant. To demonstrate the effect of this rotation, we can examine the orbital decompo sed pDOS for a four fold coordinated Pd atom on the PdO(101) surface both before and after rotation, as illustrated in Figure 43. The focus is on the two peaks centered at approximately 6 eV (filled electronic states) and +2 eV (unfilled electronic states ); these peaks are indicators of interactions with adjacent O atoms. Before the unit cell is rotated
65 (Figure 43a), the Pd dxy and dyz orbitals contribute unevenly to each of these two peaks, which agrees with the geometry of the system where the four fold coordinated Pd atom and adjacent O atoms lie approximately in the xyplane but rotated slightly upward in the zdirection about the yaxis. After rotation (Figure 43b), the Pd dxy and dyz orbitals contribute equally to each of the two peaks. This agrees with the geometry of the rotated system since the four fold coordinated Pd atom and adjacent O atoms lie approximately at a 45 angle between the xyand yzplanes, which implies that the dxy and dyz orbitals of the Pd atom would be equally involved in bonding to adjacent O atoms. Differences in bonding characteristics of threefold and four fold coordinated atoms on the PdO(101) surface can also be elucidated by examining the orbital decomposed pDOS. For example, in comparing the threefold(Figure 44a) and fourfold(Figure 43b) coordinated Pd atoms, we notice that the large dxy and dyz peaks at about 6 and +2 eV have diminished in the four fold coordinated Pd, leaving only those near the Fermi leve l, while dz 2 and dx 2 y 2 peaks appear at about 6 eV and just above the Fermi level in the threefold coordinated Pd. Again, these features agree with the geometry of the system, as the O atoms adjacent to the threefold coordin ated Pd lie to each side of th e Pd atom along the xaxis and below the Pd atom along the zaxis. The unfilled states just above the Fermi level indicate that bonding of an adsorbate is likely to occur between the adsorbate and the Pd dz 2 and/or dx 2 y 2 orbitals.
66 Figure 43. Projected density of states for a four fold coordinated Pd atom on the PdO(101) surface (a) before and (b) after a clockwise rotation of the unit cell. -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.2 0.4 0.6 0.8 1.0 pDOS (states/eV)Energy (eV) Pd dxy Pd dyz Pd dz2 Pd dxz Pd dx2-y2 -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.2 0.4 0.6 0.8 1.0 pDOS (states/eV)Energy (eV) Pd dxy Pd dyz Pd dz2 Pd dxz Pd dx2-y2 (a) (b)
67 Figure 44. Projected density of states for (a) threefold coordinated Pd, (b) three fold coordinated O, and (c) four fold coordinated O atoms on a PdO(101) surface. -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.2 0.4 0.6 0.8 1.0 pDOS (states/eV)Energy (eV) Pd dxy Pd dyz Pd dz2 Pd dxz Pd dx2-y2 -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.2 0.4 0.6 0.8 1.0 pDOS (states/eV)Energy (eV) O px O py O pz (a) (b)
68 Figure 44. Continued -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.2 0.4 0.6 0.8 1.0 pDOS (states/eV)Energy (eV) O px O py O pz (c)
69 CHAPTER 5 FIRST PRINCIPLES STUDIES O F THE ADSORPTION OF HYDROGEN ON THE OXIDES OF PALLADIUM (111) Introduction Understanding the chemistry of transition metal oxides has garnered increasing interest recently because of their possible applications as industrial catalysts In particular, studying hydrogen on transition metal oxides is of interest due to its importance in certain reactions where hydrogen adsorption and diffusion are relevant processes, such as intermediates in the synthesis or dehydrogenation of alkanes, though its presence can lead to potential contaminants as the surface becomes h ydroxylated and water is formed  The presence of stable molecularly adsorbed hydrogen suggests a dative bonding mechanism that may be important in alkane adsorption and activation processes Thus, gaining a fundamental understanding of when and why molecularly adsor bed hydrogen can be found on transition metal oxides versus clean metal surfaces would be useful in further understanding such reactions. The oxidation of Pd(111) proceeds through the formation of a series of stable and metastable structures  As o xygen is added, that is, the chemical potential of O2 increases, a chemisorbed O phase forms, followed by a twodimensional surface oxide, and finally by bulk oxide. The structure of the Pd5O4 two dimensional surface oxide on Pd(111) was resolved by Lundgren et al.  and the formation of a bulk like PdO(101) thin film on Pd(111) was recently discovered by Kan and Weaver  The PdO(101) thin film possesses unique properties and has been shown to be reactive to a variety of molecules, such as the adsorption of O2  H2O  and small alkanes  as well as in facile C H bond cleavage of propane  Experiments also indicate the presence of chemisorbed molecular hydrogen at coordinatively unsaturated (cus), or
70 undercoordinated, sites on PdO(101) as well as RuO2(110)  which is particularly interesting since chemisorbed H2 is relatively uncommon  In this chapter we focus on elucidating the interaction of molecular hydrogen, particularly the nature of the moleculesurface bonding, with two thin film oxides that form on Pd(111) as well as the clean m etal surface using density functional theory (DFT) calculations We observe chemisorbed molecular hydrogen on low index, low coordinated Pd surfaces (Pd(111) and Pd(110)), as do Schmidt et al. on Pd(210)  As a result, we wish to determine the underlying reason for this strong moleculesurface interaction. Perhaps the cus Pd atoms on the PdO(101) surface are in some way similar to the undercoordinated atoms at the Pd metal surface In addition, Sun et al. have reported chemisorbed molecular hydrogen on RuO2(110)  with qualitative si milarities in the moleculesurface bonding characteristics, specifically, a donation/backdonation mechanism, to our current results on PdO(101), and we explore whether the nature of the bonding is indeed similar Finally, we investigate why the interaction of H2 with the two dimensional surface oxide is significantly different than the PdO(101) thin film. Blanco Rey et al. also recently studied molecular hydrogen on Pd oxides  but there are a few key differences between the present study and what they have reported. While both studies examine PdO(101), BlancoRey et al. based their structure solely on the bulk PdO structure, but we have strained our thin film to match the structure reported experimentally  The structu re is illustrated in Figure 4 1 Since the PdO(101) lattice expansion required to achieve commensurability with the Pd(111) substrate is relatively small (0.46% and 3.4% in the a and b directions, respectively), we
71 do not anticipate our results to differ greatly from those already reported. On the other hand, Blanco Rey et al. have studied the Pd(100) ( O ( while we have studied Pd5O4/Pd(111) Both surface oxides possess differently coordinated Pd atoms in addition to O atom s The Pd5O4 in that smaller unit cells are required to achieve commensurability with the substrate, thereby reducing computational expense, and that its O 2Pd O trilayer geometry is more analogous to PdO(101) On the other hand, Pd5O4 is a two dimensional oxide that bears no resemblance to a bulk oxide yet forms at a close packed metal surface Thus, we expect that our results will complement those of Blanco Rey et al. in explaining the interaction of molecular hydrogen w ith Pd oxide surfaces. Calculation D etails All of the DFT calculations in this chapter are performed using VASP (Vienna ab initio simulation package)  We use the projector augmented wave (PAW) pseudopotentials [59,60] provided in the VASP database Calculations have been done using the generalized gradient approximation Perdew Burke Ernzerhof (GGA PBE) exchangecorrelation functional  A planewave expansion with a cutoff of 400 eV is used, and the total energy calculations are done using a residual minimization method with direct inversion in the iterative subspace (RMM DIIS) for electronic relaxations, accelerated using Methfessel Paxton Fermi level smearing with a Gaussian width of 0.1 eV  The positions of the atoms are relaxed using a limited memory BroydenFletcher GoldfarbShanno (LBFGS) optimization method  until the forces on all u nconstrained atoms are less than 0.03 eV/.
72 The model for the PdO(101) surface described below matches our earlier studies of H2O and alkanes on the PdO(101) thin film [86,87] While experimentally the PdO(101) film is grown on the Pd(111) surface, the oxide film is sufficiently thick (13 ) that we assume the Pd(111) substrate may be ignored 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 Instead, we represent the PdO(101) film with four layers, which corresponds to a thickness of approximately 9 The bottom layer is fi xed, but, unless noted otherwise, all other atoms in the oxide film and the H2 adsorbate are allowed to relax We use a vacuum spacing of 20 which is sufficient to minimize any spurious periodic interactions in the surface normal direction. The PdO(101) film is obtained from the relaxed PdO bulk structure, but the film is strained ( a = 3.057 b = 6.352 ) to match the reported experimental structure  The a and b lattice vectors correspond to the  and [ 1 01 ] directions of crystalline PdO, respectively A 441 (421) Monkhorst Pack k point mesh  was used for the 12 (14) surface unit cell We have confirmed that higher energy cutoffs and finer k point meshes have a negligible effect on the results reported. We use the climbing nudged elastic band (NEB) method  to find minimum energy pathways (MEP) and identify transition states (TS) for H2 dissociation. A recent DFT study of H2 adsorption and water formation on Pd oxides utilized a strainfree PdO(101) surface obtained from the relaxed, strain free PdO bulk structure ( a = 3.080 b = 6.212 )  Though these small differences in dimensions should not affect H2 adsorption or dissociation, there are some key differences between the current study and that of Blanco Rey et al. which will be discussed in more detail below The vibrational modes
73 are obtained by fixing the PdO(101) substrate and calculating frequencies for only the six degrees of freedom for the H2 molecule. For selected H2 binding sites, we performed calculations with the top PdO layer free and observe differences of only 5 to 3 0 cm1 and 0.6 kJ/mol in the vibrational frequencies of adsorbed H2 and the zero point corrections, respectively. These differences are not sufficient to modify any of the conclusions reported in this chapter. Calculations on the Pd(111) and Pd(110) surfaces are done using five layer slabs; the bottom two layers are fixed, but the top three layers and the H2 adsorbate are allowed to relax The 44Pd(111) (32Pd(110)) slab is approximately 9 (6 ) thick with a vacuum spacing of approximately 15 (19 ), which is sufficient to minimize any spurious periodic interactions in the surface normal direction. A 221 (441) Monkhorst Pack k point mesh  was used for the 44Pd(111) (32Pd(110)) surface unit cell The lattice parameter for Pd is found to be 3.953 using DFT versus an experimental value of 3.89 and the calculations were performed while fixing the inplane lattice parameter to 3.953 We constructed the twodimensional Pd5O4 surface oxide on a threelayer Pd(111) slab based on the details provided in Ref  where seven Pd5O4 unit cells are required to achieve commensurability with the Pd(111) substrate. The supercell was created such that the inplane lattice parameter for Pd was fixed to 3.953 After optimizing the Pd5O4/Pd(111) structure, the atomic positions for all atoms except the H2 adsorbate were fixed to minimize the computational expense for this large supercell Selected configurations were tested with a relaxed Pd5O4 layer, but the small differences in geometry and adsorption energy (less than 1 kJ/ mol) do not affect the
74 conclusions drawn in this chapter. Due to the size, all the calculations for the Pd5O4 Calculations were also performed on onedimensional Pt oxide chains on 2 Pt(111), similar to those re ported in Ref  and discussed in Chapter 3. The Pt(111) substrate is composed of fi ve layers, where the bottom two layers are held fixed while the remaining atoms, including the adsorbed oxygen atoms and H2 molecule, are permitted to relax The 2Pt(111) slab is approximately 9 thick with a vacuum spacing of approximately 15 w hich is sufficient to minimize any spurious periodic interactions in the surface normal direction. A total of 0.65 monolayers (ML) of oxygen atoms were adsorbed on the Pt(111) surface: 0.50 ML in a p (21) pattern and the remaining 0.15 ML arranged between the p (21) rows to form Pt oxide chains of approximately 17 in length. A 211 Monkhorst Pack k point mesh  was used for the 2Pt(111) surface unit cell The lattice parameter for Pt is found to be 3.977 using DFT versus an experimental value of 3.92  and the calculations were performed while fixing the inplane lattice parameter to 3.977 Results and D iscussion H2 on PdO(101) Molecular chemisorption of H2 We have examined H2 adsorption on the stoichiometric PdO(101) surface, which consists of alternating rows of threefoldand fourfoldcoordinated Pd or O atoms, where the threefoldcoordinated atoms are coordinatively unsaturated  As a result, half of the surface Pd and O atoms are coordinatively unsaturated (cus) and are likely to be more active than the fourfoldcoordinated atoms for binding adsorbed molecules We considered several different possible adsorption sites for H2 on the PdO(101) surface,
75 including both atop and between each of the four coordinatively distinct atom types on the surface and aligned parallel to both the a and b lattice vectors (parallel and perpendicular to the rows of surface atoms) Only two strongly bound configurations of adsorbed H2 were identified and are illustrated in Figure 5 1 : atop cus Pd both perpendicular (Figure 5 1 a) and parallel (Figure 5 1 b) to the row of cus Pd atoms The adsorption energy for H2 adsorbed parallel (perpendicular) to the cus Pd row is 46.7 (38.9) kJ/mol on the 14 surface unit cell, compared to 45.3 (37.3) kJ/mol on the 12 surface unit cell T hese values for the adsorption energies clearly demonstrate that H2 is chemisorbed on the PdO(101) surface H2 is only weakly bound on the remaining adsorption sites, as adsorption energies range from near zero to ~3 kJ/mol The H H bond distance is elongated to 0.84 (0.82) relative to gas phase H2 at 0.75 and the distance from the cus Pd atom to the center of mass of the H2 molecule is 1.75 (1.79) Figure 5 1 Top and side views of two stable configurations of molecular hydrogen adsorbed on PdO(101): (a) atop cus Pd, perpendicular to the cus Pd row and (b) atop cus Pd, parallel to the cus Pd row The adsorption energy for an H2 parallel (perpendicular) to cus Pd rows is 46.7 (38.9) kJ/mol.
76 This agrees well with that reported by BlancoRey et al. where the adsorption energy for H2 adsorbed parallel (perpendicular) to the cus Pd row is ~54 (~46) kJ/mol, and the H H bond distance is elongated to 0.84 (0.81)  This adsorbate configuration is also remarkably similar to H2 on RuO2(110), where the H2 is bound atop a cus surface metal atom, the H H bond is elongated to 0.81  and adsorption energies of ~31 kJ/mol  and ~35 kJ/mol  have been previously reported. Thus, H2 is more strongly bound on PdO(101) than on RuO2(110) For H2, we can expect zero point corrections (ZPC) to have a significant impact on the adsorption energy, but more importantly, the ZPC might be the source of the strong kinetic isotope effect observed experimentally  Table 5 1 shows vibrational frequencies calculated for both H2 adsorption configurations and the stretch frequency computed for gas phase H2. The most critical observation is that a single imaginary frequency is associated with the H2 perpendicular configuration; therefore, the perpendicular configuration represents a transition state. The H2 molecule in the perpendicular configuration experiences no forces in the a direction due to the symmetry of the surface When perturbed with a rotation around the surface normal, the configuration minimizes to H2 in the parallel orientation. Therefore, H2 in the per pendicular configuration is a transition state for rotation of the H2 molecule about the surface normal and is not an adsorption minimum This result has important implications for the H2 dissociation pathways, since the earlier DFT study by Blanco Rey et al.  focused on the H2 perpendicular configuration to map o ut an H2 dissociation pathway As a result, they report H2 dissociating from the perpendicular configuration to form an OxH group and a cus Pd H adjacent in the b direction (i.e. perpendicular to the
77 cus Pd row) The OxH group is ultimately involved i n the formation of water This dissociation pathway cannot be valid because the initial state is not a stable minimum As noted in the Calculation Details section, there are differences in the surface structure of our PdO(101) film and the PdO(101) surf ace used in the earlier DFT study In particular, our film is strained to the experimental values of the PdO(101) thin film grown on Pd(111) while Blanco Rey et al. examined a PdO(101) surface obtained from the DFTrelaxed PdO bulk structure. To test if this strain might have an effect on the nature of the H2 configuration, we examined the two H2 adsorption configur ations shown in Figure 5 1 on a PdO(101) surface generated from the DFTrelaxed bulk structure. The adsorption energies on the unstrained PdO(101) surface for the perpendicular and parallel orient ations are 39.7 and 47.4 kJ/mol, respectively These values for the adsorption energies are practically indistinguishable (less than 1 kJ/mol) from our strained PdO(101) film The associated vibra tional frequencies for the two configurations on the unstrained PdO(101) film are also reported in Table 5 1, and again, H2 atop a cus Pd atom and aligned perpendicular to the cus Pd row is not a stable minimum The identification of the imaginary mode f or the H2 configuration in Figure 51a leaves only one strongly bound, stable H2 adsorption site on the PdO(101) surface. Using the frequencies for this configuration, we compute ZPC adsorption energies of 32.5 (36.6) kJ/mol for H2 (D2). Therefore, the Z PC lowers the adsorption energy by about 610 kJ/mol, bringing the DFT values closer the TPD derived adsorption energies of 29.1 (33.5) 2.0 kJ/mol for H2 (D2). Furthermore, the DFT results reproduce the order of the H2 versus D2 desorption peak temperat ures observed in the TPD spectra.
78 Table 5 1 The vibrational frequencies (cm1) of H2 adsorbed atop cus Pd perpendicular and parallel to the cus Pd rows (see Figure 5 1 for the configurations) The stretch frequency of gas phase H2 is 4299 cm1. F requencies for H2 adsorbed in the same configuration but using a bulk PdO(101) surface are shown in parentheses Note the imaginary frequency for the perpendicular configuration indicating that this configuration is a transition state. Mode Atop cus Pd (perpendicular) Atop cus Pd (parallel) Stretch 3444 (3438 ) 3115 (3112 ) Rotation 1341 (1355 ) 1601 ( 1510 ) Translation 834 (849 ) 903 ( 912 ) Translation 410 (411 ) 477 ( 48 0) Translation 322 (319) 402 ( 399) Rotation (in plane) 281i (293i) 269 ( 272 ) The charge density difference, a visual representation of the areas of charge accumulation or depletion upon creation of an adsorbatesubstrate bond, for H2 adsorption on PdO(101) is shown in Figure 5 2 Differences of less than 0.01 electrons are not depicted in the figures for clarity Figure 5 2 Charge density difference plot for H2/PdO(101) with H2 parallel to cus Pd rows Note areas of charge accumulation between the surface and adsorbed H2, as well as areas of charge depletion both above the adsorbed H2 and around the surface Pd atom.
79 The charge density difference is calculated by subtracting the charge densities of both the isolated adsorbate molecule and bare substrate from the charge density of the adsorbed system while keeping all atomic positions fixed. Charge accumulation is evident between the surface and adsorbed H2, while charge depletion occurs both above the adsorbed H2 and around the cus Pd surface atom This combination is indicative of a donation/backdonation bonding mechanism, which is also observed in H2/RuO2(110)  ; these adsorbed states may be classified as complexes. Additionally, analysis of the Bader charges [70 72] indicates that the cus Pd atom loses 0.06 electrons, while the remaining surface atoms do not exhibit any charge transfer. The projected density of states ( pDOS ), shown in Figure 5 3a, further confirms the strong interaction between adsorbed H2 and the cus Pd surface atom The peak residing below the d band at about 8 eV, where narrow Pd d corresponds to the bonding interaction; the overlapping features between +3 and +5 eV are indicative of the antibonding interaction. Further analysis was performed on the orbital decomposed pDOS ( Figure 53b) which clarifies the specific molecular orbitals involved in the bonding and antibonding interactions As expected due to the geometry of the system, the Pd dz 2 Pd dyz contribute to the antibonding interaction. In addition, we calculated the d band center of the cus Pd surface atom to which H2 is bonded. Upon H2 adsorption, the d band center shifts 1.26 eV lower in energy to 2.80 eV. As we consider the nature of the b onding between adsorbed H2 and PdO(101), we speculate that the presence of undercoordinated Pd atoms on the PdO(101) surface serve s to enhance the bonding interactions and leads to molecularly adsorbed H2.
80 Figure 5 3 Projected density of states of Pd d states and H states for H2 adsorbed on PdO(101). The Fermi level is referenced to zero energy -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.5 1.0 1.5 2.0 pDOS (states/eV)Energy (eV) Pd d H -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.5 1.0 1.5 pDOS (states/eV)Energy (eV) Pd dxy Pd dyz Pd dz2 Pd dxz Pd dx2-y2 H (a) (b)
81 Naturally, we must also consider the adsorption of H2 on bare Pd surfaces to further determine the role of metal underc oordination to the stability of H2 on these surfaces There is much debate in the literature about the dissociative adsorption versus the presence of a molecular precursor on bare metals [90,91,96101] but we will restrict our focus to comparing the molecular H2 states on bare Pd surfaces, namely Pd(111) and Pd(110), and the PdO(101) surf ace. Pathways for dissociative chemisorption of H2 We now turn to possible pathways to dissociation for H2 adsorbed parallel to the cus Pd rows While not exhaustive, we attempted to identify pathways for direct dissociative adsorption by bringing the H2 molecule at various orientations towards the cus Pd and cus O sites In all attempts, the H2 molecule is repelled, leading to weak binding, or relaxes to the minimum shown in Figure 5 1 Thus, these calculations do not reveal a facile pathway for the dir ect dissociation of H2 on PdO(101) and instead point toward an indirect pathway with the adsorbed H2 complex acting as a precursor to dissociation, and subsequently water formation. We leave the DFT study of the pathways to water formation for the future and instead concentrate on identifying H2 dissociation pathways that can support the observed large kinetic isotope effect Before discussing the H2 dissociation mechanisms we have identified using DFT, it is worthwhile to summarize features that we should expect from such a pathway(s) if it is to explain the isotope effect Firstly, the ZPC barrier should be similar in magnitude to the ZPC adsorption energies for H2 and D2. If the dissociation barrier is too small then b oth H2 and D2 would react to form water and the observed D2 molecular peak observed in TPD would not be present If, on the other hand, the dissociation barrier is similar to the desorption energies, then ZPC could potentially result in the dissociation
82 barrier being larger than the desorption energy for D2 ( and the inverse for H2), leading to the strong kinetic isotope effect Secondly, the pathway(s) for dissociation should result in a state that will be favorable for H2O formation instead of simply reassociation to H2. To obtain possible end states for the dissociated H2 molecule, we mapped out the adsorption sites for isolated H atoms on the PdO(101) surface The three stable sites that we found are shown in Figure 5 4 Figure 5 4 The top and s ide view of H adsorption on (a) top cus O (Eads = 142.8 kJ/mol) (b) top cus Pd (Eads = 52.0 kJ/mol) and (c) between cus Pd (Eads = 30.2 kJ/mol ) sites on the PdO(101) surface. The H atom can adsorb atop a cus O site, atop a cus Pd site, and between cus Pd atoms with adsorption energies of 142.8, 52.0, and 30.2 kJ/mol of H2, respectively. Clearly, the adsorption onto cus O to form an OxH group is strongly favored, and we can expect that such a configuration would eventually lead to water formation. The b arriers to H hopping from cus Pd to cus O, cus O to cus O, and cus Pd to between cus Pd sites are all very large (> 100 kJ/mol). These large barriers to H atoms moving on the PdO(101) surface narrows the possible pathways for H2 to dissociate from the atop cus Pd site and produce an OxH group.
83 Based on the parallel orientation of the H2 molecule on the PdO(101) surface, we initially examined dissociation pathways that would lead to (1) both H atoms adsorbed between cus Pd sites (see Figure 5 5 a ) and (2) one H atom atop a cus Pd site and the other on an adjacent between cus Pd site (see Figure 5 5 b ) The dissociation (association) barriers for mechanisms (1) and (2) are 46.3 (30.9) kJ/mol and 19.7 (11.8) kJ/mol, respectively For both of these mechanisms, the molecular state is favored since it is easier to associate than dissociate, and the molecular state is more energetically favorable than the dissociated state. Figure 55. The three identified final states for H2 dissociation from H2 adsorbed in a parallel ori e n tation on the cus Pd rows (see Figure 51b ) H2 dissocia tes into (a) the two between cus Pd sites (Eads = 31.4 kJ/mol), (b) an atop cus Pd and between cus Pd site (Eads = 38.7 kJ/mol), and (c) an atop cus Pd and adjacent atop cus O site (Eads = 124.0 kJ/mol). The energetics of the H atom recombination pathways are clearly too low to account for the H2 desorption product that we observe near 300 K in TPD experiments, which supports our conclusion that the recombination product is associated with defect sites Also, the only way either dissociation mechanism could produce water at low
84 temperature is if pathways for H atom hopping to cus O sites have very low energy barriers In this case, the H2 molecule could dissociate on the cus Pd row and then quickly move to the cus O site to form an OxH group, thus avoiding facile recombination back to the molecular state. However, we find that the barriers for H atom s to hop between the cus Pd and cus O rows are very large (> 100 kJ/mol) for all of t he configurations explored here. Thus, while the two mechanisms have barriers that are feasible for H2 dissociation, they clearly do not lead to a pathway towards a stable dissociated product that can form water Probing stable configurations for two H atoms in close proximity leads to a configuration with one H atom atop a cus Pd site and the other bonded to t he adjacent cus O atom (see Figure 5 5 c), which has an adsorption energy of 124 kJ/mol of H2. Such a configuration is the end state for the H2 di ssociation reported by BlancoRey et al. but as noted above, their initial structure was H2 lying perpendicular to cus Pd rows Using four images in the climbing NEB calculation we linked the H2 molecule aligned parallel to the cus Pd row to the dissociated H2 configuration shown in Figure 5 5 c. The detailed pathway is illustrated in Figure 5 6 The H2 molecule begins to rotate, but instead of reaching the more symmetrical TS for H2 rotation, the H H bond elongates from an initial va lue of 0.84 to 0.99 at the TS and the MEP leads to the final dissociated H2 product of Figure 55c. The dissociation (association) barrier for this pathway is 37.0 (114.4) kJ/mol. These values and mechanism satisfy the two criteria for a suitable H2 dissociation pathway that will lead to H2O production namely that the dissociation barrier is comparable to the desorption barrier and that the end product is stable and will not readily reassociate back to molecular H2.
85 Figure 5 6 The pathway for H2 dissociation into H atoms bonded to atop cus Pd and atop cus O The transition state and other images from the climbing NEB calculation are shown in the inset The barrier (without zero point corrections) to dissociation (association) along this pathway is 37.0 ( 114.4) kJ/mol. Using the vibrational frequencies computed for the initial states and transition states, we find that the ZPC dissociation barriers for H2 and D2 in this pathway are 31.8 and 33.4 kJ/mol, respectively. Recollect that the ZPC deso rption barriers (i.e. the adsorption energies) for molecular H2 and D2 are 32.5 and 36.6 kJ/mol Our DFT calculations thus predict effec tive activation energies of 0.7 and 3.3 kJ/mol for the dissociative chemisorption of H2 and D2, respectively, on PdO(101) These effective dissociation barriers cannot explain the experimentally observed behavior Firstly, the facile dissociation of H2 that we observe experimentally suggests that the effective
86 dissociation barrier for H2 is negative by a few kJ/mol For H2, the apparent dissociation barrier predicted by DFT is negative but nearly zero; thus, it fails to reproduce the large dissociation yields (> 90% ) observed experimentally Furthermore, the experimentally observed kinetic isotope effect suggests that the effective dissociation barrier should be lower for H2 than D2, but the DFTpredicted barriers exhibit the opposite trend. Since our data provides strong support for the precursor mediat ed mechanism, as shown in Figure 5 6 we do not question the validity of this mechanism but rat her conclude that the strong kinetic isotope effect arises from factors other than zeropoint energy differences. Our calculations suggest that quantum mechanical tunneling is the dominant pathway for H2 and D2 dissociation on PdO(101) at the low temperatures at which these molecules dissociate during TPD (~ 100 K) For more details, see Ref  Overall, the tunneling calculations provide a consistent description of the kinetic isotope effect observed in the low temperature dissociation of H2 and D2 on PdO(101) We thus conclude that tunneling dominates the dissociation of H2 on PdO(101) at low temperature and cause s a large kinetic isotope effect for this reaction. H2 on Pd(111) and Pd(110) We examine H2 adsorption states on three bare Pd surfaces: Pd(111), Pd(110), and Pd(111) with a single Pd adatom in an fcc site These surfaces provide surface Pd atoms with varying coordination. A surface Pd atom is ninefoldcoordinated on Pd(111) and sevenfoldcoordinated on Pd(110) Similarly, a single Pd adatom on Pd(111) is threefoldcoordinated, which solely in t erms of coordination number, is the same as a cus Pd atom on PdO(101) In general, H2 prefers to adsorb atop a Pd atom on bare metal surfaces
87 Figure 5 7 (a,b) Top views of two equivalent stable configurations of H2 adsorbed atop a surface Pd on Pd( 111); (c,d) Top and side views of two stable configurations of H2 adsorbed atop a single Pd atom on Pd(111) The adsorption energy for H2 atop a surface Pd on Pd(111) [(a,b)] is 23.2 kJ/mol and invariant with respect to orientation but increases to 47.7 k J/mol when H2 is adsorbed atop a single Pd atom on Pd(111) [(d)]. On Pd(111), all surface Pd atoms are equivalent, and due to the symmetry of the system, there are only two unique configurations where H2 is lying flat on the surface (Figures 5 7 a and 5 7 b) Both configurations are equally stable; in fact, the adsorption energy of H2 on Pd(111) is 23.2 kJ/mol and invariant with respect to orientation. This agrees well with a previous study in which an adsorption energy of ~21 kJ/mol was reported  In addition, there is no discernable barrier to H2 rotation on the Pd(111) surface when H2 is adsorbed atop a surface Pd atom The H H bond distance is elongated to 0.86 and the distance from the surface Pd atom to the center of mass of the H2 molecule is 1.69 We have not explored dissociation pat hways since our focus is on the bonding interaction between molecular H2 and the metal surface, but Dong and Hafner found that H2 dissociates readily on Pd(111) by a variety of nonactivated and slightly activated pathways  Thus, the molecular states that have been identified are precursor states to the dissociative adsor ption of H2.
88 We also obtained two stable configurations where H2 is adsorbed atop a single Pd adatom on Pd(111) (Figures 5 7 c and 5 7 d) In both cases, the H2 does not reside directly atop the Pd adatom but is askew and moved closer to the Pd(111) surface The most stable configuration is such that each H atom is aligned above a surface Pd atom (Figure 5 7 d); its adsorption energy increases (relative to H2 on Pd(111)) to 47.7 kJ/mol The H H bond distance is elongated to 0.84 and the distance from the Pd adatom to the center of mass of the H2 molecule is 1.72 Figure 5 8 Top and side views of two stable configurations of H2 adsorbed on Pd(110) The adsorption energy for an H2 parallel (perpendicular) to cus Pd rows is 36.1 (33.8) kJ/mol. The Pd( 110) surface is corrugated, and the ridges are composed of cus Pd atoms It is wellknown that atomic hydrogen can induce reconstructions of the Pd(110) surface  but we are only concerned about molecularly adsorbed hydrogen. H2 again prefers to adsorb atop Pd atoms, either parallel or perpendicular to the rows of cus Pd atoms (Figures 5 8 a and 5 8 b) The adsorption energy for an H2 aligned parallel (perpendicular) to the cus Pd rows is 36.1 (33.8) kJ/mol, and there is no discernable barrier to H2 rotation on the Pd(110) surface The H H bond distance is elongated to
89 0.85 and the distance from the Pd adatom to the center of mass of the H2 molecule is 1.72 As mentioned earlier, we seek to determine the role of metal undercoordination in the stability of H2 adsorption on bare Pd surfaces Recalling from above, surface Pd atoms on Pd(111) are least undercoordinated and exhibit the weakest H2 binding, while a single Pd adatom on Pd(111) is most undercoordinated and exhibits the strongest H2 binding. A summary of re levant adsorption energies is presented in Table 5 2 Therefore, considering only the coordination of the bare Pd surfaces and the H2 adsorption energies, a trend emerges whereas more undercoordinated (or coordinatively unsaturated) metal atoms exhibit st ronger H2 binding. Table 5 2 Selected properties of adsorbed H2 on Pd metal and PdO(101) surfaces Eads (kJ/mol) H H bond length () H2 center of mass to Pd () Pd d band center after H2 adsorption (eV) Pd d band center shift (eV) Pd(111) 23.2 0.86 1.69 2.31 0.72 Pd(110) 36.1 0.85 1.72 1.85 0.37 PdO(101) 46.7 0.84 1.75 2.80 1.26 1 Pd / Pd(111) 47.7 0.84 1.72 1.49 0.42 The charge density difference for H2 adsorbed on the three bare Pd surfaces is shown in Figure 5 9 Qualitatively, the same features are observed in all three cases: significant charge depletion above and below the Pd (aligning approximately with the lobes of the dz 2 orbital), slight charge depletion above the H2, and charge accumulation between the Pd and H2. Once again, this combination is indicative of a donation/backdonation bonding mechanism As with PdO(101), a similar comparison can be made between H2/Pd and H2/RuO2(110)  where the same features are observed in the charge density
90 difference in both systems Despite these similarities, however one notable quantitative difference is that charge depletion surrounding the Pd atom is greatest for H2 on Pd(111) and least for the single Pd adatom The net effect in all three cases, however, is that 0.05 0.01 electrons are transferred from the Pd atom to the H2 molecule. Figure 5 9 Charge density difference plot for (a) H2/Pd(111), (b) H2/Pd/Pd(111), and (c) H2/Pd(110) with H2 parallel to cus Pd rows Note areas of charge accumulation between the surface and adsorbed H2, as well as areas of ch arge depletion both above the adsorbed H2 and around the surface Pd atom. The projected density of states ( pDOS ), shown in Figures 5 10a 5 10c, confirm the strong interaction between adsorbed H2 and the bare Pd surface The peak residing below the d band at about 8 eV corresponds to the bonding interaction, while the overlapping features between +4 and +7 eV are indicative of the antibonding interaction. The orbital decomposed pDOS (not shown) illustrates that, just as for H2 on PdO(101),
91 Figure 510. Projected density of states of Pd d states and H states for H2 adsorbed on Pd(111), Pd/Pd(111), and Pd(110). The Fermi level is set to zero energy -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.5 1.0 1.5 2.0 pDOS (states/eV)Energy (eV) Pd d H -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.5 1.0 1.5 2.0 pDOS (states/eV)Energy (eV) Pd d H (a) (b)
92 Fi gure 510 Continued the Pd dz 2 dxz orbital Interestingly, Pd(111) shows the most antibonding states, followed by Pd(110), and then Pd/Pd(111) In addit ion, we calculated the d band centers of the Pd surface atom to which H2 is bonded; the results are illustrated in Figure 5 11. A linear relationship exists between the adsorption energy of an H2 molecule and the Pd surface atom d band center, where H2 adsorption energy increases with d band centers approaching the Fermi level This trend is in agreement with previous studies  which indicate that d band centers can be used as a measure of reactivity and that d band centers further from the Fermi level correlate with weaker chemisorption energies. -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.5 1.0 1.5 2.0 pDOS (states/eV)Energy (eV) Pd d H (c)
93 Figure 5 11 Molecular H2 adsorption energy versus Pd d band center both before and after interaction with H2. These trends suggest that an increase in charge depletion regions surrounding the surface Pd atom, an increase in antibonding states, and Pd d band centers further from the Fermi level are all indicators of weaker H2 chemisorption energies and elongated H H bonds H2 is not merely more weakly bound on less undercoordinated Pd surfaces but instead interacts more strongly Thus, the H2 molecule is more likely to undergo dissociative adsorption via a molecular precursor on less undercoordinated Pd surfaces, such as Pd(111), than on more undercoordinated Pd surfaces, such as a Pd adatom on Pd(111). As mentioned earlier, both a cus Pd atom on PdO(101) and a single Pd adatom on Pd(111) are threefoldcoordinated. As a result, we have chosen to compare the nature of the bonding of H2 to these undercoordinated Pd atoms Despite the fact that the local bonding environment is different, the adsorption energies are remarkably similar: 48 0 10 20 30 40 50 60 -2.4 -2.2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 Pd d-band center (eV)H2 adsorption energy (kJ/mol) Bare Pd H2/Pd
94 kJ/mol for H2 on the Pd adatom and 47 kJ/mol for H2 atop the cus Pd on PdO(101) This helps to reinforce our hypothesis, as well as to extend it beyond only the bare Pd surfaces, that more undercoordinated sites result in stronger H2 binding. Additionally, comparing the charge density differences of H2 on PdO(101) (Figure 5 2 ) and t he Pd adatom (Figure 5 9 b) shows that H2 behaves quite similarly on both surfaces; the charge accumulation and depletion regions are nearly identical Even though the charges on the metal atoms are significantly different Pd is nearly chargeneutral on the bare metals but has a net charge of +0.77 in PdO(101) as a result of the interaction with the neighboring oxygen atoms the amount of charge transfer (0.06 electrons) as determined by a Bader charge analysis is the same on both surfaces Comparing the pDOS for H2 on PdO(101) (Figure 5 3a) and the Pd adatom (Figure 5 10b ) is not trivial because of the additional features introduced by the interaction between the cus Pd atom and adjacent O atoms Ultimately, regardless of differences in the local bondi ng environment, the PdH2 bonding is comparable when H2 is adsorbed on a threefoldcoordinated Pd atom. H2 on Pd5O4/Pd(111) We have also studied hydrogen adsorption on a twodimensional surface oxide, Pd5O4, atop a Pd(111) substrate, which has been previously studied and characterized in Ref  Briefly, seven Pd5O4 unit cells, each with both twofoldand fourfoldcoordinated Pd, are required to achieve commensurability with the Pd(111) substrate. While the fourfoldcoordinated Pd atoms lie along (but not necessarily directly atop) the row of close packed substrate atoms, the twofoldcoordinated Pd atoms lie above bridge or hollo w sites of the substrate. As a result, the twofoldcoordinated Pd atoms are more coordinatively unsaturated than the threefoldcoordinated Pd in PdO(101), and
95 based on our conclusions thus far, H2 should adsorb more strongly to the twofoldcoordinated Pd of Pd5O4 than on PdO(101) Interestingly, we find that the opposite is true Even though H2 adsorbs molecularly on PdO(101), we were unable to stabilize a molecular state on Pd5O4, just as Blanco Rey et al. were unable to stabilize H2 on the xide structure  Any attempts to adsorb H2 on the Pd5O4 surface oxide re sulted in a repulsive interaction which pushed the H2 molecule to a distance of more than 3 from the surface; the adsorption energies of these states were less than 2 kJ/mol On the other hand, dissociated H2 is stable at some sites, as illustrated in F igure 5 12 Figure 5 12 Top views of stable configurations of dissociated H2 on Pd5O4/Pd(111): (a) H2 dissociated across a four fold Pd ( Eads = 79.5 kJ/mol ) (b) H2 dissociated adjacent to a four fold Pd ( 47.6 kJ/mol ) (c) H2 dissociated on O and in a hollow ( Eads = 35.8 kJ/mol ) Though we have not yet been able to determine a dissociation pathway between approaching molecular H2 and stable dissociated H2 on the surface, there remains a possibility that H2 indeed undergoes dis sociative adsorption on the Pd5O4 surface oxide. In order to examine the differences in H2 bonding on the PdO(101) surface and the Pd5O4 surface oxide on Pd(111), we fixed the atomic positions of the H2 molecule at the same height above the Pd5O4 surface and in a similar orientation as the relaxed H2 state on the PdO(101) surface Illustrations of the local geometries of these states are shown in Figure 5 13.
96 Figure 5 13 Illustrations of the local geometries of PdO(101) (a,c) and Pd5O4 (b,d) both before (a,b) and after (c,d) H2 adsorption. Bader charges are also included. One significant difference in the geometries is the atom underneath the cus Pd atom In PdO(101), the cus Pd atom is surrounded by O atoms, whereas in Pd5O4, the cus Pd atom is bounded by O atoms on the surface but a Pd atom (from the underlying Pd(111)) below As a result, the charge on the cus Pd atom is more positive by 0.23 electrons in PdO(101) than in Pd5O4, but the amount of charge transfer (0.08 electrons) is nearly the same as for H2 adsorbed on PdO(101) The more covalent bond between the cus Pd atom on the PdO(101) surface and the O atom below results in fewer electrons at the surface, while the extra electrons at the twofoldcoordinated Pd atom on Pd5O4 i nteract with H2 so as to hinder molecular adsorption. Any differences in bonding
97 due to the adjacent O atoms are expected to be negligible since the local O Pd O geometry is quite similar, and the orientation of adsorbed H2 is perpendicular to the O Pd O axis In comparing the charge density difference plots for H2 on Pd5O4 (Figure 5 17a) and H2 on PdO(101) (Figure 5 2 ), we observe similar charge accumulation and depletion features except that the magnitude of donation/back donation is somewhat reduced on Pd5O4. An evaluation of the pDOS reveals that the H2 above the Fermi level are somewhat diminished for Pd5O4 compared to PdO(101), which is similar to observations by Blanco Rey et al. of the  Since backdonated electrons fill the H2 2 is forced near the Pd5O4 surface, stable molecular H2 states would not be expected on Pd5O4/Pd(111) as on PdO(101). Figure 514 Projected density of states of Pd d states and H states for H2 adsorbed on Pd5O4/Pd(111). The Fermi level is referenced to zero energy -10 -8 -6 -4 -2 0 2 4 6 8 0.0 0.5 1.0 1.5 2.0 pDOS (states/eV)Energy (eV) Pd d H
98 We can also consider the differences in H2 bonding on the bare Pd(111) surface and the Pd5O4 surface oxide on Pd(111); that is, how the Pd5O4 layer modifies the underlying Pd(111) surface toward H2 adsorption. At equal distances above the surface, the donation/backdonation mechanism between H2 and the cus Pd atom is noticeably weaker on Pd5O4/Pd(111) the charge depletion and accumulation regions are diminished in both size and intensity Based on our conclusions from the bare Pd surfaces, we would expect strong er H2 binding on Pd5O4/Pd(111), but the repulsive interaction between the extra electrons around the twofoldcoordinated Pd atom in the Pd5O4 layer results in the inability to stabilize molecular H2 states Since dissociated H2 is stable on Pd5O4/Pd(111), this suggests that H2 could undergo direct dissociative adsorption (without a molecular precursor) on the twodimensional oxide surface. H2 on Pd5O4/Pd(111) versus H2 on Pd5O4 Our inability to stabilize a molecular state for H2 on Pd5O4/Pd(111), even at t wofold coordinated Pd surface atoms, implies that some effect exists between the Pd5O4 layer and the underlying Pd(111) substrate that cannot be sufficiently explained by examining charge density difference plots or Bader charges Alternatively, we have e xamined the charge density difference plot for the bare Pd5O4/Pd(111) system in order to visualize the effect of the twodimensional surface oxide layer on the underlying met al surface, shown in Figure 515. Two notable features are apparent First, an a ppreciable amount of charge accumulation along the Pd5O4/Pd(111) interface is observed (~0.05 e /Pd atom), but this electron spillout is expected at the metal surface. Second, significant charge perturbations, especially around the twofoldcoordinated Pd a toms, are observed. Upon interaction with the Pd5O4 layer, some charge (~0.04 electrons) from the topmost Pd(111) surface atoms is transferred to the twofold-
99 coordinated, and to a lesser extent, the fourfoldcoordinated Pd atoms of the Pd5O4 layer It is possible that this charge transfer and the associated charge density perturbations are significant enough to affect H2 adsorption and prevent stabilization of a molecular state on the twodimensional surface oxide. Figure 5 1 5 Charge density difference plots for bare Pd5O4/Pd(111) illustrating the interaction between the Pd5O4 layer and underlying Pd(111): (a) through the O Pd O plane of twofold coordinated Pd, ( b) through the plane of all twofold coordinated Pd Note the charge accumulation near the Pd5O4/Pd(111) interface, especially near twofold coordinated Pd atoms. To investigate this possi bility, we have also examined H2 adsorption on only the Pd5O4 surface without the underlying Pd(111) substrate. In removing the Pd(111), any electroni c effects resulting from the interaction between the Pd5O4 and Pd(111) are removed. Interestingly, molecular H2 is still very weakly bound. Similar to our attempts on Pd5O4/Pd(111), the H2 molecule is pushed to a distance of more than 3 from the surfac e and the adsorption energies are less than 2 kJ/mol However, dissociated H2 is much more strongly bound in some cases, the adsorption energies are two to three times greater than those with the Pd(111) substrate. Also some previously unstable sites for dissociated H2, shown in Figure 5 16 are stable without the Pd(111).
100 Figure 5 16 Top views of two additional stable configurations of dissociated H2 on Pd5O4 with the underlying Pd(111) substrate removed: (a) H2 dissociated across a twofold coordinated Pd ( 111.8 kJ/mol ) and (b) H2 dissociated across adjacent twofold coordinated Pds ( 136.4 kJ/mol ) As we did previously for Pd5O4/Pd(111), we fixed the atomic positions of the H2 molecule at the same height above the Pd5O4 surface and in a similar orientation as the relaxed H2 state on the PdO(101) surface. In comparing the charge density difference plots for H2 adsorbed on only Pd5O4 (Figure 5 17a) and on Pd5O4/Pd(111) (Figure 5 17b), we observe that, with the Pd(111) removed, the areas of charge accumulation between the surface and adsorbed H2 are somewhat diminished while areas of charge depletion surrounding the surface Pd atom are significantly enhanced. Figure 5 17 Charge density difference plots for H2 fixed on Pd5O4/Pd(111) both (a) with and (b) without the underlying Pd(111)
101 As with the bare Pd surfaces, this enhancement indicates a stronger adsorbatesurface interaction. Intriguingly, we also observe a ~20% reduction in the forces on the fixed H2 molecule in both the l ateral (inplane) and surface normal directions when the Pd(111) is removed. The presence of the Pd(111) substrate induces additional repulsive effects between the Pd5O4 surface oxide and H2 molecule, weakening the adsorbatesurface interaction, and also limits the possible adsorption sites for dissociated H2. H2 on PtOx/Pt(111) To further investigate the nature of the bonding between hydrogen and an undercoordinated surface metal atom in a thin metal oxide, we also examined the adsorption of H2 at the end of a onedimensional PtOx chain on Pt(111), such as those described in Ref  and di scussed in Chapter 3. The Pt atom at the end of the oxide chain is fourfoldcoordinated to the three surrounding O atoms lying in the plane parallel to the surface and one Pt atom on the Pt(111) surface Figure 5 18 Top views of three stable configurations of dissociated H2 at the end of PtOx chains/Pt(111) The adsorption energies are (a) 28.4, (b) 88.2, and (c) 113.8 kJ/mol.
102 Similar to Pd5O4/Pd(111), molecular adsorption is weak (adsorption energies ~1 kJ/mol), even in configurations that are most closely representative of stable H2 on PdO(101), while dissociated H2 is strongly bound. Three stable configurations of dissociated H2 at the end of the PtOx chains are illustrated in Figure 5 18. Interestingly, the configuration shown in Figure 5 18a is the result of placing the H2 molecule next to the Pt atom at the end of the PtOx chain, nearly along the surface normal direction, and allowing it to relax Thus, it is possible that H2 undergoes dissociative adsorption without a molecular precur sor state on this surface as well as on Pd5O4. In fact, the adsorption energy for dissociated H2 onto two O atoms at the end of a PtOx chain is 114 kJ/mol, which is comparable to 112 kJ/mol on Pd5O4. Again, as we did for Pd5O4/Pd(111), we fixed the atomic positions of the H2 molecule at the same height above the PtOx/Pt(111) and in a similar orientation as the relaxed H2 state on the PdO(101) surface to examine the electronic effects Figure 5 19 Charge density different plots for H2/PtOx chains/Pt(111) with H2 fixed to mimic the geometry of H2 adsorbed on PdO(101): H2 across the upwardly buckled Pt atom at the end of the PtOx chain, interacting with (a) O from both the p (2 1) row and near hcp and (b) only O from the p (2 1) row Note areas of charge accumulation between the surface and H2, as well as areas of charge depletion both above the H2 and around the surface Pd atom.
103 The amount of charge transfer from the Pt atom (0.08 electrons) is the same as on Pd5O4/Pd(111), though we observe s ome subtle differences in the charge density difference plots, illustrated in Figure 5 19. Most of the charge accumulation is around the upwardly buckled Pt atom instead of between the Pt atom and H2 molecule, which may be a result of the open geometry at the end of the chain, and there is also a notable polarization of the H2 molecule.
104 CHAPTER 62MOLECULAR ADSORPTION OF SMALL ALKANES ON A PALLADIUM OXIDE (1 01) THIN FILM: EVIDENCE OF SIGMA COMPLEX FORMATION Introduction The initial CH bond activation of alkanes on metal based catalysts has attracted widespread interest due to the desire to more effectively utilize saturated hydrocarbons as energy sources as well as feedstocks for chemical production. Because initial C H bond cleavage is oft en a ratedetermining step in the catalytic processing of alkanes, extensive efforts have been devoted toward understanding alkane adsorption in detail, with most studies focusing on clean transition metal surfaces  In contrast, the interactions of alkanes with metal oxide surfaces have not been widely explored, largely because alkanes interact weakly with many oxide surfaces and are hence difficult to activate under ultrahigh vacuum (UHV) conditions Palladium oxide (PdO) is an important exception as this material is highly active toward the complete oxidation of alkanes In fact, prior studies conducted at commercially relevant pressures demonstrate that the formation of PdO is responsible for the exceptional activity of supported Pd catalysts in the catalytic combustion of natural gas in excess oxyg en [1 12] This finding provides substantial motivation for studying the surface chemistry of PdO in detail In particular, investigations with well defined PdO surfaces provide opportunities for gaining insights into the mechanisms for alkane a ctivation on transition metal oxides as well as clarifying the oxide surface properties that enhance C H bond activation. 2Reproduced in part with permission from J.F. Weaver, C. Hakano glu, J.M. Hawkins, A. Asthagiri, J Chem. Phys 132 (2010) 024709. Copyright 2010 American Institute of Physics.
105 In recent work, Weaver et al. observed that the initial C H bond cleavage of propane is highly facile on a PdO(101) thin film grown on P d(111) in UHV  The results demonstrate that the initial diss ociation of propane on the PdO(101) surface occurs by a precursor mediated mechanism in which a molecularly adsorbed state of propane acts as the precursor to initial C H bond cleavage. Temperature programmed desorption (TPD) measurements reveal that the molecular precursor is strongly bound relative to physisorbed propane on Pd(111), leadi ng them to suggest that dative bonding interactions contribute to the binding of the alkane precursor on PdO(101) In the present study, the adsorption of methane, ethane, and propane on PdO(101) is examined both experimentally and computationally, and further evidence that dative bonding strengthens the binding of small alkanes on PdO(101) is presented. Dative bonding between alkane molecules and transition metal comple xes is well known [109 111] These bonding interactions produce compounds known as alkane complexes that can serve as key intermediates in the initial C H bond cleavage of alkanes by mononuclear transition metal compounds Bond strengths for alkane complexes are typically between 20 and 50 kJ/mol  The dative bonding inter action that produces an alkane complex generally invol ves electron donation from C H bonds of the alkane molecule into empty d orbitals of the metal center, and, in some cases, back donation of electrons from filled d orbitals of the metal into C H bonds Among the few examples that have been reported, the adsorption of H2 on RuO2(110) provides the most comprehensive example of complex formation on a solid surface [92,112] Recent work by Blanco Rey et al.  as well as that presented in Chapter 5 also demonstrates that dihydrogen experiences relatively strong dative
106 bonding on PdO(101), which suggests the possibility that the surfaces of late transition metal oxides have properties that generally favo r coordinate bond formation with saturated molecules While it is reasonable to expect that dative interactions would also influence the bonding of alkanes on transition metal or transition metal compound surfaces, clear evidence for such bonding is rather limited For example, several studies have reported softening of C H vibrational modes of alkanes upon adsorption on transition metal surfaces [113 1 22] However, work by Fosser et al.  provides evidence that C H mode softening can occur even when the attractive dispersion interaction dominates the alkanesurface binding In this case, the dispersion interaction can bring the molecule close enough to the surface to cause charge flow from metal states into Rydberg molecular orbitals, resulting in C H mode softening. Thus, the C H mode softening seen on metal surfaces is not necessarily a consequence of the type of dative bonding that characterizes alkane complexes. Recent studies have established systematic trends in the desorption kinetic parameters for alkane molecules that are physically adsorbed on solid surfaces [123,124] As demonstrated in the present study, these trends provide a benchmark for comparison with systems in which dative bonding may affect the alkanesurface binding. At low coverage, straight chain alkane molecules adopt a flat lying geometry when physisorbed on a close p acked surface  This configuration maximizes the attractive interaction between the molecule and surface by allowing each CHn group to reside w ithin the attractive region of the moleculesurface potential In this simplified model, each CHn group contributes equally to the moleculesurface interaction, resulting
107 in alkane binding energies E that increase linearly with the chain length N Further, a plot of E versus N should have a zero intercept if each CHn group makes an additive contribution to the total binding energy Recently, Tait et al. [123,124] reported extensive investigations of the desorption of physisorbed alkanes from MgO(100), Pt(111) and C(0001) and confirmed that the E versus N relationships are linear with zero intercepts Th ese workers developed a so called inversionoptimization method to determine both the prefactor and (coveragedependent) activation energy for desorption from TPD spectra obtained at different initial coverages A key finding from their analysis is that the prefactors for desorption increase significantly with increasing chain length, which is consistent with earlier molecular dynamics simulations of alkane desorption [125,126] Compared with an analysis that assumes a molecule independent prefactor, the effect of the increasing prefactors is to increase the slope of the E versus N curve, causing this curve to exhibit the expected zero intercept In the context of transition state theory, the prefactors for desorption increase with chain length because the densities of translational and rotational states of gaseous alkane molecules increase substantially with increasing molecular size, resulting in larger entropy changes upon desorpt ion Tait et al.  also report formulas, derived from transition state theory, that give limiting values of the desorption prefactors for several alkanes as a function of the surface temperature. The minimu m and maximum prefactors correspond to adsorbed molecules that are fully mobile versus completely immobile, respectively Overall, the work of Tait et al. [123,124] provides important guidance in interpreting alkane TPD spectra and
108 identifying systems that deviate from the systematic trends established for linear alkanes physisorbed on close packed surface s. In the present study, the molecular adsorption of methane, ethane, and propane on PdO(101) was investigated using TPD measurements and DFT calculations The TPD results show that each of these small alkanes adsorb in a molecular state that is more strongly bound than alkanes physically adsorbed on Pd(111), and that the E versus N relationship has a nonzero intercept between 22 and 26 kJ/mol, which is cons istent with dative bonding contributing to the alkane binding energy on PdO(101) For more details, see Ref  DFT calculations predict that these small alkanes bind to coordinatively unsaturated (cus) Pd atoms through a donor acceptor interaction, resulting in an adsorbed state that is analogous to an alkane complex Computational Details All of the DFT calculations in this chapter are performed using VASP (Vi enna ab initio simula tion package)  We use the projector augmented wave (PAW) pseudopotentials [59,60] provided in the VASP database Calculations have been done using the generalized gradient approximation Perdew Burke Ernzerhof (GGA PBE) exchangecorrelation functional  W e have tested select configurations with both the Perdew Wang (GGA PW91) and local density approximation (LDA) functional. GGA PW91 yields adsorption energies that are 24 kJ/mol higher than those obtained with GGA PBE, but the differences are insufficient to affect the conclusions discussed in this chapter. LDA yields significantly larger adsorption energies (four to five times those predicted with GGA) as expected, but the trends observed are the same. Additionally, we are mindful that these exchangecorrelation functionals fail to describe longrange van der Waals energies which are important for organic molecules such as alkanes
109  As discussed in further detail later in the chapter we use DFT to probe the dative bonding between alkane molecules and the PdO(101) surface While the dispersive interactions provide a substantial portion of the adsorption energy, we would expect these interactions to be less sensitive to the moleculesurface geometry compared with the dative bonding interaction. Therefore, the adsorption geometry is expected to be relatively accurate, whi le the DFT adsorption energy can be approximately assigned to the entire dative bonding contribution and will be a lower bound. A planewave expansion with a cutoff of 400 eV is used, and the total energy calculations are done using a residual minimization method with direct inversion in the iterative subspace (RMMDIIS) for electronic relaxations, accelerated using Methfessel Paxton Fermi level smearing with a Gaussian width of 0.1 eV  The positions of the atoms are relaxed using a limited memory BroydenFletcher GoldfarbShanno (LBFGS) optimization method  until the forces on all unconstrained atoms are less than 0.03 eV/ While experimentally the PdO(101) film is grown on the Pd(111) surface, the oxide film is sufficiently thick (13 ) that we assume the Pd(111) substrate may be ignored 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. Instead, we represent the PdO(101) film with four layers, which corresponds to a thickness of approximately 9 The bottom la yer is fixed, but all other atoms in the oxide film and the alkane adsorbate are allowed to relax We use a vacuum spacing of 20 which is sufficient to minimize any spurious periodic interactions in the surface normal direction. The PdO(101) film is o btained from the relaxed PdO bulk structure, but the film is strained ( a = 3.057 b = 6.352 ) to match the reported experimental structure  A
110 421 (241) Monkhorst Pack k point mesh  was used for the 14 (22) surface unit cell We have evaluated the zero point corrections for methane i n its favored configuration on PdO(101) and find the contribution is less than 2 kJ/mol The zero point correction is sufficiently small that it does not affect the conclusions drawn in this chapter and is excluded in our DFT values for adsorption energy. Results and Discussion Methane A dsorption on PdO(101) We used DFT to investigate the bonding of alkanes on the PdO(101) surface Our main goal is to determine the extent to which dative bonding contributes to the alkanesur face interaction on PdO(101) ve rsus Pd(111) We emphasize that the optimized geometries and binding energies are likely to deviate from those of the actual adsorbed molecules due to neglect of dispersion interactions in the DFT calculations Nevertheless, the DFT calculations reveal k ey differences in the dative bonding of alkanes on PdO(101) versus Pd(111) and thereby help to clarify the origins for the stronger binding that is observed on the oxide surface Figure 6 1 shows a representation of the optimized bonding configuration of CH4 on the PdO(101) surface The CH4 molecule adopts a so called 2(H,H) configuration  on PdO(101) in which a H C H bond angle straddles a cus Pd atom, resulting in a four center interact ion The plane defined by the downward oriented H C H bond angle aligns parallel to the cus Pd row The CH4 molecule can also bond such that one of the C H bonds aligns parallel to the surface normal direction, but these configurations are significantly less stable than the 2(H,H) configuration. Prior electronic structure calculations predict that the 2(H,H) config uration is preferred in alkane complexes
111 formed with Pd atoms  Pd+2 ions  and neutral PdO dimers  as well as various mononuclear Pd compounds  Figure 6 1 Top and side view of the 2(H,H) configuration of CH4 adsorbed on PdO(101) determined by DFT. The binding energy is 16.2 kJ/mol. The DFT calculations predict a binding energy of 16.2 kJ/mol for CH4 adsorbed on PdO(101), which is significantly less than the 39 kJ/mol binding energy estimated from the TPD measurements The low binding energy predicted by DFT results largely from the absence of dispersion interactions in the electronic structure calculations For example, DFT predicts a binding energy of less than 3 kJ/mol for methane on Pd(111), whereas the actual value is es timated as 16.4 kJ/mol  Thus, although the values are underestimated, DFT correctly predicts stronger binding of methane on PdO(101) compared with Pd(111) Interestingly, the DFT calculations suggest that the dispersion interaction accounts for nearly 60% of the methane binding energy on PdO(101), which agrees reasonably well with the nonzero offsets observed in the E versus N plots 
112 Figure 6 2 (a) Charge density difference plot (electrons/3) of the plane defined by the upward oriented H C H bond angle predicted by DFT for CH4 adsorbed on PdO(101) in the 2(H,H) configuration, and (b) pDOS of CH4 electronic states and Pd d states for the 2(H,H) configuration of CH4 on PdO(101).
113 The DFT calculations reveal that the bonding of CH4 on PdO(101) is analogous to the bonding in an alkane complex For example, the charge density difference plot ( Figure 6 2 a ) shows charge accumulation between a cus Pd atom and the CH4 molecule as well as depletion of charge near the Pd atom and the H atoms of the methane molecule. The decrease in electron density near the H atom s suggests that electron donation oc curs from bonding orbitals of the CH4 molecule toward the cus Pd atom, with charge accumulating between the molecule and the cus Pd atom In addition, the pronounced charge depletion around the Pd atom implies that the bonding also involves back donation from the Pd atom to anti bonding orbitals of the CH4 molecule. The projected density of states (pDOS) plot confirms that the CH4PdO(101) bonding interaction involves both donation and back donation ( Figure 6 2 b ) A relatively strong interaction between t he highest occupied CH4 molecular orbitals (2t1) and the Pd d states is evident from the overlap of peaks at 6.1 eV Inspection of the orbital decomposed p DOS reveals that a preferred interaction between Pd d states and the C 2 py and 2pz orbitals breaks the degeneracy of the three 2t1 hybrid molecular orbitals and causes the associated peak in the pDOS to split into two components centered at 6.1 and 5.6 eV The C and H s states and the C 2pz state contribute to both peaks, whereas the C 2 px state contributes mainly to the peak at 5.6 eV and the C 2py state s contribute mainly to the peak at 6.1 eV The latter peak overlaps a sharp peak in the d band, indicating an interaction between these states The pDOS also shows that the d stat es of the surface interact only weakly with the 2a1 orbital of methane located at 13.1 eV.
114 Several small peaks between 3 and +6 eV are also evident in the pDOS of CH4 and overlap with Pd d states The CH4 states between +3 and +6 eV appear to arise prim arily from the original anti bonding molecular oribtals However, the peaks at about 2 and +0.5 eV likely represent new anti bonding states that result from the relatively strong interactions between filled d states and the occupied 2t1 orbitals of the m olecule. In this case, the pDOS calculation indicates that one of anti bonding states, centered at 2 eV, is occupied due to back donation from the surface. Changes in the molecular geometry and vibrational frequencies demonstrate that the CH4PdO(101) bonding interaction weakens the two C H bonds that coordinate with the cus Pd atom For example, the lengths of the downward oriented C H bonds stretch by 1.55% relative to their gas phase value, and the corresponding H C H bond angle increases by 5.6 The calculations also predict pronounced changes in the vibrational motions of the downward oriented C H bonds Three modes an asymmetric stretch, a symmetric stretch and an asymmetric bend mainly involve motions of the downward oriented C H bonds and the frequencies of these modes are lower by 204, 161, and 107 cm1, respectively, than their gas phase values For comparison, the frequencies of the remaining vibrational modes change by less than 38 cm1 upon adsorption. Also, the vibrational frequencies of CH4 adsorbed on Pd(111) are lower by no more than 30 cm1 from the gas phase values The redshifts of the three strongly perturbed modes of CH4 on PdO(101) are considerably larger than those reported for alkanes physisorbed on m etal surfaces [108,121] E thane and P ropane A dsorption on PdO(101) Figure 6 3 shows the favored 2(H,H) configurations of ethane and propane adsorbed on the PdO (101) surface. For ethane, the 2(H,H) configuration has a binding
115 energy of 21.9 kJ/mol and represents the lowest energy configuration that co uld be identified with DFT. In this configuration, the upward directed methyl group points toward the neighboring cus O atom An 2(H,H) configuration can also be obtained by rotating the ethane molecule such that the spectator methyl group points toward the 4fold O atom The binding energy of this configuration is only 0.4 kJ/mol l ower than that shown in Figure 6 3 Other less stable configurations were identified where the ethane molecule aligns parallel to the cus Pd r ows but have binding energies between 14.2 and 15.4 kJ/mol According to DFT, the binding energy of C2H6 on PdO(101) is 5.7 kJ/mol higher than that of CH4 adsorbed in the 2(H,H) configuration ( Figure 6 1 ) We attribute this small difference to electrostatic interactions between the spectator methyl group of C2H6 and the neighboring O atoms Figure 6 3 Top and side view of the favored 2(H,H) configurations of C2H6 and C3H8 adsorbed on PdO(101) determined by DFT. The binding energies for ethane and propane are 21.9 and 23.7 kJ/ mol, respectively.
116 As seen in Figure 6 3 the favored 2(H,H) configuration for propane involves bonding between the CH2 group and the cus Pd atom, with the molecular plane oriented perpendicularly to the cus Pd row DFT predicts a binding energy of 23.7 kJ/mol for this adsorbed configuration of C3H8, which is only slightly higher than the binding energy for the favored 2(H,H) configuration of C2H6. For comparison, the binding energy is about 13.5 kJ/mol when propane binds to the cus Pd atom through a CH3 group. A key difference between the CH3 versus CH2 down configurations of propane is that the latter puts the spectator CH3 groups closer to neighboring surface oxygen atoms Figure 6 4 Top and side view of the 1(2H) configuration of C3H8 adsorbed on PdO(101) determined by DFT. The binding energy of this configuration is 27.0 kJ/mol. By aligning its molecular plane parallel to the cus Pd atom row, the propane molecule can bind in a configuration that is slightly (~3.3 kJ/mol) more favorable than the CH2 2(H,H) configuration sh own in Figure 6 3 In the resulting adsorbed state
117 (Figure 6 4 ), one H atom of each CH3 group coordinates with a cus Pd atom, producing two coordinate bonds per molecule where each H P d bond may be classified as an 1 configuration. The total binding energy for this adsorbed geometry is 27.0 kJ/mol Because their total binding energies are similar, the DFT results suggest that the CH2 2(H,H) and 1(2H) configurations will be nearly equally populated. The similar bi nding energies also indicate that each 1 H Pd interaction is weaker than the bonding achieved in the CH2 2(H,H) configuration. However, the C H bonds stretch by similar amounts (2.2% versus 2.4%) in the CH2 2(H,H) and 1(2H) configurations, respectivel y, indicating that the secondary C H bonds in the 2(H,H) configuration are perturbed to a similar extent as the pri mary C H bonds involved in the 1(2H) interaction. An implication is that C H bond softening resulting from the moleculesurface interaction has a negligible effect on the selectivity of propane C H bond activation on PdO(101) Comparison between Experiment and C omputation A comparison of the experimentally and computationally determined binding energies indicates that the dispersion inter action contributes significantly to the alkanePdO(101) interaction, accounting for between 50% and 70% of the total binding energy for the small alkanes studied here An implication is that both the dispersion interaction and dative bonding are important in determining the optimum binding configurations, and hence that the actual configurations may deviate from those reported here. However, even though the dispersion interaction contributes significantly to the alkane binding energies, its more diffuse c haracter compared with the dative interaction may cause only minor perturbations from the optimal configurations determined using DFT. The good agreement between the DFT predictions and the experimental results generally supports this idea.
118 The DFT result s show that the strength of the coordinate bonding increases only weakly with increasing chain length for the alkanes studied here. In fact, electrostatic interactions between spectator groups and the surface atoms appear to be responsible for the slight increases that are observed. This trend agrees with the experimental results which suggest a nearly constant enhancement in the binding energies of methane, et hane, and propane on PdO(101) versus Pd(111) Furthermore, the DFTpredicted binding energies are similar in magnitude to the differences in alkane binding energies on PdO(101) versus Pd(111) that we estimated from the experimental TPD data. Thus, the DFT calculations support the co nclusion that the formation of complexes is responsible for the stronger binding of alkanes on PdO(101) compared with Pd(111), and that the strength of this bonding is similar for methane, ethane, and propane. Methane Dissociation on PdO(101) We briefly discuss mechanisms for C H bond breaking as CH4 dissociates into C H3 and OxH groups On PdO(101), we evaluated three possible mechanisms in which the CH4 molecule begins in an 2(H,H) configuration. In the first mechanism, illustrated in Figure 6 5 a, the CH4 molecule rotates on an axis parallel to the cus Pd rows such that one of the upwardoriented H atoms approaches the adjacent cus O atom, and then dissociates. This pathway is least favored; the dissociation barrier is 167 kJ/mol. In the second mechanism, illustrated in Figure 6 5 b, the CH4 molecule rotates on an axis approximately parallel to the surface normal while one of the downwardoriented H atoms dissociates onto the secondnearest cus O atom. This pathway is somewhat more favored than the first; the dissociation barrier is 122 kJ/mol. In the third mechanism, illustrated in Figure 6 5 c, the CH4 molecule rotates on an axis
119 approximately parallel to the surface normal such that one of the downwardoriented H atoms approaches the adjacent cus O atom, and then dissociates. This pathway is the most favorable, as the dissociation barrier is only 64.2 kJ/mol. Figure 65. Initial, transition, and final states for the dissociation of CH4 into CH3 and OxH groups on PdO(101). The dissociation barriers are (a) 167 kJ/mol, (b) 122 kJ/mol, and (c) 64.2 kJ/mol. (a) (b)
120 Figure 6 5 Continued Methane Adsorption on Modified PdO(101) Earlier, we considered CH4 adsorption on the PdO(101) surface, but we also wish to explore the effects of metal substitution and metal adatoms on PdO(101) Specifically, we consider substituting one cus Pd atom with similar transition metals in groups 811 (Pt, Cu, Ir, Ru, Rh, Ag, Au, and Os), a fully substituted analogous PtO(101) surface, as well as Pt and Pd adatoms on the PdO(101) surface Since we have previously est 2(H,H) configuration (see Figure 6 1 ) is most stable for CH4 adsorption, we will only study this configuration on the modified PdO(101) surfaces The binding energies for CH4 on each of the modified PdO(101) surfaces are listed in Table 6 1 In general, substitutions of group 11 ions resulted in very weak bonding while substitutions of group 8 and 9 ions, as well as Pt, resulted in stronger bonding. Group 9 ions exhibit the greatest enhancement, where binding energies are approximately double those on unmodified PdO(101). (c)
121 Table 6 1 Binding energies for CH4 adsorbed on modified PdO(101) surfaces in a n 2(H,H) configuration. X/PdO(101) indicates that X is an adatom on PdO(101), while X PdO(101) indicates that X is substituted for one cus Pd atom of PdO(101) Values in parentheses are for CH4 adsorbed atop an adatom on PdO(101) in a 2(C,H) configuration. Surface Binding energy (kJ/mol) PdO(101) 16.2 PtO(101) 20.9 Pd/PdO(101) 18.3 ( 21.5 ) Pt/PdO(101) 12.1 ( 35.9 ) Pt PdO(101) 28.1 Cu PdO(101) 1.7 Ir PdO(101) 36.4 Ru PdO(101) 27.2 Rh PdO(101) 32.2 Ag PdO(101) 1.5 Au PdO(101) 0.6 Os PdO(101) 25.8 We attempted to determine a trend for the binding energies based on electronegativity or Bader charges but were unable to do so In all cases, except for substitutions of group 11 ions, the C H bonds stretch to a similar extent (2.12.8%) as on the unmodified PdO(101) surface Interestingly, rotating the CH4 molecule atop an adatom by 90 about the surface normal results in a so called 2(C,H) configuration which is more strongly bound than the 2(H,H) configuration Additional work should be performed to determine if the 2(C,H) configuration is the most stable structure of CH4 adsorbed atop a metal adatom on PdO(101). Upon substitution of the cus Pd atom with Pt (Pt PdO(101)), the dissociation mechanism is nearly identical to PdO(101), but the dissociation barrier is reduced to 46.3 kJ/mol Similarly, for a Pt adatom on PdO(101) (Pt/PdO(101)), the mechanism is again nearly identical, but the dissociation is barrierless Clearly, modifying the PdO(101) surface alters its ability to activate C H bond breakage in methane, but more work must be performed to fully understand the effects of surface modifications.
122 Summary We inves tigated the molecular adsorption of methane, ethane, and propane on a PdO(101) thin film using TPD and DFT calculations The TPD data shows that alkanes 1 state on PdO(101) have higher binding energies than alkanes physisorbed on Pd(111) Based on an analysis of the TPD spectra using limiting desorption prefactors, we estimate that the alkane binding energies on PdO(101) increase linearly with increasing chain length up to N = 3, but that the E versus N relation has a nonzero intercept between about 22 and 26 kJ/mol We attribute this constant offset to a dative bonding interaction that is similar in magnitude for each alkane, and suggest that the strength of the dispersion interaction with PdO(101) increases with alkane chain length for the molecules studied. We estimate that the formation of complexes accounts for about 30% to 50% of the total binding energy of these small alkanes adsorbed on PdO(101) DFT calculations predict that the alkanes bind on PdO(101) by forming coordinat e bonds with cus Pd atoms The resulting adsorbed s pecies are analogous to alkane complexes in that the alkanePd coordination involves electron donation from C H bonds to the Pd center as well as back donation from Pd d states into localized anti bonding states According to DFT, methane and ethane achieve maximum binding energies on PdO(101) by adopting an 2(H,H) configuration on top of a cus Pd atom Propane also binds favorably in the 2(H,H) configuration through the CH2 group and can attain a comparable bin ding energy by adsorbing in an 1(2H) configuration along the cus Pd row The bindi ng energies for alkanes in the 2(H,H) configuration lie in a range from about 16 to 24 kJ/mol, which agrees well with the constant of fsets es timated from TPD data Because complex formation weakens alkane C H bonds, the DFT
123 calculations suggest that the moleculesurface interaction may assist in the initial activation of C H bonds on PdO(101) This finding is consistent with recent evidence that the strongly bound molecular state of propane on PdO(101) serves as the precursor to initial bond cleavage  Efforts are currently underway to continue study ing how surface modifications influence the coordinate bonding of alkanes on PdO(101) and the barriers for C H bond cleavage.
124 CHAPTER 7 CONCLUSIONS Density functional theory (DFT) calculations have been used to examine the initial stages of oxidation of the Pt(111) surface, the adsorption of hydrogen on Pd(111) and its oxides, the dissociative chemisorption of hydrogen on PdO(101), and the molecular adsorption of methane, ethane, and propane on PdO(101). In the initial stages of oxidation of Pt(111), we predict that subsurface oxygen is not the precursor to oxidation, but instead, there is a strong preference for the formation and growth of onedimensional Pt oxide chains As oxygen atoms aggregate between the close packed oxygen rows, they induce large buckling and charge modification of the surface Pt atoms, and the resulting oxide compound grows as a onedime nsional chain running parallel to the oxygen rows of the p (2 1) structure These new Pt oxide chain structures are more stable than previously reported configurations of chemisorbed oxygen atoms [45,46] The structures we have identified reproduce many of the features of recent STM images of O/Pt(111)  and suggest a novel precursor mechanism to the oxidation of metal surfaces involving Pt oxide chain formation and growth on terrace s at moderate oxygen coverages Unfortunately, DFT calculations are insufficient to fully explore the behavior of such oxide chains due to system size limitations The development of accurate charge transfer potentials will be useful in this pursuit. We predict that H2 binds relatively s trongly on both Pd(111) and PdO(101) by forming complexes on coordinatively unsaturated Pd sites The nature of the bonding between the H2 molecule and metal center is explored. Our calculations support experimental evidence that H2 dissociates on PdO(101) by a precursor mediated
125 pathway that generates stable products on PdO(101) W e also determine that quantum mechanical tunneling dominates the dissociation of H2 on PdO(101) at low temperature and that differences in tunneling rates are responsible for the large kinetic isotope effect that is o bserve d experimentally. We also predict that small alkanes bind on PdO(101) by forming dative bonds with coordi natively unsaturated Pd atoms, and t he resulting adsorbed s pecies are analogous to alkane complexes in that the alkanePd coordination involves electron donation from C H bonds to the Pd center as well as back donation from Pd d states into localized anti bonding states We conclude that both the dispersion interaction and the formation of complexes contribute to the binding of small alkanes on PdO(101) and estimate that complex formation accounts for 3050% of the total binding energy for the molecules s tudied. The predicted weakeni ng of C H bonds resulting from complex formation may help to explain the high activity of PdO surfaces toward alkane activation.
126 LIST OF REFERENCES  R.B. Anderson, K.C. Stein, J.J. Feenan, L.J.E. Hofer, Ind. Eng. Chem. 53 (1961) 809.  C.F. Cullis, B.M. Willatt, J. Catal. 83 (1983) 267.  R.J. Farrauto, M.C. Hobson, T. Kennelly, E.M. Waterman, Appl. Catal. A 81 (1992) 227.  R.J. Farrauto, J.K. Lampert, M.C. Hobson, E.M. Waterman, Appl. Catal. B 6 (1995) 263.  R. Burch, F.J. Urbano, Appl. Catal. A 124 (1995) 121.  R. Burch, F.J. Urbano, P.K. Loader, Appl. Catal. A 123 (1995) 173.  P. Salomonsson, S. Johansson, B. Kasemo, Catal. Lett. 33 (1995) 1.  J.G. McCarty, Catal. Today 26 (1995) 283.  J.N. Carstens, S.C. Su, A.T. Bell, J. Catal. 176 (1998) 136.  A.K. Datye, J. Bravo, T.R. Nelson, P. Atanasova, M. Lyubovsky, L. Pfefferle, Appl. Catal. A 198 (2000) 179.  R.S. Monteiro, D. Zemlyanov, J. M. Storey, F.H. Ribeiro, J. Catal. 199 (2001) 291.  H. Gabasch, K. Hayek, B. Klotzer, W. Unterberger, E. Kleimenov, D. Teschner, S. Zafeiratos, M. Havecker, A. KnopGericke, R. Schlogl, B. Aszalos Kiss, D. Zemlyanov, J. Phys. Chem. C 111 (2007) 7957.  A. Alavi, P.J. Hu, T. Deutsch, P.L. Silvestrelli, J. Hutter, Phys. Rev. Lett. 80 (1998) 3650.  A. Eichler, J. Hafner, Phys. Rev. B 59 (1999) 5960.  P. Salo, K. Honkala, M. Alatalo, K. Laasonen, Surf. Sci. 516 (2002) 247.  X.Q. Gong, Z.P. L iu, R. Raval, P. Hu, J. Am. Chem. Soc. 126 (2004) 8.  M.D. Ackermann, T.M. Pedersen, B.L.M. Hendriksen, O. Robach, S.C. Bobaru, I. Popa, C. Quiros, H. Kim, B. Hammer, S. Ferrer, J.W.M. Frenken, Phys. Rev. Lett. 95 (2005) 255505  K. Nakao, S.I. Ito K. Tomishige, K. Kunimori, Catal. Today 111 (2006) 316.  E.D. German, M. Sheintuch, J. Phys. Chem. C 111 (2007) 9184.
127  M.S. Chen, Y. Cal, Z. Yan, K.K. Gath, S. Axnanda, D.W. Goodman, Surf. Sci. 601 (2007) 5326.  J.F. Weaver, H.H. Kan, R.B. Sh umbera, J. Phys.: Condens. Matter 20 (2008) 184015.  D.H. Parker, M.E. Bartram, B.E. Koel, Surf. Sci. 217 (1989) 489.  N. Saliba, Y.L. Tsai, C. Panja, B.E. Koel, Surf. Sci. 419 (1999) 79.  R.B. Shumbera, H.H. Kan, J.F. Weaver, Surf. Sci. 601 (2 007) 4809.  J.F. Weaver, J.J. Chen, A.L. Gerrard, Surf. Sci. 592 (2005) 83.  C.R. Parkinson, A. Walker, C.F. McConville, Surf. Sci. 545 (2003) 19.  E. Lundgren, J. Gustafson, A. Resta, J. Weissenrieder, A. Mikkelsen, J.N. Andersen, L. Kohler, G Kresse, J. Klikovits, A. Biederman, M. Schmid, P. Varga, J. Electron. Spectrosc. Relat. Phenom. 144 (2005) 367.  J. Gustafson, A. Mikkelsen, M. Borg, E. Lundgren, L. Kohler, G. Kresse, M. Schmid, P. Varga, J. Yuhara, X. Torrelles, C. Quiros, J.N. And ersen, Phys. Rev. Lett. 92 (2004) 126102.  E. Lundgren, G. Kresse, C. Klein, M. Borg, J.N. Andersen, M. De Santis, Y. Gauthier, C. Konvicka, M. Schmid, P. Varga, Phys. Rev. Lett. 88 (2002) 246103.  J. Klikovits, E. Napetschnig, M. Schmid, N. Serian i, O. Dubay, G. Kresse, P. Varga, Phys. Rev. B 76 (2007) 045405.  W. Li, B. Hammer, Chem. Phys. Lett. 409 (2005) 1.  N. Seriani, W. Pompe, L.C. Ciacchi, J. Phys. Chem. B 110 (2006) 14860.  T.M. Pedersen, W.X. Li, B. Hammer, Phys. Chem. Chem. Ph ys. 8 (2006) 1566.  R.B. Getman, W.F. Schneider, J. Phys. Chem. C 111 (2007) 389.  A.D. Smeltz, R.B. Getman, W.F. Schneider, F.H. Ribeiro, Catal. Today 136 (2008) 84.  D.S. Sholl, J.A. Steckel, Density Functional Theory: A Practical Introduction, Hoboken, New Jersey, John Wiley and Sons, Inc., 2009.  R.M. Martin, Electronic Structure: Basic Theory and Practical Methods, Cambridge, UK, Cambridge University Press, 2004.  K. Honkala, A. Hellman, I.N. Remediakis, A. Logadottir, A. Carlsson, S. Dahl, C.H. Christensen, J.K. Norskov, Science 307 (2005) 555.
128  R. Schweinfest, A.T. Paxton, M.W. Finnis, Nature 432 (2004) 1008.  K. Umemoto, R.M. Wentzcovitch, P.B. Allen, Science 311 (2006) 983.  P. Hohenberg, W. Kohn, Phys. Rev. B 136 (1964) B864.  W. Kohn, L.J. Sham, Phys. Rev. 140 (1965) 1133.  J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 13244.  J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.  P. Legare, Surf. Sci. 580 (2005) 137.  R.B. Getma n, Y. Xu, W.F. Schneider, J. Phys. Chem. C 112 (2008) 9559.  S. Ovesson, B.I. Lundqvist, W.F. Schneider, A. Bogicevic, Phys. Rev. B 71 (2005) 115406.  A. Eichler, F. Mittendorfer, J. Hafner, Phys. Rev. B 62 (2000) 4744.  H.R. Tang, A. Van der V en, B.L. Trout, Phys. Rev. B 70 (2004) 045420.  H.R. Tang, A. Van der Ven, B.L. Trout, Mol. Phys. 102 (2004) 273.  D.I. Jerdev, J. Kim, M. Batzill, B.E. Koel, Surf. Sci. 498 (2002) L91.  S.P. Devarajan, J.A. Hinojosa Jr., J.F. Weaver, Surf. Sci. 602 (2008) 3116.  M. Todorova, K. Reuter, M. Scheffler, Phys. Rev. B 71 (2005) 195403.  N. Seriani, F. Mittendorfer, J. Phys.: Condens. Matter 20 (2008) 184023.  G. Kresse, J. Furthmuller, Comput. Mater. Sci. 6 (1996) 15.  G. Kresse, J. F urthmuller, Phys. Rev. B 54 (1996) 11169.  G. Kresse, J. Hafner, Phys. Rev. B 49 (1994) 14251.  G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.  G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.  P.E. Blchl, Phys. Rev. B 50 (1994) 17953.  M. Methfessel, A.T. Paxton, Phys. Rev. B 40 (1989) 3616.  D. Sheppard, R. Terrell, G. Henkelman, J. Chem. Phys. 128 (2008) 134106.  H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188.
129  A.D. Polli, T. Wagner, T. Gemming, M. Rhle, Surf. Sci. 448 (2000) 279.  R.A. Olsen, G.J. Kroes, G. Henkelman, A. Arnaldsson, H. Jonsson, J. Chem. Phys. 121 (2004) 9776.  G. Henkelman, B.P. Uberuaga, H. Jonsson, J. Chem. Phys. 113 (2000) 9901.  G. Henkelman, H. Jonsson, J. Chem. Phys. 113 ( 2000) 9978.  Z.H. Gu, P.B. Balbuena, J. Phys. Chem. C 111 (2007) 9877.  D.C. Ford, Y. Xu, M. Mavrikakis, Surf. Sci. 587 (2005) 159.  G. Henkelman, A. Arnaldsson, H. Jonsson, Comput. Mater. Sci. 36 (2006) 354.  E. Sanville, S.D. Kenny, R. Smith, G. Henkelman, J. Comput. Chem. 28 (2007) 899.  W. Tang, E. Sanville, G. Henkelman, J. Phys.: Condens. Matter 21 (2009) 084204.  S. Helveg, H.T. Lorensen, S. Horch, E. Laegsgaard, I. Stensgaard, K.W. Jacobsen, J.K. Nrskov, F. Besenbacher, Surf. Sci. 430 (1999) L533.  J.G. Wang, W.X. Li, M. Borg, J. Gustafson, A. Mikkelsen, T.M. Pedersen, E. Lundgren, J. Weissenrieder, J. Klikovits, M. Schmid, B. Hammer, J.N. Andersen, Phys. Rev. Lett. 95 (2005) 256102.  B.K. Min, A.R. Alemozafar, M.M. Biener, J. Biener, C.M. Friend, Top. Catal. 36 (2005) 77.  J. Biener, M.M. Biener, T. Nowitzki, A.V. Hamza, C.M. Friend, V. Zielasek, M. Baumer, ChemPhysChem 7 (2006) 1906.  M.M. Biener, J. Biener, C.M. Friend, Langmuir 21 (2005) 1668.  G. Zheng, E.I. Altman, Surf. Sci. 462 (2000) 151.  H. Gabasch, W. Unterberger, K. Hayek, B. Klotzer, G. Kresse, C. Klein, M. Schmid, P. Varga, Surf. Sci. 600 (2006) 205.  J.G. Yu, S.B. Sinnott, S.R. Phillpot, Phys. Rev. B 75 (2007) 085311.  J. Tersof f, D.R. Hamann, Phys. Rev. B 31 (1985) 805.  H.H. Kan, J.F. Weaver, Surf. Sci. 602 (2008) L53.  D. Ciuparu, M.R. Lyubovsky, E. Altman, L.D. Pfefferle, A. Datye, Catal. Rev. Sci. Eng. 44 (2002) 593.
130  G. Ketteler, D.F. Ogletree, H. Bluhm, H. Liu, E.L.D. Hebenstreit, M. Salmeron, J. Am. Chem. Soc. 127 (2005) 18269.  J.A. Hinojosa Jr., H.H. Kan, J.F. Weaver, J. Phys. Chem. C 112 (2008) 8324.  H.H. Kan, R.J. Colmyer, A. Asthagiri, J.F. Weaver, J. Phys. Chem. C 113 (2009) 1495.  J.F. Weave r, C. Hakanoglu, J.M. Hawkins, A. Asthagiri, J. Chem. Phys. 132 (2010) 024709.  J.F. Weaver, S.P. Devarajan, C. Hakanoglu, J. Phys. Chem. C 113 (2009) 9773.  M. Knapp, D. Crihan, A.P. Seitsonen, E. Lundgren, A. Resta, J.N. Andersen, H. Over, J. Phy s. Chem. C 111 (2007) 5363.  K. Christmann, Surf. Sci. Rep. 9 (1988) 1.  P.K. Schmidt, K. Christmann, G. Kresse, J. Hafner, M. Lischka, A. Gross, Phys. Rev. Lett. 87 (2001) 096103.  Q. Sun, K. Reuter, M. Scheffler, Phys. Rev. B 70 (2004) 235402.  M. Blanco Rey, D.J. Wales, S.J. Jenkins, J. Phys. Chem. C 113 (2009) 16757.  J.M. Hawkins, J.F. Weaver, A. Asthagiri, Phys. Rev. B 79 (2009) 125434.  C. Hakanoglu, J.M. Hawkins, A. Asthagiri, J.F. Weaver, submitted to J. Phys. Chem.  A.S. Mrtensson, C. Nyberg, S. Andersson, Phys. Rev. Lett. 57 (1986) 2045.  W. Dong, J. Hafner, Phys. Rev. B 56 (1997) 15396.   M. Salmeron, Top. Catal. 36 (2005) 55.  H.F. Busnengo, W. Dong, A. Salin, Phys. Rev. Lett. 93 (2004) 236103.  M.A. Di Cesare, H.F. Busnengo, W. Dong, A. Salin, J. Chem. Phys. 118 (2003) 11226.  V. Ledentu, W. Dong, P. Sautet, G. Kresse, J. Hafner, Phys. Rev. B 57 (1998) 12482. [ 103] H. Niehus, C. Hiller, G. Comsa, Surf. Sci. 173 (1986) L599.
131  R.J. Behm, V. Penka, M.G. Cattania, K. Christmann, G. Ertl, J. Chem. Phys. 78 (1983) 7486.  B. Hammer, J.K. Norskov, Surf. Sci. 343 (1995) 211.  J. Greeley, J.K. Nrskov, M. Mavrikakis, Annu. Rev. Phys. Chem. 53 (2003) 319.  V. Pallassana, M. Neurock, L.B. Hansen, B. Hammer, J.K. Nrskov, Phys. Rev. B 60 (1999) 6146.  J.F. Weaver, A.F. Carlsson, R.J. Madix, Surf. Sci. Rep. 50 (2003) 107.  C. Hall, R.N. Perutz, Ch em. Rev. 96 (1996) 3125.  A.E. Shilov, G.B. Shul'pin, Chem. Rev. 97 (1997) 2879.  J.A. Labinger, J.E. Bercaw, Nature 417 (2002) 507.  J.H. Wang, C.Y. Fan, Q. Sun, K. Reuter, K. Jacobi, M. Scheffler, G. Ertl, Angew. Chem. Int. Ed. 42 (2003) 2151.  J.E. Demuth, H. Ibach, S. Lehwald, Phys. Rev. Lett. 40 (1978) 1044.  F.M. Hoffmann, T.E. Felter, P.A. Thiel, W.H. Weinberg, Surf. Sci. 130 (1983) 173.  F.M. Hoffmann, T.H. Upton, J. Phys. Chem. 88 (1984) 6209.  N.R. Avery, Surf. S ci. 163 (1985) 357.  R. Raval, M.A. Chesters, Surf. Sci. 219 (1989) L505.  S. Lehwald, H. Ibach, Surf. Sci. 89 (1979) 425.  M.K. Weldon, P. Uvdal, C.M. Friend, B.C. Wiegand, Surf. Sci. 355 (1996) 71.  K.A. Fosser, R.G. Nuzzo, P.S. Bagus C. Woll, Angew. Chem. Int. Ed. 41 (2002) 1735.  K.A. Fosser, R.G. Nuzzo, P.S. Bagus, C. Woll, J. Chem. Phys. 118 (2003) 5115.  M.J. Hostetler, W.L. Manner, R.G. Nuzzo, G.S. Girolami, J. Phys. Chem. 99 (1995) 15269.  S.L. Tait, Z. Dohnalek, C.T. Campbell, B.D. Kay, J. Chem. Phys. 122 (2005) 164708.  S.L. Tait, Z. Dohnalek, C.T. Campbell, B.D. Kay, J. Chem. Phys. 125 (2006) 234308.
132  K.A. Fichthorn, R.A. Miron, Phys. Rev. Lett. 89 (2002) 196103.  K.E. Becker, K.A. Fichthorn, J. Chem. Phys. 125 (2006) 184706.  W. Kohn, Y. Meir, D.E. Makarov, Phys. Rev. Lett. 80 (1998) 4153.  M.R.A. Blomberg, P.E.M. Siegbahn, M. Svensson, J. Am. Chem. Soc. 114 (1992) 6095.  M.R.A. Blomberg, P.E.M. Siegbahn, M. Svensson, J. Phys. Chem 98 (1994) 2062.  D.Y. Hwang, A.M. Mebel, J. Phys. Chem. A 106 (2002) 12072.
133 BIOGRAPHICAL SKETCH Jeffery Michael Hawkins was born in November 1982 in Warren, Ohio, to Jeffery and Teresa Hawkins He was raised in Newton Falls, Ohio, where he graduat ed as salutatorian of Newton Falls High School in June 2000 After high school, Jeffery attended The University of Akron in Akron, Ohio, to study chemical engineering. During his undergraduate studies, he completed a cooperative work assignment at the Bridgestone Americas Center for Research and Technology in Akron, Ohio, where he assisted in both improving existing and developing new synthesis processes for polybutadiene polymers He passed the Ohio Fundamentals of Engineering exam in April 20 05 In May 2005, Jeffery graduated magna cum laude with a Bachelor of Science in chemical engineering with a polymer specialization and minors in m athemat ics and chemistry After a relaxing summer, he moved to Florida in August 2005 to begin his graduate studies in chemical engineering at the University of Florida in Gainesville, Florida. Under the supervision of Prof. Aravind Asthagiri, Jeffery used atomic scale, first principles calculations to explore surface science phenomena such as the oxidation and chemical reactivity of platinum and palladium surfaces In May 2010, Jeffery earned his Doctor of Philosophy in chemical engineering. He has joined KBR, Inc., in Houston, Texas, where he is employed as a chemical process engineer.