Oxidation electronics: bond–band–barrier correlation and its applications

Progress in Materials Science 48 (2003) 521–685 www.elsevier.com/locate/pmatsci

Oxidation electronics: bond–band–barrier correlation and its applications§ Chang Q. Sun* School of Electrical and Electronic Engineering, Block S2, Nanyang Technological University, Singapore 639798, Singapore Institute of Advanced Materials Physics and Faculty of Science, Tianjin University, 300072, China

Abstract This report features the recent progress in understanding the behaviour of atoms and valence electrons involved in the process of oxidation, and some technological development driven by the new knowledge. It is initiated and verified that a chemical bond contracts spontaneously at a surface associated with magnitude rise of the bond energy due to the coordination imperfection and that an oxygen atom hybridizes its sp orbitals upon reacting with a solid surface. The former leads to the bond order–length–strength (BOLS) correlation for the physical aspect of a surface and a nano-solid and the latter to a bond–band–barrier (BBB) correlation for chemical reaction. In the process of oxidation, non-bonding lone pairs, anti-bonding dipoles and hydrogen-like bonds are involved, which add corresponding densityof-states (DOS) features to the valence band of the host. Bond forming also alters the sizes and valencies of the involved atoms and causes a collective dislocation of these atoms, which corrugate the morphology or the potential barrier of the surface. Based on the above premises, the oxidation of the low-index surfaces of transition metals Cu, Co, Ag and V, noble metals Rh, Ru, and Pd and non-metallic diamond has been consistently analyzed. Identities probed with various techniques, such as STM, LEED, XRD, STS, PES, TDS, EELS and Raman, have been systematically defined in terms of atomic valencies, bond geometry, valence DOS, bond strength and bond forming kinetics. It is understood that formation of the basic oxide tetrahedron, and consequently, the four discrete stages of bond forming kinetics and the oxygen-derived DOS features, are intrinsically common for all the analyzed systems though the patterns of observations may vary from situation to situation. What differs one oxide surface from another in observations are: (i) the site selectivity of the oxygen adsorbate, (ii) the order of the ionic bond formation and, (iii) the orientation of the tetrahedron at the host surfaces. The valencies of oxygen, the scale and geometrical orientation of the host lattice and the electronegativity of the host elements determine these specific differences extrinsically. §

Supplementary multimedia movie showing the quantified four-stage Cu3O2 bonding kinetics can be found at doi:10.1016/S0079-6425(03)00010-0. * Tel.: +65-6790-4517; fax: +65-6792-0415. E-mail addresses: [email protected] (C.Q. Sun). 0079-6425/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0079-6425(03)00010-0

522

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Extending the premise of sp-orbital hybridization to the reactions of (C, N)–Ni(001) surfaces has led to a novel approach neutralizing the diamond–metal interfacial stress and hence strengthening the diamond–metal adhesion substantially. The BOLS correlation has provided consistent insight into the shape-and-size dependence of a number of properties for nanosolids. The BBB correlation has led to new findings in designing and fabricating materials for photoluminescence, electron emission and ultrahigh elasticity, etc. # 2003 Elsevier Science Ltd. All rights reserved. Keywords: Surface interface; Crystal growth; Chemisorption; Materials design; Oxygen

Contents 1. Introduction ....................................................................................................................525 1.1. Scope ......................................................................................................................525 1.2. Overview.................................................................................................................526 1.3. Challenges...............................................................................................................528 1.3.1. Bond nature and bond forming kinetics.....................................................528 1.3.2. Alteration of atomic valencies .................................................................... 529 1.3.3. Spectroscopes correspondences ..................................................................529 1.3.4. Driving forces behind reconstruction .........................................................530 1.3.5. Work function and inner potential change.................................................532 1.3.6. Factors controlling bond formation ...........................................................532 1.4. Objectives ...............................................................................................................533 2. Principle: bond–band–barrier (BBB) correlation ............................................................534 2.1. Foundations ...........................................................................................................534 2.1.1. Basic concepts.............................................................................................534 2.1.2. Bonding effects ........................................................................................... 536 2.2. Chemical bond: the basic tetrahedron .................................................................... 539 2.3. Valence density-of-state (DOS) .............................................................................. 542 2.4. Surface potential barrier (SPB) .............................................................................. 543 2.4.1. One-dimensional SPB model ...................................................................... 543 2.4.2. 3-D effect with DOS contribution ..............................................................545 2.4.3. Physical indications .................................................................................... 545 2.5. Summary ................................................................................................................546 3. STM and LEED: atomic valencies and bond geometry..................................................547 3.1. Phase ordering ........................................................................................................547 3.2. O–Cu{(001), (110), (111)}....................................................................................... 548 3.2.1. Observations............................................................................................... 548 3.2.2. Analysis ......................................................................................................556 3.2.3. Quantification: bond geometry and bonding kinetics ................................561 3.2.4. Summary ....................................................................................................568 3.3. O–(Rh, Pd)(110) .....................................................................................................570 3.3.1. Observations............................................................................................... 570 3.3.2. Analysis ......................................................................................................573 3.4. O–(Co, Ru)(1010) ...................................................................................................576 3.4.1. Observations............................................................................................... 576 3.4.2. Analysis ......................................................................................................577

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

523

3.5. O–Rh(111) and O–Ru(0001) .................................................................................. 584 3.5.1. Observations............................................................................................... 584 3.5.2. Analysis ......................................................................................................586 3.6. O–Rh(001) and (N, C)–Ni(001).............................................................................. 590 3.6.1. Observations............................................................................................... 590 3.6.2. Analysis ......................................................................................................594 3.6.3. Quantification: driving force and bond stress ............................................597 3.7. O–(Ag, V)(001) .......................................................................................................599 3.7.1. O–Ag(001) ..................................................................................................599 3.7.2. O–V(001) ....................................................................................................601 4. STS and PES: valence DOS ............................................................................................ 604 4.1. Signature generality ................................................................................................ 604 4.1.1. STS .............................................................................................................604 4.1.2. PES, IPES and XPS ................................................................................... 606 4.1.3. Indication ...................................................................................................611 4.2. Specification ...........................................................................................................611 5. TDS: bond nature and bond strength ............................................................................. 612 5.1. Identity similarity ...................................................................................................612 5.2. Specification ...........................................................................................................618 6. EELS and Raman: fingerprints of weak interaction .......................................................619 6.1. EELS: dipole vibration........................................................................................... 619 6.2. Raman: lone-pair in oxides, nitrides and bio-molecules.........................................620 6.3. Confirmation: ultra-elasticity of nitride surfaces ....................................................620 7. Kinetics of bond forming and bond switching................................................................ 622 7.1. Four-stage bond forming kinetics .......................................................................... 622 7.2. Bond switching: O-floating and O-diffusing ...........................................................624 8. Application: I. Bond contraction and charge transport ..................................................625 8.1. Introduction ...........................................................................................................625 8.2. Nano-solid: bond order–length–strength (BOLS) correlation................................626 8.2.1. Principle......................................................................................................627 8.2.2. Application: lattice strain and surface mechanics ......................................628 8.2.3. Other applications ...................................................................................... 634 8.3. Catalytic effect on band-gap expansion..................................................................636 8.3.1. Blue light emission of PZT ......................................................................... 636 8.3.2. O-induced blue-shift in PL ......................................................................... 639 8.3.3. PL of III- and IV-nitride ............................................................................ 641 8.4. Joint size and catalytic effects: PL of nanometric SiO2 ..........................................642 8.5. Work function reduction: cold cathode field emission ...........................................647 8.5.1. Current understanding ............................................................................... 647 8.5.2. Explanation ................................................................................................ 649 8.6. Magnetic enhancement ........................................................................................... 649 9. Application: II. Synthetic diamond................................................................................. 650 9.1. Thermal oxidation ..................................................................................................651

524

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

9.2. Adhesion improvement........................................................................................... 654 9.3. Dielectric relaxation and transition ........................................................................ 655 10. Summary .........................................................................................................................655 10.1. General understanding ........................................................................................... 655 10.1.1. Essential events at a surface ....................................................................... 655 10.1.2. Bond nature and bond forming kinetics.....................................................656 10.1.3. Orientation specificity of the tetrahedron...................................................657 10.1.4. Consequences of bond forming ..................................................................659 10.1.5. Driving forces behind reconstruction .........................................................660 10.1.6. Factors controlling bond formation ...........................................................661 10.2. Capability-enhancement of probing techniques .....................................................661 10.2.1. STM and STS............................................................................................. 661 10.2.2. PES, TDS, EELS and VLEED ..................................................................662 10.3. Findings in applications ......................................................................................... 663 11. Recommendations ...........................................................................................................663 Acknowledgements...............................................................................................................665

Nomenclature
Y (ML) gi a.u. AES APECS AR ARIPES BA BBB BL BOLS BR BZ CN CNT CVD DFT DLC DOS DSIC

Electronegativity Oxygen coverage (unit in Monolayer) Surface-to-volume ratio Atomic unit (e=m=
h=1; 1 a.u.=1 Bohr radii=0.529 A˚; E=27.21 eV) Auger electron spectroscopy Auger photoelectron coincidence spectroscopy Added-row Angular-resolved inverse photoemission spectroscopy Bond angle Bond–band–barrier Bond length Bond order–length–strength Buckled-row Brillouin zone Coordination number Carbon nano-tube Chemical vapor deposition Density function theory Diamond like carbon Density of states Deep submicron integrated circuit

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

E/LDS EELS EF EMT FWHM GGA H-/M-/LEIS HREELS ICISS L LDA L LDOS LEISS MR PED PEEM PES PL PZT Qi RSGF SBC SEM SEXAFS SIB SPA-LEED SPB STM/S TDS TOF UPS V/LEED XPD/S

525

Electron/laser stimulated desorption spectroscopy Electron energy loss spectroscopy Fermi energy Effective-medium theory Full width at half maximum Generalized gradient approximation High-, medium-, and low-energy ion scattering High-resolution electron-energy-loss spectroscopy Impact-collision ion-scattering spectrometry Langmuir (10
6 torr.sec) Local density approximation Local work function Local DOS Low-energy-ion scattering spectroscopy Missing row Photoelectron diffraction Photoelectron emission microscopy Photoelectron spectroscopy Photoluminescence PbZrTi oxide Bond contracting factors Real-space Green’s function method Surface-bond contraction Scanning electron microscopy Surface extended X-ray absorption fine-structure spectroscopy Saturated image barrier Spot analysis LEED Surface potential barrier Scanning tunneling microscopy/spectroscopy Thermal desorption spectroscopy Time-of-flight Ultraviolet photoelectron spectroscopy Very/low-energy electron diffraction X-ray photoelectron diffraction/spectroscopy

1. Introduction 1.1. Scope The report will start, in Section 1, with a brief overview on oxygen adsorption studies. The long-lasting puzzles in this field are summarized, which challenge the current efforts towards generalizing knowledge from various systems observed with various techniques. Section 2 will describe the original approaches

526

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

to oxide-tetrahedron bonding and its effect on the valence density-of-states (DOS) and the surface potential barrier (SPB). The basic conditions for an oxide tetrahedron formation and the effects of bond forming on the charges of surrounding atoms are classified. Some crucial yet often-overlooked events, such as non-bonding lone pairs, anti-bonding dipoles and the hydrogen-like bonds will be emphasized. In Sections 3–7, observations using STM, LEED/XRD, STS, UPS/XPS, TDS, EELS and Raman of a number of typical samples are systematically analyzed based on the core idea of the chemical-bond–valence-band–potential-barrier (BBB) correlation for deeper insight and generalized information. Emphasis will be given to deriving: (i) (ii) (iii) (iv)

The formulae of reaction with specification of individual atomic valencies; Kinetics of charge transportation/polarization; Bond geometry and atomic dislocation; The driving forces and bond strength for surfaces with chemisorbed oxygen; and, (v) The correspondence between the BBB correlation and various signatures of observations.

As a result, two essential concepts are developed. One is the bond contraction at surface or sites surrounding defects where the atomic coordination number (CN) is reduced, and the other is the essentiality of sp-orbital hybridization for an oxygen atom upon interacting with a solid surface, which widens the band-gap by charge transportation and polarization. It is shown that this premise can also be applicable to reactions involving carbon and nitrogen. Sections 8 and 9 will introduce some findings in practical applications driven by the developed BBB and BOLS correlation knowledge. In Section 10, a summary of the main conclusions will be given in responding to the challenges addressed in Section 1. The report will end (Section 11) with recommendations on further extension of the current approaches in materials design. 1.2. Overview The atomic and electronic process of catalytic oxidation plays an essential role in many fields such as environmental chemistry (CO and NO oxidation, radiation protection and ozone layer protection), bioelectronics (DNA folding and protein signaling) and pharmacology (NO regulating and messaging). Oxygen interaction with solid surfaces of metals and non-metals relates to the technical processes of corrosion, bulk oxidation, and heterogeneous catalysis. Studies of these processes laid the foundations for applications in microelectronics (MOSFET gate devices and DSIC (deep submicron integrated circuit) technologies), photo-electronics (photoluminescence, photo-conductance and field emission), magneto-electronics (superconductivity and colossal magneto-resistance) and dielectrics (ferro-, piezo-, pyro-electrics). For both scientific and technological reasons, oxygen interaction with solid surfaces has formed the subject of extensive study over many years [1,2]. Solid surfaces with chemisorbed oxygen have been examined in detail from a macroscopic to an atomistic point of view, and both experimentally and theoreti-

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

527

cally. Various techniques have been used to characterize the atomic and electronic properties. The experimental techniques include:  Crystallography includes low-energy-electron diffraction (LEED), surface X-ray diffraction (XRD), X-ray photoelectron diffraction (XPD), high-, medium-, and low-energy ion scattering (HEIS, MEIS, LEIS)  Microscopy contains scanning tunneling microscopy (STM), photoelectron emission microscopy (PEEM).  Spectroscopy includes the angular-resolved X-ray or ultraviolet photoelectron spectroscopy (ARUPS, or XPS) for the valence DOS features, X-ray photoelectron spectroscopy (XPS) for the energy shift of a core-band, scanning tunneling spectroscopy (STS) for the on-site DOS and the inverse photoelectron spectroscopy (IPES) for surface image states. It also contains the surface extended X-ray absorption fine-structure spectroscopy (SEXAFS) and the impact-collision ion-scattering spectroscopy (ICISS).  Techniques for bond activation and lattice vibration include the thermal, electron and laser stimulated desorption spectroscopy (TDS, EDS, and LDS), as well as high-resolution electron energy loss spectroscopy (EELS). Numerous theoretical approaches have been employed to investigate the details of oxygen chemisorption. Theoretical methods include the semi-empirical effectivemedium theory (EMT) [3], the tight binding theory [4–6], and the first principle method [7,8], including the density function theory (DFT) [9–13]. Usually, the process in which an oxygen atom exchanges electrons with the solid surface is defined as chemisorption otherwise it is physisorption. The kinetics of surface oxidation is generally believed to involve dissociation of the initial oxygen molecules at the surface followed by trapping of the oxygen atoms into the chemisorption-well of potential energy of the surface. The chemisorption of oxygen breaks the host–host surface bonds and then creates new kinds of oxygen–host bonds [14]. In the oxidation of metals, oxygen in the atmosphere is adsorbed onto the surface and reacts with the metallic atoms to form an ionic or ionic–covalent type of compound. To a certain extent, the degree of adsorption and reaction is a function of the orientation of the crystal face exposed to the gas and the partial pressure or activity of the oxygen in the atmosphere. The actual mechanism for the oxidation of each surface was thought to be quite different and very complicated [15]. Therefore, the atomic processes involved in the oxidation were far from clear [16]. The invention of STM and STS has led to enormous impact onto studying the oxidation of metal surfaces on an atomic scale and in real time. In spite of the difficulties in interpreting the STM images, valuable, direct, yet qualitative, information for systems with chemisorbed oxygen has been gained from such observations [17,18]. It is possible to investigate the kinetic and the static features of the chemisorbed systems with the STM and STS and hence to [14]: (a) Distinguish between the different reconstruction models and thus eliminate the inappropriate ones, and, (b) Elucidate the driving force behind such surface phase transitions.

528

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Models derived from STM observations or from the fitting of diffraction (such as LEED and XRD) provide information about static atomic structures of the surface. These structural models succeeded in describing specific situations in terms of the static positions of the adsorbates that were often assumed rigid spheres. The general characteristics of electronic structures and atomic arrangement on a variety of metal oxide surfaces have been now fairly determined [19,20]. There have been many landmark reviews on the progress in this field published recent years [14,16,19,21–31]. 1.3. Challenges Although the physical picture of the oxidation process is now fairly understood, the underlying mechanism for the various observations still needs to be established. Much more needs to be known about the correlation between the chemical bonds, valence DOS, surface morphology and the corresponding properties of an oxide. Generally, our understanding of the nature and kinetics of oxide bonding and its consequences on the behavior of atoms and valence electrons at surfaces is far from complete [14–30]. The following issues have formed the long-lasting puzzles that are still a great challenge. 1.3.1. Bond nature and bond forming kinetics Controversies remain regarding the nature of the oxide bond and the long-range ordered O–M(metal)–O chains that appear on the oxygen chemisorbed surfaces. The O–M–O chain formation was believed to provide the major forces that stabilize the reconstructed surface. For instance, one opinion [32] suggests that oxygen bonds more covalently to a Cu atom than to a Ni, and with very small 3d-electron participation in the bonding between oxygen and copper. An alternative opinion [33] is that the O–Cu bond is an ionic one with a significant Cu-3d electron contribution. A co-linear O–Cu–O chain model [34,35] suggests that the chain is linked through the O(2pxy)–Cu(3dx2
y2 ) interaction; in comparison with this idea, it is suggested [36] that the O–Cu–O chain is connected by the delocalized O–Cu anti-bonding states. The latest investigations [37–40] suggest that the O–M bond have a mixture of ionic– covalent character and that the O–M bond transforms from ionic/covalent to covalent/ ionic p inpnature when the Cu(001)-c(22)-2O phase transforms into the Cu(001)( 22 2)R45 -2O structure (Section 3.2). The covalent bond character is found to be weaker in the case of Cu(001)-c(22)-2O than the case of Ni(001)-c(22)-2O [39,40]. Through a study of dc resistance and infrared reflectance changes induced in epitaxial Cu(100) films by adsorbed oxygen, McCullen et al. [41] found that the standard surface resistivity models based on free electrons and point scatters are inadequate, even if the adsorbate-induced changes in conduction electron density are considered. They found that interpreting their findings within a free electron model would require that each adsorbate localize an unreasonably large number of conduction electrons. However, it is yet to be known how the electron transports among the bonding constituent atoms or how the adsorbate localizes the electrons or even what the exact nature of the interaction between the oxygen and the Cu is. Oxidation is actually a kinetic process of bond forming and it is difficult to determine accurately

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

529

the static position of the moving surface atoms with altered valencies and sizes. Therefore, the nature and kinetics of oxide bonding and its consequences on the behavior of atoms and valence electrons at the surface is of key importance. 1.3.2. Alternation of atomic valencies The common features of the STM images of metal surfaces with chemisorbed oxygen are the pronounced dimensions and contrasts of protrusions compared with those on clean metal surfaces. STM studies have confirmed the presence or absence of the O–M–O chains. It has been accepted that the oxygen adsorbates ‘squeeze out’ atoms at a certain number of surfaces so that metal rows are missing from these surfaces. Furthermore, the shapes of the protrusions and the orientations of the O–M–O chains vary considerably with the material andpthe ffiffiffi crystal pffiffiffi orientation. For instance, O–Cu pairing chains form on the Cu(001)-( 2  2 2)R45 -2O surface and the ‘dumb-bell’ protrusions are as high as 0.45 A˚ [36]. In contrast, zigzagged O–O chains form between two Co rows along the close-packed direction on the Co(1010)-c(42)4O surface [42]. The ‘honeycomb’ protrusions on the O-Co(1010) surface are up to 1.0 A˚ [43]. The resulting reconstructed phases of (Ag, Ni, Cu, Pt)(110)-(21)-O surfaces [9,44–46] possess ‘a high degree of similarity in the sense that they all are stabilized by single O–M–O strings perpendicular to the close packed direction’. The ‘spherical’ protrusion on the Cu(110)-(21)-O surface is about 0.8 A˚ in height contrasting to that of 0.15 A˚ for the clean Cu(110) surface [47]. The ‘oval’ protrusions are observed in the Cu(110)-c(62)-8O phase [48] and the ‘honeycomb’ protrusions composed of the ‘dumbbells’ are observed from the Cu(111)-O surfacepat higher p temperature [49]. All the (Ag, Cu, Ni, Pt)(110)-(21)-O and the Cu(001)-( 22 2)R45 -2O phases have missing rows. However, some others have no atoms that are missing during the reaction. These include Ni(001)–O [50], Pd(001)–O [51], (Co, Ru)(1010)–O [42,43], Cu(111)–O [49], Rh(001)–O [52], Rh(111)–O [53,54] and Ru(0001)–O surfaces [55,56]. STM studies should be able to reveal inherently common features caused by oxygen adsorption from all these many specific forms. The determination of the behavior of surface electrons is far beyond the scope of models in terms of rigid spheres. It has been noted [18] that the STM features for metal surfaces with chemisorbed oxygen can hardly be explained crystallographically. What needs to do first is to define correctly the correspondence between the STM and STS signatures and the valencies of surface atoms. The atomic valencies may alter from metallic to ionic, polarized, or to missing-row vacancies upon reaction. The definition of surface atomic valency may then enable the reaction of a specific surface to be formulated. The patterns of morphology and crystallography may vary from situation to situation; the oxide bond configuration and the valence DOS distribution modified by oxidation should be naturally common for all the oxide surfaces. This would eventually lead to deeper insight into the various observations of different oxidized systems for generalized information. 1.3.3. Spectroscopes correspondences Spectroscopes such as STS, PES, and TDS as well as EELS are important tools commonly used in chemisorption studies. UPS with E < 50 eV (He-I and He-II

530

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

excitation) is often used to obtain direct information about the distribution of valence electrons below the Fermi level (EF) of a specimen. XPS at higher energy (102 eV) reveals the energy shift of the core bands of the surface. The core-level shift depends on the strength of the crystal field experienced by the core electrons. Shortened bond length and alternatively charged ions will enhance the binding energy and hence the crystal field. Charge transportation alters the nature of atomic interaction and also weakens the effect of valence screening on the particular core-level states [57]. Therefore, the chemical shift in XPS reflects the occurrence of electron transport and atomic dislocation due to the oxidation. STS provides on-site DOS information around the EF level of atoms at the surface. UPS and STS also provide information about any work-function change caused by dipole layer formation at the surface [58]. For example, STS profiles from the O–Cu–O chain region on the O-Cu(110) surface have revealed two DOS features around EF [47]. One is the empty energy states located at 0.8–1.8 eV above EF and the other is the newly occupied state that is 1.4–2.1 eV below EF. The STS DOS features below EF are substantially the same as those detected using UPS from the Cu(110)–O [59], Cu(001)–O [35] and Pd(110)–O [60] surfaces. It is unlikely that oxygen adds simply its 2s and 2p states to the valence bands of the host metals without charge exchange. Therefore, the oxygen-derived DOS features need yet to be classified. On the other hand, as remarked by Redhead [61], the pioneer of TDS, it is possible to identify, from the TDS features, the individual process of bond breaking, i.e., the opposite process of bond forming. TDS profiles possess several peaks and the intensities of the peaks oscillate with increasing oxygen exposure. For example, TDS from O–Pd [62] and O–Rh [173] surfaces show a similar number of peaks (4–5) with slight difference of peak temperatures. A correspondence needs to be identified between: (i) the TDS peaks and the bond strength and, (ii) the peak intensity oscillation and bond forming kinetics. High-resolution EELS from O–Ru [63,64] and O–Rh surfaces showed that the stretch modes of dipole vibration were around  0.05 eV energy. The peak shifts towards higher binding energy when oxygen coverage increases. However, the nature of the weak interaction and the origin for the peak shift are yet to be defined. It is interesting to note that the value 0.05 eV is at the same energy level as that for the hydrogen bond vibration detected from protein, H2O and DNA molecules. Clear and consistent definition of the outstanding features of STS, UPS, TDS and EELS is essential. These spectral features should correspond to the bond forming kinetics, bond strength and the oxygen-derived valence DOS features. 1.3.4. Driving forces behind reconstruction Questions still remain such as what mechanism generates a force which is so strong that it enables the oxygen adsorbates to ‘push’ or ‘pull’ the entire first atomic layer outward by 8–30%. Where do the forces come from that ‘push’ the second and the third atomic layers closer by  5%? For a pure-metal surface the first interlayer spacing often contracts by 3–30%, instead [65–72]. It is hard to imagine that the oxygen adsorbates resting above the surface are able to ‘pull out’ the entire first atomic layer without external forces ‘pulling’ the adsorbates. It is not clear yet how

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

531

the adsorbates remove metal atoms from the surface to form the missing rows. It is not certain yet how the oxygen atoms weaken the interaction between the top and the second atomic layer, and how the oxygen adsorbates enhance the interaction between the second and the third substrate atomic layer. Theoretical studies [39,40] suggest that the long range Coulomb interaction between the O overlayer and the metal surface provides the driving force for the p p phase reconstruction from the Cu(001)- c(22)-2O phase to the ( 22 2)R45 -2O. The long-range electrostatic interaction that increases with reducing the O–Cu layer separation, also controls the charge transfer and chemical binding in the system [73]. The O valence charge density is found to be anisotropic and non-monotonically dependent on the separation between the O overlayer and the Cu surface, varying from 0.54 and 1.08 A˚. Based on the effective-medium theory, Jacobsen and Nørskov [3] assumed that oxygen atoms penetrate into the Cu(001) and the Cu(110) surfaces and push the top Cu layer outward. Such a subsurface-oxygen structural configuration on the Cu(110) surface agrees with the conclusion drawn by Feidenhans’l et al. [74] from their surface XRD studies. A first-principles study of the O–Al(111) surface by Kiejna and Lundqvist [75,76] recently revealed that the oxygen adsorbate prefers the hcp tetrahedral site, 1.92 A˚ below the topmost Al layer that has relaxed by 25–37%. For the simultaneous sub-surface and on-surface adsorption at Y=1.0, the binding energy in the hcp-hollow sub-surface site is 0.2 eV/atom lower than the binding energy in the on-surface fcc-hollow sites. The hcp-hollow subsurface oxygen is apparently favourable in the Al(111) surface due to the lower binding energy. These sub-surface-oxygen hypotheses should be reasonably true, as the oxygen adsorbate needs to move into the surface to form bonds with its surrounding atoms in both the top first and the second atomic layers and then penetrate into the bulk, preceding the oxygen-attacked corrosion. The oxygen adsorbates push up the entire top layer and squeeze some metal atoms away from their original sites, as a result of bond formation. The interaction between metal ions in the second layer with metal atoms in the third layer should be stronger than the original pure metallic interaction, which may drive the second and the third atomic planes come closer. The driving force for reconstruction were attributed to the formation of the O–M– O chains and the formation of the missing rows at the surface. However, neither missing rows nor O–M–O strings form on the (Co, Ru)(1010)–O, Rh-(111)–O, Ru(0001)–O and Rh(001)–O surfaces. Therefore, formation of the missing row and the O–M–O chain may not be an essential mechanism driving the reconstruction. Jacobsen and Nørskov [3] related the driving force to a ‘stronger O–metal bond’ formation on the reconstructed surface, because they noticed that the oxygen 2p states hybridise (bond) more strongly with the d states of metal atoms. Further, they noted that the O–M bond becomes stronger if the oxygen bonds with metal atoms of lower CN, though the bond contracts insignificantly at a curved surface of a nanosolid according to their EMT calculations [77]. These ideas provide highly possible mechanisms for the forces that drive the reconstruction. Further correlations between the binding energy (driving force) and the bond nature and the extent of bond contraction are still needed.

532

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

1.3.5. Work function and inner potential change The work function of a surface is the separation in energy between EF and the vacuum level. The work function often changes when dipoles form at the surface. If the negative end of the dipole is directed into the vacuum the work function reduces, and vice versa. The muffin-tin inner potential constant corresponds to the net quantity of electrons around the atom [78]. For metal surfaces with chemisorbed oxygen, the work function often reduces by about 1.2 eV and such a reduction depends on the phases and oxygen exposure [23,79]. Zhang et al. [80] detected that the work function of a Gd(0001) surface reduces from 3.3 to 2.5 eV upon oxygen chemisorption. They noted that the work function initially changes quickly with exposure but becomes slower at overages over 0.5 L. Occasionally, the work function of a surface is observed to increase at higher oxygen exposures. The inner potential constants for the Cu [81] and the Ru [82] surfaces were found to reduce upon oxygen adsorption. Pfnu¨r et al. [83] found it necessary to assume such a reduction in analyzing the VLEED (very-low-energy electron diffraction) spectra from the O–Ru system. The VLEED calculations with rigid-sphere models by Thurgate and Sun [81] showed a 1.2 eV or higher reduction of the inner potential for the top layer of the O–Cu(001) surface. The inner potential reduction also varies for different phases due to the different possible crystal structures [84]. VLEED optimization with the current bond model revealed that the inner potential for the top Cu atomic layer reduces by 9.5% (from 11.56 to 10.5 eV) upon oxygen chemisorption [78]. Besides the strong localization of surface charges due to the missing-row formation and charge transportation, the residual ion cores provide an additional likely mechanism for reducing the inner potential. Quantification and explanation of the reduction of both the work function and the inner potential constant for the O-chemisorbed surfaces are also big challenges. 1.3.6. Factors controlling bond formation Models of rigid spheres for a specific reconstructed system can illustrate the static atomic positions at a certain moment of snapshot. However, such models reveal little information about the kinetics and dynamics of atoms and electrons at the surface. During the reaction, atomic position and atomic size change; atomic valencies and the form of atomic interaction change; electrons are strongly localized by transporting from one specimen to the other, and from one energy level to another; some occupied energy states in the valence band are emptied and some empty ones are filled up. It seems impractical to locate accurately the static positions of the individual atoms at a surface. It is not realistic to base all observations on atomic dislocation or crystal-structure change. As pointed out by King [22], the task in the future decades is to grips with the factors controlling bond forming and breaking. It is worth noting that, from an experimental and theoretical point of view, bond-forming kinetics and dynamics are beyond the scope of currently available instrumentation and theoretical approximations. For example, results of numerical optimizations are subject to the assumptions made or to the initial conditions taken, [55,85,86] and simulation of diffraction data often involves a huge number of strongly correlated parameters [87]. The independent treatment of the correlated

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

533

parameters often leads to an infinite number of numerical solutions [88–90]. In addition, theoretical calculations often consider the electrostatic interaction between the adsorbates and host atoms, being treated as rigid spheres, rather than the true bond formation [91], with essential contraction of bonds at the surface. Therefore, physical constraints on all the observations and their interdependence are necessary. It is quite often that some undetectable factors play the dominant roles behind the observations. It would be interesting and rewarding to cope with the above-mentioned challenges, and thus perhaps to discover methods to control the processes of bond making or breaking. 1.4. Objectives The aforementioned long-lasting challenges and the availability of the advanced VLEED methods (the data and the calculation code) at Murdoch University, Australia, and some outstanding STM/S observations [47,49] brought the present practitioner into this field in 1992. At the starting point of time, the crystallography, electronic spectroscopy, and the surface morphology were treated independently as convention and the hard sphere models with electrostatic interaction were dominant in chemisorption studies. In the trial-error VLEED calculations, numerous (around 30) independent parameters were needed to be dealt with. Therefore, an attempt allowing the sp orbitals of an oxygen atom to hybridize when the oxygen interacts with a solid surface and an alternative way of modeling considerations and decoding techniques were timely necessary. Based on decoding the kinetic VLEED data from the O–Cu(001) surface, a compact model has been developed for the oxide tetrahedron bonding [92,93] and its subsequent effects on the valence DOS [94,95] and the SPB of surfaces with chemisorbed oxygen [88]. The developed BBB correlation has enabled in turn the capacity and reliability of VLEED to be fully explored [96], and the outstanding STM images (see Fig. 4 in Section 3) to be explained in terms of bond geometry and atomic valencies. The corresponding decoding technique and the BBB models have enabled the kinetic VLEED from the O–Cu(001) surface to be quantified and consistently understood in terms of four-stage Cu3O2 bonding kinetics and its effects on the valence DOS [97]. Hence, the reliability of both the BBB theory and the advanced VLEED technique has been justified. For more details about the VLEED quantification of the O–Cu(001) bond forming kinetics the reader may be referred to recent reports [98,99]. For the purpose of completeness, we need to highlight some key points in the present report. With respect to the literature documented and previous reviews of this practitioner [98–100] the current report focus more on extending the BBB correlation mechanism to the electronic process of surface oxidation of metals and thermal oxidation of nonmetallic diamond, and enhancing the capacity of STM, STS, PES, TDS and EELS for general understanding. This has led to some designer process and materials with desired functions. The main objectives of this report are to share with the community what the practitioner experienced and learnt in the past decade as the following:

534

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

(i) Surface oxidation is a kinetic process in which charge transportation/polarization dominates and atomic dislocation is merely one of the consequences. The events of O–M bonding, oxygen lone pair non-bonding, metal dipole anti-bonding, and H-like bonding are crucial to the process of oxidation. Meanwhile, atomic sizes and atomic valencies change, and the bonds at the surface contract. These events dislocate surface atoms collectively and modify the valence DOS of the host. These events also roughen the surface (or SPB) and change the physical properties of an oxide surface. (ii) The electronegativity, the scale of lattice constant and the geometrical orientation of the surface determine the specific details of the adsorbate site, bond ordering and the orientation of the oxide tetrahedron. This gives rise to the versatile modes of crystal reconstruction and surface morphology change. For the analyzed representatives of transition metals, noble metals and a nonmetallic diamond it is consistently concluded that the phase ordering, crystallography and surface morphology vary from situation to situation. However, formation of the basic oxide tetrahedron, with sp-orbital hybridization and lone pair production, oxygen derived valence DOS features and the kinetic processes of bond formation are all the same by nature. (iii) The combination of STM, LEED/VLEED, PES/STS, TDS and EEELS/ Raman are essential for a comprehensive insight into the electronic process occurring at a surface. Furnished with the new BBB correlation premise, these techniques allow one to extract information about the surface atomic valencies, bond geometry, valence DOS, bond strength and bond forming kinetics. (iv) The tetrahedron bond model can be applied to reactions involving carbon and nitrogen. The concepts of bond contraction and band-gap expansion have been extended to some practical applications, which has led to some innovative findings in technical applications, as will be discussed in the context.

2. Principle: bond–band–barrier (BBB) correlation 2.1. Foundations 2.1.1. Basic concepts As will be demonstrated, patterns of observations of the oxidized surfaces depend on the scale and geometry of the surface lattice, and the difference in electronegativity between the bonding constituents. Therefore, it is necessary to classify these basics first. Fig. 1 illustrates the typical coordination environment of the low-index fcc and hcp surfaces. Host atoms are arranged at the first two planes of the fcc{(001), (110), (111)} and the hcp{(1010), (0001)} surfaces in the regular lattice sites. The C4v, C3v and C2v point-group symmetries can be applied to the unit cells. The shortest atomic

535

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 1. Possible coordination environment for the oxide bond formation. (a) Atoms surrounding the fcc(001) fourfold (C4v) hollow-site form an upside-down pyramid. (b) On the fcc(111) and hcp(0001) surfaces, there are two types of threefold (C3v) hollow sites. Atoms surrounding the hcp(0001) hollow (I) form a tetrahedron. No atom exists in the substrate second layer below the fcc(111) hollow site (II). (c) The fcc(110) and its analogue (d) hcp(1010) surfaces possess alternate hcp(0001) (I) and fcc(111) (II) facet sites along the close packed direction.

separation (atomic diameter) is a. These structures represent majority of coordination environments so far documented for surface oxidation studies. Table 1 compares the lattice geometry of the unit cells. In the fcc(001) surface unit cell [see Fig. 1(a)], five atoms surrounding the C4v hollow site form an upside-down pyramid. The atomic structure of the fcc(111) and the hcp(0001) surfaces in Fig. 1(b) are the same in the top two atomic planes where atoms arrange in the same AB order. Table 1 Comparison of the lattice geometry of the unit cells of various surfaces (unit in atomic diameter, a) a1 fcc(001) fcc(110) hcp(1010) fcc(111) hcp(0001)

1 1 1 1 1

a2 p

1 2 1.747 1 1

a3 (layer spacing) p 1/ 2 1/2 0.2887 0.6934 0.8735

536

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Atoms surrounding the hcp(0001) hollow (indicated I) site form a tetrahedron while atoms surround the fcc(111) hollow (indicated II) site cannot because there is no atom in the second layer. Atoms surrounding the fcc(110) and the hcp(1010) hollow sites [in Fig. 1(c and d)], form a rectangular-pyramid of C2v symmetry. Besides the long-bridge hollow site, there are two facet sites along the close packed direction in the fcc(110) and the hcp(1010) surfaces. One is the hcp(0001) facet hollow site (I) that involves one atom in the top layer and two atoms in the second layer; the other is the fcc(111) facet (labeled II) that contains two atoms in the top layer and one in the second layer along the close packed direction. The fcc{(110), (111)} surfaces are analogous to the hcp{(1010), (0001)} surfaces with a slight difference in the interatomic spacing. Table 2 lists the values of electronegativity (
), possible valencies and the atomic radius of representative elements of different electronic structures. The difference in electronegativity between atoms of two elements determines the nature of the bond between them. If the
is sufficiently high (around 2), the bond is ionic, otherwise it is covalent or polar-covalent [102]. Normally, the atomic size of a noble (4d) metal is greater than that of a transition (3d) metal and the electronegativity of the noble metals is higher than that of transition metals. It is noted that an atomic radius is not a constant but varies with the coordination number of this atom. Importantly, atomic radii change with alteration of valencies. It will be shown that, these basics play important roles in specifying the site of the adsorbate and the orientation of the basic oxide tetrahedron and hence the patterns of observations for the chemisorbed surfaces. 2.1.2. Bonding effects Bond formation is a process in which valence electrons transport. This should have enormous effects on the surroundings by polarization and mass transportation. Alteration of atomic sizes will change the atomic distances and modify the surface morphology. Besides the well known bonding states of metallic, covalent, ionic and Van der Waals bonds in nature, polar-covalent bonds, non-bonding lone pairs, antibonding dipoles, H-like bonds and hydrocarbon-like bonds also exist.

Table 2 Electronegativity, possible valencies and the CN-related atomic radius of typical elements after Goldschmidt [101] and Pauling [102] Element

C

N

O

Si

Co

Cu

Electronic structure

2s2p2

2s2p3

2s2p4

3s2p2

3d74s2 3d104s1 4d105s1 4d75s1 4d85s1 4d105s0 3d34s2


Rion (Valency) Rm (CN=1) Rm (CN=12)

2.5 2.6 (
4) 0.771 0.914

3.0 1.71 (
3) 0.70/0.74 0.88/0.92

3.5 1.32 (
2) 0.66/0.74 -

1.9 1.9 0.41 (4) 0.82 (2) 1.173 1.157 1.316 1.252

1.9 0.53 (1) 1.173 1.276

Ag

1.9 1.00 (1) 1.339 1.442

Ru

2.2 – 1.241 1.336

Rh

2.2 – 1.252 1.342

Pd

2.2 – 1.283 1.373

V

1.6 – 1.224 1.338

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

537

Despite the well known bonding events illustrated in Fig. 2(a and b), Fig. 2(c–h) describes the formation of an ionic bond, non-bonding lone pairs and their consequences on the wave functions of their atomic neighbors. The electronegative adsorbate or additive (smaller broken circle labeled A) interacts with the heavier host atoms (bigger broken circle labeled B) by either capturing electrons from the host B atom or polarizing the electrons of B. The energy of the polarized electrons will rise to higher energy levels because the polarization provides additional energy to the polarized electrons. Electron transport alters the atomic valencies and atomic sizes of both the adsorbate A and the host B. For example, an oxygen atom changes its radius from 0.66 to 1.32 A˚ when the oxygen atom evolves into an O
2 ion. A copper atom alters its radius from 1.278 to 0.53 A˚ when the Cu atom becomes a Cu+ ion. All the ions, whether positive or negative, and the non-bonding lone pairs are apt to polarizing their neighbors giving rise to the host dipoles. Dipoles are formed with

Fig. 2. Schematic illustration of the possible bond configurations and their consequences on the electron clouds of surrounding atoms (shaded areas stand for dipoles). (a) and (b) are the well-known bonding events. (c) Ionic bond formation alters atomic sizes (broken circles) and valencies. (d) Non-bonding lone pair formation (represented by ‘:’) induces Bdipole. (e) H-like bond forms if B+/dipole replaces the H+/dipole. (f) O–M bonds involve non-bonding lone pairs, bonding electron pairs and (g) anti-bonding dipoles. (h) Hydrocarbon-like bonds can form by replacing the H+ with B+, which also induces anti-bonding dipoles.

538

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

expansion of atomic sizes and elevation of the DOS in energy space. The production of the dipoles and the dipole–dipole interaction in the opposite direction will raise the system energy. It is therefore reasonable to term such an event as anti-bonding dipole formation—an extreme case of the Van der Waals bond interaction. Antibonding is a by-product of reaction and it never forms between atoms of different electronegativity [108]. Non-bonding lone pairs form when a pair of electrons of a specific atom occupies a directional bonding orbital. This happens to electronegative elements in the upperright part of the periodic table, such as nitrogen, oxygen and fluorine when the 2s, 2px, 2py 2pz orbitals of these elements are hybridised [103]. It is often the case that part of the hybridized orbitals are occupied by shared electron pair (bonding) between A and B and the remaining orbitals by the lone electron pairs (non-bonding) of the electronegative additives. The number of lone pairs of an adsorbate follows a ‘4
n’ rule and the n is the valence value of the adsorbate. For oxygen (n=2), two lone pairs are present while for nitrogen (n=3) only one lone pair forms during the sp-orbital hybridization. The ‘4
n’ rule holds for any elements in which the sp orbitals hybridize. The lone pair requires an interaction with a B atom through polarization without any charge transport. The lone pair is actually not a bond but the weaker part of the hydrogen bond. The classical hydrogen bond (O
2–H+/dipole : O
2), known for over 50 years, plays an essential role in the structure and function of biological molecules. The ‘–’ and ‘:’ represent the bond and the lone pair, respectively. Hydrogen bonds are responsible for the strength and elasticity of materials, such as wood or a spider’s web, molecular binding, as well as base pairing and folding in DNA. Hydrogen bonds are also responsible for the synthesis and transferring of protein signaling [104,105]. It is to be noted that the formation of the hydrogen bond is not due to the existence of atoms of hydrogen or oxygen but due to the existence of the non-bonding lone pairs. If the lone-pair-induced Bdipole bonds further to an electronegative element A, then an H-like bond (O
2–B+/dipole :O
2) forms. H-like bonding differs from the classical hydrogen bond simply in that, the B+/dipole replaces the H+/dipole in the hydrogen bond (see Fig. 2e). If an atom of another electronegative element, such as C, replaces one of the oxygen ions then the (C-4–B+/dipole : O
2) configuration forms, which was specified in some cases as the anti-hydrogen bond [106]. This is also a H-like bond. Formation of such a H-like bond depends merely on the existence of the lone pair rather than the particular elements involved. Hence, the H-like bond is more generally applicable though it is not often referred to as such. The same is true for the hydrocarbon-like bonds. The hydrocarbon bond is polar covalent in nature. The naked H+ also polarizes and attracts electrons of its neighboring atoms. Hydrocarbon-like bond can form by replacing the H+ with B+. The B+ is less electronegative than the carbon. Unfortunately, the production of nonbonding lone pairs, anti-bonding dipoles, H-like bonds and the hydrocarbon-like bonds are often overlooked. However, these events indeed play crucial roles in determining the physical properties of a system that involves electronegative additives. Quite often, a system contains several kinds of chemical bonds, such as in graphite and in an oxide. Because of the sp2-orbital

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

539

hybridization of carbon, the Van der Waals bond dominates in the [0001] direction while the stronger covalent bond dominates in the (0001) plane of the graphite. As can be seen from Fig. 2(f and g), O–B bond formation involves sharing pairs of electrons (bond), non-bonding lone pairs and anti-bonding dipoles. The electronic environment surrounding an oxygen atom varies from site to site at the subatomic scale. From an energy point of view, bond formation lowers the system energy and stabilizes the system. Anti-bond (dipole) formation requires additional energy. Although it is energetically less favorable, the anti-bond can still form as a by-product of the events of bonding and non-bonding. Occupation of the orbitals by nonbonding electron lone pairs of an electronegative element, in principle, neither raises nor lowers the system energy with respect to the energy level of the isolated atoms of the electronegative element [107,108]. From the band structure point of view, the anti-bond derived DOS (or polaron) should locate at energy above EF or near to it due to the energy rise of the polarized electrons. The DOS features for bonding are located below the originally occupied levels of the electronegative element; while the DOS features of non-bonding lone pairs are located between that of the bond and that of the anti-bond. Hydrogen-like bond formation will stabilize the system as electrons transport from the high-energy anti-bonding states to the lower bonding states. Bond and anti-bond formation will produce holes below the EF of the host material [93], which should be responsible for the transition from metal to semiconductor when a compound forms. 2.2. Chemical bond: the basic tetrahedron The original idea of the model is to extend the H2O molecular structure to a solid surface with chemisorbed oxygen by replacing the H atom with a host atom of an arbitrary element B, as illustrated in Fig. 3(a). Two factors are taken into account in the modeling considerations. First, the atomic radius is not constant but varies with changes in not only its atomic valency, but also, its CN. Second, the sp orbitals of oxygen hybridize and a quasi-tetrahedron forms. The bond angles and the bond lengths are not constant but vary within limits. Therefore, an oxygen atom can react with atoms, in any gaseous, liquid or solid states of an arbitrary element B through two bonding electron pairs and two non-bonding lone pairs. Besides the well-known fact that an atom changes its radius when its valency alternates, both the ionic and metallic radii of an atom contract with reducing the CN of this atom. Goldschmidt [101] suggested that, if an atom changes its CN from 12 to 8, 6 and 4, then the ionic radius would be reduced by 3, 4 and 12% correspondingly. Pauling [102] also noted that the metallic radius contracts considerably with reduction of the CN of the metal atom (see examples in Table 2). One may extend the CN-imperfection induced radius contraction to atoms at a solid surface or sites surrounding defects (such as point defects and stacking errors) though no such attempt has been reported previously. It is understandable that the surface provides an ideal environment for CN reduction. Termination of the lattice periodicity in the surface normal direction reduces the CN of an atom at the surface. Such a CN-reduction

540

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 3. (a) The primary oxide quasi-tetrahedron and (b) the corresponding DOS features of bonding, nonbonding, anti-bonding and holes [94]. Each of the two ions, 1 and 2, donates one electron to the central oxygen to form the Goldschmidt-contraction ionic bonds. Atoms labeled 3 are the lone-pair-induced metal dipoles with expansion of sizes and elevation of energy states. Arrows represent the process of charge transportation. The arrow from the anti-bonding sub-band to the bond states corresponds to the process of H-like bond formation.

shortens the remaining bonds of the surface atom. It is reasonable to consider the CN-effect on the atomic radius as a Goldschmidt-contraction for an ionic bond or a Pauling-contraction for a metallic. There exists sufficient evidence for the bond contraction at metal surfaces [65–72]. For instance, about a 10% reduction of the first layer spacing of the (Ru [109], Co [110] and Re [111])(1010) surface has been detected using LEED measurements and DFT calculations. The interlayer distance between the first and second layer of the diamond (111) surface is  30% smaller than the interlayer separation in the bulk, which leads to a substantial reduction of the surface energy [112]. However, it has been reported that the dimer bond lengths for the IIA (Be and Mg(0001) surface) and IIB (Zn, Cd, and Hg) elements are longer than the corresponding nearest neighbor atomic separation of the bulk values [72], which is conflicting with Pauling’s premise [102]. Table 3 summarizes the surface interlayer relaxations of some metals caused by the Pauling-contraction. Oxygen interacts with atoms of element B [Fig. 3(a)] and hybridizes its sp orbitals to form four directional orbitals. Oxygen captures two electrons from B atoms and the 2s and 2p levels of oxygen are fully occupied with eight electrons that will repopulate in the four directional orbitals. Therefore, two of the four hybridized orbitals are occupied by shared electron pairs (bonding orbitals). The remaining two orbitals are occupied by the lone electron pairs of oxygen (non-bonding orbitals). It may be necessary to point out that the sp orbitals of oxygen hybridize independently without the involvement of orbitals of atoms of other elements, but the hybrid orbitals may be occupied by electrons of the others. Therefore, the orbital to be occupied is one thing; the actual occupancy of the hybridized orbital is another. The orbital can

541

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Table 3 Summary of the observed surface interlayer relaxation of clean metal surfaces. Origin for the conflicting relaxation of the Be(0001) surfaces is yet not clear [72] Metal

Method

d12/d12

Metal

Rh(001) W(001) W(110) W(320) Al(001) Al(210) Ti/Zr(0001)

LEED [113,114] DFT [116] LEED [114,117] DFT [118] DFT [68] LEED [122] DFT [72,123] LEED [114] (110)[127,128] (111), LEED (100) [67] DFT [393,133] LEED [134]


1.2;
1.4
5.7
3.0
22.3
10
16
6.1
7.8 -4.9
6
9
1
2 [131]
2
3
10.0;
13.0 +5.8

Fe(210) Fe(310) Pd(310) Pt(210) Cu(331) Cu(551) Cu(211)

Ag/Cu/Ni Ag/Cu/Ni Ag/Cu/Ni Fe/W(110) Be(0001) a

Method

LEED [115] LEED [115] DFT [71] LEED/EAM [119] DFT [125,120] RSGF [121] RSGFa [121] DFT [124,125,121] LEED [126] Cu(117) LEED [129] DFT [130] Al(113) DFT [132] Al(115) DFT [132] Al(117) DFT [132] Al(331) LEED [135]

d12/d12 (%)
22
16
14.1
23.0/
31;
22.0;
13.8;
10.4
9.8
14.4;
28.4;-10.8
14.9
13.0;-9.5
6.8
8.0
8.3
11.7

RSGF: Real-space Green’s function method. d12 is the separation between the top two atomic layers.

be occupied by any kind of electron pairs (sharing or non-sharing). In a bonding orbital, the extent of electron sharing, or the nature of the bond, depends on the difference in electronegativity (
) between the oxygen and element B. Due to the high
value (see Table 2), oxygen catches an electron from B (labeled 1 and 2) to form the Goldschmidt-contraction ionic bond at the surface. Formation of the nonbonding lone pairs, however, is independent of the nature of element B. The lone pairs are likely to polarize atom B (labeled 3) and the B atom becomes a Bdipole with associated expansion of size and elevation of energy of the polarized electrons that occupy the anti-bonding energy levels. In an oxide tetrahedron, the plane (3O3) composed of the lone pairs and the oxygen nucleus should be ideally perpendicular to the plane (1O2) that consists of the two bonding orbitals. The distance (1–2) between the two B+ ions and the spacing (3–3) between the Bdipole and Bdipole match closely the first and second shortest atomic spacing at a surface, which involves two atomic layers. The Bdipole tends to locate at the open end of a surface due to the strong repulsion between the dipoles. The B2O primary tetrahedron is not a standard one but it is distorted due to the following two effects: (i) The difference in repulsion between the occupied orbitals varies the bond angles [BAij (angle ffiOj), i, j=1, 2, 3 correspond to the atoms as labeled; BA124104.5 , BA33 > 109.5 ] and, (ii) the difference in CN of atoms at different sites adjusts the bond length [BLi=(RM++RO
2 )(1-Qi), i=1, 2; Qi are the effective bond contracting factors]. The length of BL3 and the angle BA33 may vary with the coordination circumstances in a real system. It is unavoidable that the oxide tetrahedron formation dislocates the B atoms collectively in the otherwise regular lattice sites. Moreover, oxygen always seeks four neighbors to form a stable quasi tetrahedron. On the other hand, the expansion of atomic radius and the energy rise of the dipole electrons are responsible for the protrusions in the STM images and the reduction of the local work function. The

542

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

localized dipole electrons are also responsible for the non-Ohmic rectification at the surface, even though the local work function reduces significantly. The strong localization of dipole electrons at the surface increases the surface contact resistance because these electrons cannot move easily. The consequences of dipole formation in this model agree with Lang’s theory [136–138] that an oxygen adsorbate affects the STM current predominantly by polarizing metal electrons, as a result of anti-bonding dipole formation. At the initial stage of oxidation, the oxygen molecule dissociates and the oxygen atom interacts with the host atoms through a single bond. It will be shown later (Section 3) that the O
1 occupies a specific position where the O
1 bonds directly to one of its neighbors and polarizes the rest. For the transition metals, such as Cu and Co, of lower electronegativity (
< 2) and smaller atomic radius (< 1.3 A˚), oxygen often bonds to an atom at the surface first. For noble metals, such as Ru and Rh, of higher electronegativity (
> 2) and larger atomic radius ( > 1.3 A˚), oxygen tends to sink into the hollow site and bonds to the atom underneath first. The ordering of bond formation leads to different patterns of reconstruction. The O
1 also polarizes other neighbors and pushes the Bdipole at the surface radially outward from the adsorbate. Because of oxide tetrahedron formation with lone pair non-bonding and dipole anti-bonding, the electronic structure surrounding a certain atom varies from site to site. 2.3. Valence density-of-state (DOS) The formation of bonds, non-bonding lone pairs and anti-bonding dipoles as well as the H-like bonds generates corresponding features adding to the DOS of the valence band and above of the host, as illustrated in Fig. 3(b). Arrows represent the kinetic processes of electron transportation. Initially, energy states below the EF of a metal are fully occupied in the ideal case at T=0. The work function, 0, Fermi energy, EF, and the vacuum level, E0, follow the simple relation: E0=0+EF. For Cu, as an example, E0=12.04 eV, 0=5.0 eV and EF=7.04 eV. The Cu-3d band locates at energies range over from
2.0 to
5.0 eV below EF. The oxygen 2p states are around
5.5 eV with respect to EF for Cu. At the initial stage of reaction, an electron from a metal is transported from its outermost shell to the unoccupied 2p orbital of the oxygen, which produces a hole in the outermost shell of the metal. The O
1 polarizes its rest neighbors to form a polaron, as a result. This simple process creates additional DOS features of bonding ( < < EF), holes (4EF) and anti-bonding dipoles (4EF). With the full occupancy of the p-orbital of oxygen, the sp orbitals of the O
2 hybridize, which brings about four additional DOS features, as illustrated in Fig. 3(b):  Electronic vacancies are produced right below EF, generating a gap between the conduction band and the valence band of a metal. The electron transportation can also expand the original band-gap of a semiconductor from EG0 to EG1.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

543

 The non-bonding (lone pair) states of O
2 locate below EF without apparent energy change, in principle, compared to the 2p-level of an isolated atom of oxygen [107].  The bonding states are close to the originally occupied 2p-level of the isolated oxygen.  The anti-bonding (lone-pair-induced dipole) states are located above EF or near to it. The oxygen-induced dipole reduces the work function from 0 to  1.  Upon being overdosed with oxygen, H-like bonds form at the surface. The overdosed oxygen gets electrons from the dipoles and the Bdipole becomes B+/dipole. The arrow from the anti-bonding states above EF to the deeper bonding sub-band represents the process of H-like bond formation. Apparently, this process lowers the system energy. It is noted that the hole-production and the lone-pair production are independent but simultaneous, which result in the joint DOS features below EF. If the products of both processes are compatible in quantity, the joint DOS features derived by the two processes may not be easily separated. The hole-production is due to two mechanisms: bonding and anti-bonding. For the Cu example, the 4s electrons (in the conduction band, CB) either contribute to oxygen for the bonding or jump up to the outer emptyshell (Cu 4p for example) for the anti-bonding dipole. Such bonding and antibonding processes empty the states just below EF, which result in the Cu-oxide being a semiconductor with a known band-gap ranging from 1.2 to 1.5 eV [139,140]. STS and VLEED revealed that the states of anti-bonding of the O–Cu system range over 1.3 0.5 eV above the EF and the non-bonding states
2.1 0.7 eV below. Angular-resolved inverse PES [141] detected that the features of empty states at +2.0 eV decrease with increasing oxygen coverage on the Cu(110) surface. The PEEM studies of O–Pt surfaces [79,142–144] have detected the conversion of the dark islands, in the scale of 102 mm, into very bright ones with work functions  1.2 eV lower than that of the clean Pt surface. As will be shown in Section 4, the bonding states are around
5.5 eV below EF which is shifted slightly towards an energy lower than the 2p-level of the oxygen because the hybrid bond forming lowers the system energy. Most strikingly, all the oxygen-derived DOS features are strongly localized in real space. 2.4. Surface potential barrier (SPB) 2.4.1. One-dimensional SPB model The SPB experienced by electrons traversing the surface region contains two parts [145]: Vðr; EÞ ¼ ReVðrÞ þ iImVðr; EÞ ¼ ReVðrÞ þ iIm½VðrÞ  VðEÞ

ð2:4:1Þ

The real part, ReV(r), describes the elastic scattering of the incident electron beam. Integration of the ReV(r) along the moving path of the electron beam determines the phase-shift of the electron beam.

544

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

It is adequate to consider the surface normal direction and the Re(r) will be ReV(z) which has the form [146]:   ReVðzÞ ¼f
V0 = 1 þ Aexp½
Bðz
z0 Þ ; z 5 z0 ða pseudo-Fermi-z functionÞ   1-exp½lðz
z0 Þ =½4ðz
z0 Þ; z < z0 ðthe classical image potentialÞ ð2:4:2Þ where A and B are constants given by B=V0/A and A=
1+4V0/l. The z-axis is directed into the crystal. V0 is the muffin-tin inner potential constant of the crystal and z0 is the origin of the image plane. l describes the degree of saturation. The imaginary part, ImV(r), describes the spatial decay of the incident beams. ImV(E) represents the joint effects of all the dissipative processes including excitation of phonons, photons and single-electron as well as plasmon excitation. Plasmon excitation occurs at energy much higher than EF (normally  15 eV above EF). At very low energy, plasmon excitation does not come into play. Excitation of phonon and photon requires energy smaller than the work function. Single-electron excitation occurs at any beam-energy that is greater than the work function and in the space occupied by electrons. The spatial distribution of electrons is described by (r) (charge density) which relates to the inelastic damping potential, ImV(r). Spatial integration of the ImV(r, E) determines the amplitude-loss of the scattered electron waves. An ImV(z, E) can be defined to include the effects that damping occurs in the electron-occupied space (Fermi z decay) and that damping takes place at incident beam energy being greater than the work function, which depends on the occupied DOS [88]: ImVðz; EÞ ¼ Im½VðzÞ  VðEÞ ImVðE; zÞ ¼  ðzÞ  exp½½E
L ðEÞ=       E
L ðEÞ z
z1 ðz0 Þ = 1 þ exp
¼  exp

ðz0 Þ

ð2:4:3Þ

and, L ðx; yÞ ¼ E0  EF ; and EF / n

ð2:4:4Þ

where and are constants depending on the calibration of the measured spectral intensities. The terms z1(z0) ((z1)=0.5bulk) and (z0) (saturation degree) involved in the Fermi z function describe the spatial distribution of electrons contributing to the damping of incident beams. The spatial integration of (z) from a position inside the crystal to infinitely far away from the surface gives the local DOS (n(x, y)). Therefore, L(E) can extend to cover situations that are E dependent and to large surface areas over which the LEED method integrates.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

545

2.4.2. 3-D effect with DOS contribution The ReV(r) correlates with the ImV(r) through the Poisson equation [147]: r2 ½ReVðrÞ ¼
ðrÞ; and; ImVðrÞ / ðrÞ:

ð2:4:5Þ

The gradient of the ReV(r) relates to the intensity of the electric field "(r): ![ReV(r)]=
"(r). If (r)=0, then the ReV(r) corresponds to a conservative field in which the moving electrons will suffer no energy loss and the spatial variation of the inelastic potential ImV(r) / (r)=0. It is to be noted that the ReV(z) transforms at z=z0 from the pseudo-Fermi z function to the 1/(z
z0) dominated classical image potential. Therefore, r2 ½ReVðz0 Þ ¼
ðz0 Þ ¼ 0:

ð2:4:6Þ

The origin of the image-plane, z0, acts as the boundary of the surface region occupied by electrons. If we permit z0 to vary with the surface coordinates, then the z0(x, y) provides a contour of the spatial electron distribution, which should be similar to that plotted using STM imaging method. The SPB features are thus characterized by the z0 and this effect allows us to choose z0 as the parameter in the single-variable parameterization of the nonuniform-SPB [88]. In order to correlate the parameters that used to be treated as independent, and to ensure the uniqueness of solutions, we define the SPB parameters as functional dependents of z0. They are supposed to be correlated with z0 through a Gaussiantype function: n  2 o z1 ðz0 Þ ¼ z0  exp
ðz0
z0M Þ= 1 ; ð2:4:7Þ n  2 o ðz0 Þ ¼ 1=lðz0 Þ  exp
ðz0
z0m Þ= 2 ; n 2  o   lðz0 Þ ¼ l0M x þ ð1
xÞ  exp
ðz0
z0M Þ=lz ; ðx ¼ 0:4732Þ Constants 1 and 2 are the full width at half maximum (FWHM) of the Gaussian functions and optimized to be 0.75 and 1.50, respectively [148]. The minimum z0m is estimated to be
1.75 Bohr radii. The l0M=1.275 is the maximum of l corresponding to maximum z0M=
3.425 Bohr radii and to lz=0.8965 for the O–Cu(001) surface. These constants may vary with materials considered. 2.4.3. Physical indications Eqs. (2.4.2–2.4.4) represent not only correlation among the SPB parameters but also correlate to STM profiles. At the dipole site, z1Mz0M, l
1, while in the atomic vacancy or ion positions, z1m < < z0m due to the strong localization of electrons at the surface. The ImV(z) is much less saturated than is the ReV(z) in the site giving STM depressions. The SPB increases its degree of saturation with the

546

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

outward-shift of the image plane z0. This means that formation of metal dipoles shifts the electron clouds outward and enhances the density of the shifted clouds. Except for the inner potential constant V0, all the SPB parameters l,  and z1 become functional dependents of z0. The number of SPB variables is thus reduced from four to one. In addition, decoding VLEED using the correlation produces results of clearer physical indication. More importantly, the electronic structure described with the SPB ties in closely with the valencies and positions of atoms at the surface. In order to reflect the interdependence of the SPB and bond geometry, we have developed a decoding technique used for the particular O–Cu(001) surfaces [98,99]. In calculations, the VLEED code reads in the SPB constants and the atomic positions that were converted from the bond geometry. The code then automatically optimizes the z0 towards a duplication of the measured intensity at each point of energy. Besides the bond geometry and the SPB constants, the calculation duplicates the spectrum and produces a z0(E) profile of which the shape varies with the atomic structure employed. The z0(E) profile shows joint features of surface morphology and the valence DOS as VLEED integrates over large (mm level) areas of surface. One can judge an optimal atomic geometry by analyzing the shape of the z0(E) profile against physical constraints [98]. A combination of the parameterized SPB and the z0-scanning calculation method reconciles the bond geometry (atomic positions), surface morphology [ReV(z) and ImV(z)] and the DOS change [z0(E) and ImV(E)] as a coherent system. This original approach may reflect the real process of reaction because electrons are tied in closely with the positions and valencies of surface atoms, and the reaction is a kinetic process in which atoms move collectively. Typical results will be shown in Section 3.2.3. 2.5. Summary We may end up the modeling section with a brief summary. The H2O molecule structure and the concept of CN-imperfection induced atomic radius contraction have been extended to the oxidation of a solid surface. This leads to the current model for oxide-tetrahedron bond formation and its effects on the valence densityof-states and the surface potential barrier, as well as their interdependence:  Oxygen can interact with atoms of an arbitrary element B to form a tetrahedron with bonding and non-bonding states, as well as anti-bonding dipoles due to polarization.  Oxygen derives four additional DOS features that add to the valence band and above of the host. These features of bonding, non-bonding, anti-bonding and electron holes are strongly localized.  Oxygen induces the non-uniformity of the SPB. All the SPB parameters are correlated to the image plane (z0) which corresponds to the boundary of the surface region occupied by electrons.  The bond geometry, atomic valency, valence DOS and the SPB are interdependent. One may need to pay equal attention to these categories in dealing with the surface chemisorption.

547

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

 Some important yet often-overlooked events, such as non-bonding lone pairs, anti-bonding dipoles and the H-like bond are crucial in practice. In the next section, we report our exercises of applying the original BBB model to results drawn from a database of surface oxidation. It will be shown that the signatures of STM/S, LEED, PES and TDS as well as the EELS and Raman can be explained in terms of atomic valencies, bond geometry, valence DOS and bond strength as well as bond forming kinetics. Formation of the oxide tetrahedron and the valence DOS and the bond forming kinetics are found to provide common base to explain all the systems analyzed.

3. STM and LEED: atomic valencies and bond geometry 3.1. Phase ordering Table 4 summarizes the phase orderings occurred at the low-index O–(Cu, Rh){(001), (110), (111)} surfaces and their analogues of O–(Co, Ru){(1010), (0001)} surfaces. The O–Pd(110) surface performs in the same way as the O–Rh(110) surface. It will be shown later that the O–(V, Ag)(100) surfaces perform differently from the O–(Cu, Rh)(001) surfaces despite the same geometrical configuration. Incorporating the primary bond model to these phase structures, we can derive the formulae of Table 4 Oxygen induced phase ordering on the typical metal surfaces. The hcp(0001) and hcp(1010) faces are analogues to the fcc(111) and fcc(110) faces, respectively. Please refer to Fig. 1 for the lattice geometry Cu(fcc) fcc(001)

Refs.

Rh(fcc)
1

c(22)-2O offcentered pyramid p p ( 22 2)R45 -2O
2 missing row [21,36,38,81,94,97, 149–155]

fcc(110) or hcp(1010)

Refs.

(21)-O
2 (Ni, Ag, Pt) c(62)-8O
2 [2,16,21,34,35,9,47, 48,3,74,94,168–170]

fcc(111) or hcp(0001) ‘29’ & ‘44’ structures Refs.

[188,49]

Co(hcp)

Ru(hcp)

Disordered (21)-2O
1 (atop) c(24)-4O
2

c(24)-4O
2


1

c(22)-2O radial and inverse pyramid c(22)p4g-2O
2 or p p c(4 24 2)-16O
2 [52,156,–167] Disordered (11)-O
1 (hollow) c(22n)p2mg-nO
2 (n=2,3,4) (Pd) [17–182] radial (22)-O
1 (hcp-hollow) c(22)-2O
2 pairing row (11)-O
2 [52,53,54,189–194]

(21)p2mg-2O
2 [42,43,183,184,109, 185]

(21)p2mg-2O
2 [110,168,187] radial (22)-O
1 (hcp-hollow) c(22)-2O
2 pairing row (11)-O
2 [55,56,257,85,189, 195–200]

548

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

reaction at various stages, which enables the atomic valencies in the first two atomic layers to be specified. As will be discussed in detail subsequently, the phase transition originates from nothing more than the alteration of the oxygen valency (as assigned in Table 4) under a certain geometrical configuration at different oxygen coverage. In the following sections, we will analyze these well-determined representatives from the perspective of bond forming. 3.2. O–Cu{(001), (110), (111)} 3.2.1. Observations 3.2.1.1. O–Cu(001). Since 1956 when Young et al. [1] found that the Cu(001) surface is more easily oxidized than other faces of the copper single crystal, there have been many conflicting opinions regarding the oxygen-induced Cu(001) surface reconstruction. Different atomic superstructures have been derived with various experimental techniques p [201] p and theoretical approaches [33,202]. The missing-row (MR) type Cu(001)-( 22 2)R45 -2O reconstruction, first proposed by Zeng and Mitchell in 1988 [149,150], has been elegantly accepted. The latest conclusion [19,38,94,15] about the phase forming kinetics was that a short-ordered nanometric Cu(001)-c(22)-O
1 precursor phase forms at 25 L (Langmuir=10
6 torr.sec) oxyp p gen exposure or lower, and then the ordered MR ( 22 2)R45 -2O
2 structure follows (the valencies of oxygen are denoted based on the current model for readers’ convenience). It was resolved from the STM images in Fig. 4 that oxygen prefers the next-nearest-neighboring hollow site throughout the course of reaction. This implies the strong site-specificity of the oxygen adsorbate [151,203]. At the initial stage of reaction (Fig. 4a), nanometric c(22)-2O
1 domains dominate with zigzag and U-type protruding boundaries. Upon increasing oxygen exposure, the short ordered c(22)p p O
1 phase evolves into the ordered ( 22 2)R45 -2O
2 structure by every fourth row of Cu atoms becoming missing. In the second phase, the ‘dumb-bell’ shaped protrusions in Fig. 4b bridge over the missing rows. The ‘dumb-bell’ protrusions was interpreted as [36]: ‘the pairing of Cu–O–Cu chains by displacing the Cu and/or O atoms next to the missing row by about 0.35 A˚ towards the missing row’ and ‘the O-Cu-O chain is formed by the delocalized anti-bonding states’. The separation of the paired rows was estimated to be 2.9  0.3 A˚. The length of the bright spot was about 5.1 A˚. The height of the bright spot was 0.45 A˚ compared to the 0.3 A˚ protrusions on the pure Cu(001) surface [36]. Clearly, as can be resolved from the STM images for the precursor, the atomic valencies of the Cu atoms sitting inside the domain of ‘depression’ differ completely from that of the Cu atoms composing the protruding boundaries. The atomic valencies of the paired protrusions should also differ from the depressions in the second phase. This was the starting point that triggered the original exercise to identifying the atomic valencies at the surface. We may note that obtaining a highquality STM image is infrequent and the form of the image is often subject to the tip conditions and the applied bias voltages [204,205]. For instance, an asymmetrical tip

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

549

Fig. 4. STM images and the corresponding models [94] for the O–Cu(001) bi-phase structures. (a) and (c) correspond to the nanometric c(22)-O
1 domains with zigzag- and U-shaped protruding boundaries p p [151]; (b) and (d) are the fully developed ( 22 2)R45 -2O
2 structure [36]. The O–Cu–O pairing chains (dumbbell-shaped bright spots) lie along the <010> direction.

or a different bias may modify the appearance of the image but it has less effect on the nature of the bonding occurring at the surface. However, the STM images documented for the O–Cu(001) second phase [206] show a similarity to that in Fig. 4b. The tip effect on the STM image should be very interesting, but it is beyond the scope of the BBB model for oxygen-metal interaction. Therefore, a model for the interaction between the tip and the surface is necessary in this regard. Nevertheless, it would be useful to note that the two sequential phases on the O–Cu(001) surface p p are reversible. It has been found [207] that the ( 22 2)R45 -2O
2 phase reverses

550

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

+ into the initial precursor state after sputtering by an energetic followed p Ar pbeam and by flashing the sample up to 520 K. When annealing the ( 22 2)R45 -2O
2 surface at a temperature corresponding to the ‘dull-red’ color of the Cu surface for half an hour [208], sp-orbital de-hybridization occurs to the oxygen [209]. The atomic geometry of the O–Cu(001) surface has been intensively studied using various techniques. The divergence of opinions is summarized in Table 5.

3.2.1.2. O–Cu(110). The O–Cu(110) surface has also been investigated intensively ever since the pioneer LEED study reported by Ertl in 1967 [2]. LEIS [210,211] and SEXAFS [169,212,213] studies suggest a missing-row type reconstruction where every other [010] Cu row on the surface is absent. In comparison, HEIS [214] and STM [47,215] investigations support a buckled-row (BR) model in which the every other row of Cu atoms is not missing but is shifted outwards instead. Moreover, XRD [57], LEED [216] and ICISS [217,218] measurements indicate an added-row (AR) type reconstruction. STM imaging [219–221] could finally settle the disagreement in favor of the missing-row reconstruction, which, however, should be viewed as an ‘added-row’ phase due to the mass transportation mechanism. The MR and AR models were thought to be the same at the saturation coverage, although the mass transport for these two models is different [222]. The MR model requires removal of alternate [010] Cu rows while the AR model requires addition of alternate [010] Cu rows by Cu-surface diffusion. Step edges presumably serve as sinks or sources for Cu atoms at lower oxygen coverage. However, discrepancies exist on the height of the oxygen atom and on the extent of the interlayer relaxation between the two topmost Cu layers. A summary of the structural parameters pertaining to the O–Cu(110) surface is given in Table 6. It shows clearly that the results vary considerably depending on the data sources. XRD [74] gives a considerably large first-layer expansion D12=1.65 A˚ Table 5 Structural discrepancies for O-Cu(001) surface (Unit in A˚)a Ref.

Method

[81]

VLEED

DCux

DOx

DOz

DCuz

D12

Bond length


0.8 1.90 c(22) 0.3 0.0
0.2
0.1 1.90 [38 SEXAFS 0.25 0.03 0.2 0.1 1.86 2.07
0.15
0 3
0.1 0.8 0–0.25 c(22) [206] STM, LEED 0.3 0.0 0.1 0.1 1.94 1.91 [150] LEED 0.3 0.0 0.1 0.1 1.94 1.81, 2.04 1.84(2) [154] LEED, PED, NEXAFS 0.10.2 0
0.25 0.25 0.10.2 2.05 1.94 [152] LEED 0.10 0.0 0.10
0.05 2.06 [3] EMT >0.0 <0.0 0.3
0.5 [36] STM 0.35 0.35 [93,97] VLEED Bond length: 1.63(1); 1.77(1); 1.94 (2) A˚, ionic bond angle: 102.5, lone pair angle: 140.0 a All are missing row structures unless otherwise denoted c(22). Parameters are the first interlayer spacing, D12, the shift of the atom near the missing row (DCux, DCuz) and the displacement of the adsorbate (DOx, DOz).

551

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685 Table 6 The geometrical details of the Cu(110)-(21)-O surface (unit in A˚)a

ICISS [170] ICISS [210] ICISS [217,218] LEED [216] SEXAFS [213] SEXAFS [283] XRD [48] Theory [8] Theory [23] Model [93] +data [48] Theory [293]

D12

D23

1.60 0.13 1.51 0.04 1.51 0.04 1.49 0.03

1.15 0.06

1.31 1.65 0.05 1.331 1.60 0.05 1.655 1.8770.2

D

1.28 0.20 1.21 0.03

0.12 0.07 0.03 0.03

1.28 0.01

0.0310.005

0.0310.005

DO

DCu(1)-O

DCu(2)-O


0.100.10
0.080.15 0.04 0.03 0.35 0.2
0.340.17 0.48
0.210.1
0.60
0.10.2

1.810.01 1.810.01 1.810.01 1.820.02 1.840.02 1.840.06

1.900.10 2.000.14 2.010.05 1.990.02 2.000.05 1.850.16

1.675

1.92

a Listed are layer spacing D of the corresponding layers, the lateral shift of the second-layer atoms towards the missing first-layer rows, the vertical O position DO relates to the first layer, and the bond length to the nearest and the next-nearest Cu neighbors of the O adsorbate [21].

(+30%) and a second-layer contraction D23=1.15 A˚ (
11%), and the oxygen is located, DO=0.34 A˚, beneath the missing-row layer. In contrast, a theoretical optimization [9] suggested a smaller first layer-expansion D12=1.331 A˚ (+4%) with oxygen located  0.5 A˚ above the top layer. For the clean Cu(110) surface, D12=1.17 A˚ (
8.5%) and D23=1.307 A˚ (+2%) [66,170], as compared to the bulkinterlayer distance Dbulk=1.278 A˚. The lateral displacements (parallel to the surface) of the Cu atoms remain very small. The measured vertical position of the oxygen is quite uncertain and it varies considerably from positions above to below the expanded first-layer of Cu atoms. An average of the reported values is often used. Pouthier et al. [223] examined the reported data and found a strong correlation between the values of DO and D12. By taking parameters for charge transfer [223], and the surface potential barrier [89,93] into account, numerical solutions would be more complex. As can be seen from Fig. 5, the overall trend of the correlation between D0 and D12 can be accounted for by a simple linear expression [88]: 

DO ¼
1:982  D12 þ 2:973  0:136 A : The coefficients have no apparent physical meanings. The correlation between DO and D12 has generated numerous mathematical solutions that have caused longlasting arguments. Nevertheless, all the numerical solutions covered by the correlation region (Fig. 5) could be correct from a numerical point of view. This should be true as the diffraction intensity depends on the arrangement of the scattering centers with various diffraction cross-sections. The cross-section of diffraction should vary with the effective number of electrons of the scattering atom. The arrangement of scattering centers determines the phase-shift of the diffracted beams. The cross-section of the scatter determines the amplitude of the diffracted waves. Therefore, the diffraction

552

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 5. Correlation between the interlayer spacing D12 and the oxygen position DO with respect to the topmost Cu plane (see inset) [88]. The data are given in Table 6. The symmetry points for O and Cu are 0.010.59 and 1.500.16 A˚. Case [8] transforms to [10] can be realized by rotating the O–Cu–O chain around its axis 180 and shifting the axis from 1.166 (O above) to 1.540 A˚ (O below). This indicates the uncertainty of atomic positions derived from diffraction.

can identify nothing about the nature of the scattering centers but only show the resultant effect of scattering from the geometrical arrangement of the scattering centers of different cross-sections. With XRD and DFT, Vlieg et al. [224] determined for the Cu(410)–O surface two detailed structural solutions which give equally acceptable fits to the XRD data after imposition of a Leonard–Jones penalty factor. These models differ especially in the O positions, but one is found to be more favored by comparison with the results of the DFT calculations, and by considerations based on bond lengths and valence. It can be found from Fig. 5 that the vertical position of the adsorbate ranges from DO=
0.01  0.59 A˚ relative to the buckled Cu layer. The spacing between the second and the top Cu layer D12 is 1.50  0.16 A˚. If the top Cu layer buckles upward, the adsorbate will buckle inward, and vice versa. For each pair of values of (D12, DO) that fit the diffraction data there must exist a counterpart set that also fit. For example, the value at point [8] couples with the value at point [10]. Both are the numerical solutions for the same system. This couple can be obtained by rotating the O–Cu–O chain 180 around its axis, and followed by offsetting the O–Cu–O chain in the vertical direction. A half lattice-constant glide shift along the O–Cu–O chain is also necessary. The exchange in the vertical position of O and Cu (0.59/0.16=3.68) seems to approach the ratio of atomic numbers ZCu/ZO=58/16=3.63. This relation infers indeed, to a certain extent, that the cross-section depends on the charge of the specific scattering center. Therefore, all the solutions along the line from [8] to [10] can find a counterpart that could fit the diffraction data. The vertical change of

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

553

positions for both the partners should satisfy the numerical relation between D0 and D12. However, the simple operation on the O–Cu–O chain provides entirely different physical meanings. Therefore, physical constraints should be necessary for the multiple solutions. For instance, the on-surface oxygen should be excluded for the fully developed surface oxidation. This is because the Cu+ and O-2 are hardly detectable using an STM due to the altered atomic valencies. The on-surface oxygen should produce no protrusions in the STM images. As will be discussed later, the vertical positions of the sub-surface oxygen also give different geometrical configurations of the oxide tetrahedron. Considerable efforts with STM imaging have also been made on studies of the O–Cu(110) surface [47,168,215,217,218,220]. A typical STM image and the corresponding model for the (21)-O phase are illustrated in Fig. 6. In the STM image, the round bright spots of 0.8  0.2 A˚ in height are separated by 5.1 A˚ in the [110] direction and 3.6 A˚ in the [010] direction. Single O–Cu–O strings are formed along the [010] direction, or perpendicular to the close-packed Cu row. In contrast, the protrusion for the clean Cu(110) surface is about 0.15 A˚ [47]. Exposing the Cu(110)-(21)-O
2 surface to higher amounts of oxygen at T > 300 K initiates a second structural phase of c(62)-8O
2. STM images of the c(62)8O
2 structure [48,168,225,226] show a corrugation pattern consisting of a quasihexagonal arrangement of ‘ellipsoid’ protrusions. Fig. 7 shows the c(62) domains adjacent to the (21) phase. Higher resolution images of the c(62) structure in the right-hand panel display additional weaker features between the large ‘bumps’. The height of the bright spots was reported to be 0.6  0.1 A˚, slightly lower than the

Fig. 6. STM image [47] and the corresponding models [93] for the surface atomic valencies of Cu(110)(21)-O
2 [219]. The STM grey-scale is 0.85 A˚, much higher than that of metallic Cu on a clean (110) surface (0.15 A˚). The single ‘O
2 : Cudipole : O
2’ chain is zigzagged by the non-bonding lone pairs and composed of the tetrahedron.

554

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 7. STM image [48] and the corresponding models [93] for the surface atomic valencies of Cu(110)c(62)-8O
2 phase. The STM grey-scale is 0.66 A˚. The ‘O
2 : Cudipole : O
2’ chains are paired by the rotation of every other tetrahedron that produce the dipole bridge over the missing row. The pairing chains interlock the Cu+/dipole at the surface with regularly protruding dipoles.

round spot in the (21)-O
2 surface, but much higher than that of the clean Cu(110) surface. Feidenhans’l et al. [48,226] derived a structural model based on their studies using the combination of STM and XRD measurements and EMT computations. The model illustrates that, in the second phase of the O–Cu(110) surface, there are two kinds of oxygen atoms with different vertical positions; every third row of Cu atoms is missed and additional Cu atoms are buckled up, crossing over the missing rows (see Fig. 7). Liu et al. [227] and Dorenbos et al. [228] further confirmed this structure by using the tensor-LEED and LEISS (low energy ion scattering spectroscopy), respectively. The vertical positions of the atoms are listed in Table 7. In the model of Fig. 7, the sublayers (labeled 1 and 3) are composed of Cu and the layers 2 and 4 are oxygen. The atomic valencies in various layers will be specified in the next section based on a reaction formula.

555

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685 Table 7 Summary of the vertical positions of ion cores for the Cu(110)-c(62)-8O surface

STM, XRD, EMT [48,226], Tensor-LEED [227] LEIS [228]

D13

D23

D34

1.2 0.450.12

0.4 0.40 0.12

0.2 0.120.12

3.2.1.3. O–Cu(111). It has been understood that the dissociate adsorption of oxygen on the Cu(111) surface roughens the surface with short-range order [2,229]. HREELS studies [230] suggest that oxygen chemisorbs at the threefold hollow site, either in, or below, the outermost plane of Cu atoms, resulting in a small work function change. LEISS studies [231] revealed that oxygen induces a restructuring that involves a 0.3 A˚ displacement of Cu atoms. SEXAFS [232] determined that oxygen atoms adsorb in the threefold hollow sites with the O–Cu bond of 1.83 0.02 A˚ in length, close to the corresponding bulk value for Cu2O (also the average bond length determined in the current study). The Cu–Cu distance is considerably relaxed (3.15 0.1 A˚) with respect to the Cu-Cu separation in the bulk (2.555 A˚). Furthermore, an IPES study [233] revealed that an extra band around 3.0 eV is produced above EF at %, the center of the first Brillouin zone. These earlier observations provide strong evidence that the O–Cu(111) surface reconstructs with short-range order. The rearrangement of the Cu atoms indicates that the impinging oxygen adsorbate ‘pushes out’ the Cu atoms that roughen the surface of the disordered oxide precursor [21]. With a combination of STM, HEISS, and LEED, Jensen et al. [188] found two new well-ordered, O-induced commensurate reconstructions with extremely large unit cells, 29 and 44 times the (11) surface lattice, after an exposure of 300 L oxygen at 673 K and post-annealing to 723 and 773 K, respectively. Matsumoto et al. [234] confirmed these structural phases in a recent LEED and STM study. As shown in Fig. 8, there is a series of ‘double-holes’ and distorted ‘honey-comb’ frames composed of the ‘dumb-bell’ shaped bright spots as observed in the O–Cu(001) surface.

Fig. 8. (a) STM topography (It=2.9 nA, Vs=55 mV) of a 6060 A˚2 region of O–Cu(111) surface showing the ‘29’ structure. Grey scale is 0.27 A˚. (b) Topography (It=4.9 nA, Vs=7 mV) of a 6060 A˚2 region showing the ‘44’ structure. The grey-scale from black to white corresponds to 0.55 A˚ [188].

556

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

3.2.2. Analysis We now turn to revisiting the well-determined O–Cu{(001), (110), (111)} surface reconstructions from the perspective of bond forming. It is shown that the reaction can be formulated by specifying the atomic valencies at the surfaces. The phase ordering and the derivatives originate from the formation of the O
1 and subsequently the O
2 with sp-orbital hybridization. The various observations from Cu{(001), (110), (111)} surfaces with chemisorbed oxygen are purely the result of geometrical coordination effects on bond forming, because of the same atomic size and electronegativity. 3.2.2.1. Cu(001)-c(22)-2O
1: off-centered CuO2 pairing pyramid. Initially, the O2 molecule dissociates and then the oxygen atom immediately bonds to one atom at the surface. Bond theories [103,235] indicate that it is forbidden for an oxygen adsorbate to form identical bonds with four atoms which are located in the same plane, because oxygen possesses at most four directional orbitals. The model in Fig. 4c states that the O
1 forms a bond with one of its four neighbors at the surface
1 and polarizes the rest. This gives rise to two typical domains in a p the pO ( 22 2)R45 complex unit cell, which can be represented by: O2 ðadsorbateÞ þ 7Cu ðsurfaceÞ  ) 2O
1 þ Cuþ2 ðdomainÞ þ 6Cudipole ðdomain boundaryÞ; ðCuO2 pairing pyramidÞ p p or a (2 22 2)R45 complex unit cell: 2½O2 ðadsorbateÞ þ 6Cu ðsurfaceÞ  ) 4 O
1 þ Cuþ1 ðdomainÞ þ 8Cudipole ðdomain boundaryÞ

ð3:2:1aÞ

ð4CuO off-centered pyramidÞ The [2O
1+Cu+2] or the 4[O
1+Cu+1] domains form the depressed domains and the O
1-induced Cudipole builds up the ‘engaged-cogwheels’ domain boundary that are detected as U- or zigzag-shaped patches of protrusions in STM imaging (Fig. 4a). Larger domains consisting of c(22)-2O
1 unit cells can be constructed by adding more oxygen adsorbates to the surface. The additional oxygen catches one electron from a dipole and polarizes its rest neighbors in the top layer. The model of the surface matrix in Fig. 4c shows the alternative sign of charge distribution both within the domains and in the domain boundaries. Therefore, the surface is fully covered with dipoles in a way that stabilizes not only the domains but also the domain boundaries. At this phase, the surface stress should be tensile though further confirmation is needed. Clearly, the atomic valencies of the copper atoms within the domain (O
1, Cu+1 or Cu+2) differ from that at the domain boundary (Cudipole). It is to be noted that the surface reaction takes place without the second atomic layer being involved at the precursor stage. There are no atoms missing in the short-ordered

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

557

c(22)-2O
1 surface. As optimized with VLEED, the small off-centered shift of the adsorbate (DOz  0.40 A˚. DOx  0.18 A˚ is about 5% of the fourfold hollow dimension) is beyond the resolution of an STM. p p Cu(001)-( 22 2)R45 -2O
2: Cu3O2 pairing tetrahedron. Upon increasing oxygen-exposure the ‘disordered’ nanometric Cu(001)-c(22)-2O
1 domain develp p ops into the ‘ordered’ ( 22 2)R45 -2O
2 phase in which every fourth row of Cu atoms is missing. As an intermediate and quasi-stable state, the O
1 tends to catch another electron from its neighbors. Once its two bonding orbitals are fully occupied, the sp-orbital hybridization follows [97]. As a consequence, a tetrahedron forms with two additional orbitals that are occupied by lone electron pairs of the oxygen. The Cu(001) surface geometry allows the intriguing Cu3O2 pairing tetrahedron to form in such a way that the substrate is involved, as shown in Fig. 4d. A perspective view of the Cu3O2 pairing tetrahedron is illustrated in Fig. 9. We may formulate the O–Cu(001) bi-phase ordering as the effect of the O
1 and subsequently the hybridized-O
2 formation. The complete process of reaction is described as follows:

p p Fig. 9. Perspective of the Cu3O2 pairing tetrahedron in the Cu(001)-(2 2 2)R45 -2O
2 unit cell [97]
2 [reaction formula is given in Eq. (3.2.1b)]. O prefers the center of a quasi tetrahedron. Atoms 1 and 2 are Cu+2 and Cu+. Atom 3 is Cudipole and M is the vacancy of the missing Cu. Atom 4 is metallic Cu atom. The pairing dipoles 3–3 cross over the missing row.

558

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

O2 ðadsorbateÞ þ 4Cu ðsurfaceÞ þ 2Cu ðsubstrateÞ ) 2O
1 þ Cu2þ ðsurfaceÞ - - - - - - - - - - - - - - - - ðCuO2 pairing pyramidsÞ þ 3Cudipole ðO
1 -inducedÞ þ 2CuðsubstrateÞ - - - - - - - - - - - ðbonding effectÞ then, upon increasing oxygen exposure, ) 2O
2 ðhybridÞ þ Cu2þ ðsurfaceÞ þ 2Cuþ ðsubstrateÞ - - - - - - - - - - ðCu3 O2 pairing tetrahedronÞ þ 2Cudipole ðlone-pair inducedÞ þ Cu ðMR vacancyÞ - - - - - - ðbonding effectÞ ð3:2:1bÞ In the first phase, the Cu2+ connects the two off-centered pyramids on the surface to form a CuO2 (see lower side in Fig. 4c). In the second phase, the Cu2+ couples two Cu giving rise to the Cu3O2 structure. The top layer of the Cu(001)p 2O-tetrahedra, p ( 22 2)R45 -2O
2 surface contains rows of Cu2+, a pairing [O
2 : Cudipole : O
2] row and the missing-row vacancy; the second layer is composed of alternate rows of Cu and Cu+. In the second phase, the surface stress seems to be compressive due to the repulsion among the surface dipoles. The adsorbate–adsorbate interaction is always repulsive throughout the course of phase transformation because the identical valencies of them. It is pointed out that as the origin of the observations, the bond forming processes are hardly detectable with currently available means. What one is able to observe is the effect of bonding–dipole protrusions and missing row vacancies. 3.2.2.2. Cu(110)-(21)-O
2 : single O–Cu–O chain. Comparatively, the Cu(110)(21)-O-2 surface reaction, as shown in Fig. 6, can also be expressed as an effect of the hybridized-O
2 formation. The characteristics of this phase is the combination of the alternate missing-row (MR) of Cu and the buckled row (BR) of the ‘O
2 : Cudipole : O
2’ chain. The ‘:’ represents the lone electron pair of oxygen that zigzags the buckled string. The formula for the MR+BR reconstruction is given as (for a c(22) unit cell): O2 ðadsorbateÞ þ 4 Cu ðsurfaceÞ þ 4 Cu ðsubstrateÞ  ) 2 O
2 ðhybridÞ þ 2Cuþ ðsubstrateÞ - - - - - - - - - - - - - ðCu2 O bondingÞ  þ 2 Cu ðMR vacancyÞ þ Cudipole ðBRÞ - - - - - - - - - - - ðthe bonding effectÞ ð3:2:2aÞ or even the added row (AR) model: O2 ðadsorbateÞ þ 4Cu ðsurfaceÞ þ 2Cu ðterrace edgesÞ  ) 2 O
2 ðhybridÞ þ 2Cuþ ðsurfaceÞ - - - - - - - - - - - - - - - ðCu2 O bondingÞ þ 2Cudipole ðbuckled ARÞ - - - - - - - - - - - - - - - - - - - - ðthe bonding effectÞ ð3:2:2bÞ

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

559

From the mass-transport point of view, the Cu(110)-(21)-O
2 phase is made simply by every other Cu-row missing or an O–Cu–O raw being added which is buckled up. Although the mass-transport mechanisms [222] in the MR+BR and the AR models are different, as reflected in the formulae, the cause of the reconstruction (Cu2O bonding) is the same. Formulae (3.2.2) indicate that all the MR, BR and the AR models are correct in the sense of bonding effect, but the combination with their cause (Cu2O bond forming), would be complete. The model can also be applied to the similar fcc(110)-c(n1)-O
2 surface of Ni, Ag [21], Pt [236] and Cu(112) surface, as well as the bcc(112) surface of Mo and Fe with chemisorbed oxygen. Recently, Schroeder et al. [237] and Santra et al. [238] noted that the O–Mo(112) surface shares similar STM patterns to that of the O–Cu(110) surface. Tan et al. [239] have found with STM and LEED similar one-dimensional Cu–O–Cu strings added on the Cu(112) surfaces upon oxygen chemisorption. LEED study [240] revealed a missing-row type reconstruction forms at the Fe(112)-p(21)-O phase where the O–M–O chain runs perpendicular to the close-packed rows of the substrate. From the bond forming point of view, the bi-phase and related phenomena on the O–Cu{(001), (110)} surfaces can be originated from charge transportation, that is, the O
1 and the hybridized-O
2 formation. Oxygen catches electrons one by one from its two nearest neighbors. As the effect of hybridized-O
2 on both the Cu(001) and the Cu(110) surfaces, the lone-pair-induced dipoles are responsible for the protrusions in the STM images. It is clear that the [: O
2 : Cudipole : O
2 :] chain is ‘zigzagged’ by the lone pairs (Fig. 6) rather than it is co-lined by states of anti-bonding, covalent bonding or even ionic bonding. The removal of the missing-row atom from both surfaces results from the isolation of the specific Cu atom as its neighbors have already bonded
2 to the oxygen. The Cu(110)-(21)-O phase, whether added-row or missing-row, p p differs in origin from the Cu(001)-( 22 2)R45 -2O
2 surface by nothing more than that the [O
2 : Cudipole : O
2 :] chain rotates around its axis by 45 to fit itself to the coordination surroundings. The simple rotation has indeed yielded different valencies of atoms in the surface planes. The difference in the coordination geometry between the two surfaces appears to be the origin of the complexity of surface oxidation. Instead of an intermediate-O
1 state like that on the Cu(001) surface, O
2 forms directly at the Cu(110) surface. The oxygen adsorbate may catch two electrons from its nearest neighbors (shortest spacing, 2.552 A˚) in sequence. Meanwhile, the adsorbate also requires extra atoms for the Cudipole to complete the tetrahedron. Due to the expansion of dipole dimension and the repulsion between the non-bonding lone pairs, dipoles tend to expose to the open end of the surface. Apparently, oxygen atoms can hardly find edge atoms from a perfectly finished Cu(110) surface. The situation may change, however, if the (110) face has sufficient terrace defects. This is suggested to be the origin as to why the Cu(001) face is oxidized easier or harder. We may therefore infer that the tetrahedron is the basic and stable building block in oxidation. If the tetrahedron is destroyed, oxygen will seek new partners for a new tetrahedron—a process of re-bonding or bond switching. This may provide a mechanism for O-penetrating in bulk oxidation and O-floating in the epitaxial growth of metals on the oxygen pre-covered metal surfaces [241,242].

560

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

We may also suggest a formula for the Cu(110)-c(62)-8O
2 phase with specification of the atomic valencies in different sublayers (Fig. 6d): 4O2 þ 12Cu ðsurfaceÞ þ 12Cu ðsubstrateÞ ) 4Cu ðMR vacancyÞ þ 2Cudipole ðsupplied by the MRÞ - - - - - - - ðlayer 1Þ þ 4O
2 ðhybrid; large filled circleÞ - - - - - - - - - - - - - - - - - - - - - - ðlayer 2Þ þ 8Cuþ=dipole ðserve as Cuþ for the upper O
2 Þ - - - - - - - - - - - - - - ðlayer 3Þ þ 4O
2 ðhybridÞ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ðlayer 4Þ þ 4Cudipole ðinteracting with the upper O
2 Þ - - - - - - - - - - - - - - - - ðlayer 5Þ þ 8Cuþ ðsmall open circleÞ - - - - - - - - - - - - - - - - - - - - - - - - - - - ðlayer 6Þ ð3:2:3Þ Instead of varying the valencies of the oxygen adsorbate, further exposure at raised temperature increases the oxygen-coverage from 1/2 to 2/3 ML. The process of oxygen re-bonding, which reorganizes the tetrahedron and the Cu atoms at the surface, changes the structural pattern. Self-organization of the Cu2O tetrahedron yields two kinds of oxygen positions (layer 2 and 4), as can be seen from the side view of Fig. 7. The lower O
2, the Cu+ at layer 6 and the Cudipole at layer 3 build up the first Cu2O tetrahedron, which orientates the same as the Cu2O in the (21)-O
2 phase. Acting as a donor for further bonding, the Cudipole at layer 3 provides an electron to the upper O
2 and the Cudipole becomes Cu+/dipole. The non-bonding lone pairs of the upper O
2 then induce the Cudipole at layer 1 and 5 to form another Cu2O tetrahedron with altered orientation, which interlocks the O–Cu–O chains near to the missing row. Differing from the (21)-O
2, the substrate (sublayer 5 and 6) is composed of Cu+ and Cudipole alternatively along the original O–Cu–O chain. They are arranged in a much more complicated array than even that of the second layer of the Cu(001)p p ( 22 2)R45 -2O
2 phase. It is to be noted that two of the four missing Cu atoms become the Cudipole that bridge over the missing row and they are responsible for the recorded STM ‘ellipsoid’ protrusions. The atomic arrangement specified by the tetrahedron agrees with that determined by other researchers (refer to Table 7). One oxygen adsorbate atom locates below the Cu layer labeled 3 and the other oxygen above. The possible mechanism causing such a more complicated phase is suggested to be that thermal energy relaxes the Cu2O bonding and further exposure enhances the self-organization. The reversibility of the bond forming and a suitable ambient produces a phase with more close-packed Cu2O and H-like bond involvement, which is even stable. Fig. 7 also shows the molecule structure of the pairing O–Cu–O chain bridges over the missing row. The protruding dipoles are responsible for the ‘ellipsoid’ STM protrusions. It is worth mentioning that the lone-pair-induced Cudipole (layer 3) interacts further with the upper O
2. Such a set of interactions (O
2 : Cu+/dipole-O
2) forms an identical system to the hydrogen bond by definition. Such configuration has been

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

561

defined as an H-like bond to discriminate the Cu+/dipole from the H+/dipole in the hydrogen bond. According to the bond theory of Atkins [107], one may further infer that the H-like bond forms if the lone-pair-induced dipole combines further with other electronegative specimen through bonding orbitals. Otherwise, an anti-bonding dipole is retained. It is known that the major contribution to an anti-bonding state is made by the less electronegative element [107]. Therefore, anti-bonding dipole-dipole interaction hardly ever forms between specimens with large differences in electronegativity. A typical example of the anti-bonding configuration the quadruples in Fig. 4d, or the ‘dumb-bell’ p is p protrusions in the Cu(001)-( 22 2)R45 -2O
2 STM image. This inference may be necessary in understanding the heterogeneous catalysis. For example, the formation of H-like bonds may provide a more feasible mechanism for forming isocyanate (N, C, O) on the Ru(0001) surface. As found by Kostov et al. [203], the isocyanate could only be formed on the Ru(0001) surface in the presence of pre-adsorbed oxygen. It is expected that the anti-bonding electrons of the dipoles readily combine with new adsorbates (nitrogen or carbon), lowering the system energy. 3.2.2.3. O–Cu(111). Reconstruction phases on the Cu(111)-O surface are most complicated and the corresponding reaction formulae need to be defined. The Cu– Cu lattice distortion (3.15  0.10 A˚ being the scale of dipole-dipole separation) is indicative of oxide tetrahedron formation. The STM protrusions (Fig. 8) and the oxygen-derived DOS ( 3.0 eV) above the EF imply the presence of anti-bonding dipoles in the Cu(111)-O surface. 3.2.3. Quantification: bond geometry and bonding kinetics 3.2.3.1. Kinetic VLEED from O–Cu(001). It is known that energetic electron beams in LEED (E5102 eV) interact with a stack of ordered layers of ion cores. Information derived from LEED patterns is the symmetry and size of the unit cells. Decoding LEED I–V profiles generates knowledge about atomic positions but the accuracy is always under question. In the VLEED (LEED with E < 30 eV, or lower than the plasma excitation energy), electron beams interact with the valence electrons of a few surface atomic layers and with the potential barrier of the surface. VLEED provides highly informative data but sophisticated data processing is required. It has been shown that VLEED (at E < 16 eV) could simultaneously collect non-destructive information from the top layer of a surface about the behavior of atoms (bond geometry) and valence electrons (DOS features) and the morphology (SPB) of the surface [99]. Intensive calculations have been carried out on VLEED from an O–Cu(001) surface [208,243]. The calculations used the code developed by Thurgate [81]. The codep involves multiple-diffraction events and multi-atoms in a complex p Cu(001)-( 22 2)R45 -2O
2 unit cell. Fig. 9 gives the Cu3O2 structure used in the calculations. First, the calculation code reads in the parameters of the first interlayer spacing, D12, the shift of the Cudipole (DCux, DCuz) and the displacement of the adsorbate (DOx, DOz). These data are converted from variables of bond

562

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

geometry. The bond variables vary within a range that is subject to physical constraints. The bonds contract by Q1=0.12, Q2=0.04  0.04, the Cudipole shifts within DCux=0.25  0.25 A˚, and the bond angle BA12 varies within 104.5 . Secondly, the VLEED code optimizes the SPB constants such as the inner potential constant, V0, the amplitude of the inelastic potential, g, and the slope of the energy dependence, d, of the imaginary part of the SPB (Section 2.4). Calculations are performed based on the parameterized SPB by varying z0 over-2.5  1.25 [for pure Cu(001)] in steps of 0.25 (atomic unit). A contour plot of z0 versus E is then drawn with a matching between the calculated intensity and the measurement, Ic(z0i, E)/Im(E)=1.0 0.05. The z0(E) contour plot is the unique yield of the calculation of this method. The plot shows the shape of the z0(E) curve that gives the desired best fit of the measurement and shows all the possible solutions within the limits of parameter variation. This z0-scanning method is also a convenient way to compare different models and to refine the parameters of the SPB and the bond geometry. Once the refinement of the z0(E) plot is complete, the program automatically fits the value of the z0(Ei) to give a desired level of agreement (for example 3% error bar). In contrast to the z0-scanning method, the step of z0 automatically varies from 0.25 to 0.0005 depending on the ratio of =Ic(Ei)/Im(Ei). If  reaches the required precision, calculation will automatically turn to the next energy step Ei+1 of measurement. This method yields simply the geometry-dependent z0(E) profile and reproduces the measured VLEED spectrum. Quantities such as the bond geometry, work function, barrier shapes and energy band structure are automatically given as the product of the data processing. Decoding of the angular-resolved VLEED profiles from the O–Cu(001) surface has yielded static information about the Brillouin zones, band structure and bond geometry [244,245]. Calculations also reveal the non-uniformity and anisotropy of the SPB, and the distribution of the DOS in the upper part of the valence band. It has been revealed that oxygen adsorption has reduced the inner potential constant (V0) by 9.6% (from 11.56 a.u. to 10.5 a.u) and the work function by 1.2 eV from 5.0 to 3.8 eV. The oxygen-reduced V0 is due to the transport of atoms and valence electrons at the surface. The lowered work function arises from surface dipole formation that increases the local charge density. VLEED revealed that, apart from long-term aging, the reaction upon aging and annealing differs only slightly from that upon increasing the exposure of the O–Cu(001) surface. It was discovered from the decoding that annealing at a temperature of ‘dull red’ supplies energy for oxygen de-hybridization rather than the driving force that enhances bond formation. Calculations excluded both the offcentered pyramid structure with oxygen higher than 0.4 A˚ above the top layer and the centered pyramid structure with four identical O–Cu bonds for the precursor c(22)-O
1 phase [209]. More details about the VLEED quantification of the O–Cu(001) bonding kinetics have been reported in Ref. [99]. For comprehensiveness, here we introduce briefly the results that are of immediate relevance. Fig. 10(a–c) shows the VLEED (00) beam reflectance I00/I0 versus the incident beam energy measured at 70 incidence and 42 azimuth angles [208]. It is obvious

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

563

Fig. 10. Exposure-resolved VLEED spectra (a–c) measured at 70.0 incidence and 42.0 azimuth from the O–Cu(001) surface [208], and the calculated results (d–f) from varying the individual bond variables of Q2, BA12 and DCux. Variations of intensities in panel (a–c) at 7.1, 9.1 and 10.3 eV show four reaction stages. Calculations (d–f) for 400 L oxygen-exposure by using bond variables can reproduce the trends of measurement at different stages [97].

that the VLEED spectral features are very sensitive to the oxygen-exposure. The variations of the typical peaks at 7.1, 9.1 and 10.3 eV show that the reaction progresses in four discrete stages:  YO 430 L: When the exposures are smaller than 30 L, the peak at 7.1 eV decreases in magnitude until oxygen exposure reaches 30 L, while other peaks show little change.

564

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

 30 L < YO 435 L: In the region of 30–35 L, the decreased peak intensity at 7.1 eV recovers a little.  35 L < YO 4200 L: From 35 to 200 L, the first peak attenuates while one new peak at 9.1 eV emerges; and then both the peak at 9.1 eV and the peak at 10.3 eV have increasing to maximum values up to 200 L.  YO > 200 L: When the exposures are greater than 200 L, a general attenuation of the entire spectrum occurs. The calculations for the 400 L oxygen-exposure by individually varying the bond variables BA12, Q2 and BL2 can reproduce the measured trends, as summarized in Fig. 10(d–f). Results indicate that:  The oxygen-coverage is maintained at 0.5 ML throughout the course of reaction.  The results, especially, in panels (e) and (f), agree remarkably well with the measured trends given in panels (b) and (c), respectively. This indicates that the process of Cu2O bond forming dominates in the reaction while the SPB is relatively insensitive to the oxygen exposures.  The four discrete stages of reaction can be simply simulated by varying the bond variables individually. For instance, features appearing in the range from 35 to 200 L are dominated by either the increase of ff1O2 or BL2 extension. However, ff1O2 expansion and BL2 contraction is physically reasonable process. The spectral features for samples with oxygen exposure greater than 200 L result from the increase of DCux alone. From 30 to 35 L, the recovery of the peak at 7.1 eV can be realized by increasing the Q2 with smaller ff1O2 and smaller DCux.  Variation of DCuz gives a little change of the spectral intensity between 9.5 and 11.5 eV. Varying the DCuz produces no features that could match measurements. Therefore, the individual shift of an atom at a time does not occur in the real process of bond formation. Fig. 11 shows the offset of the oxygen-exposure resolved z0(E) profiles, which reproduce the measured spectra in Fig. 10(a–c). It is seen that the z0(E) profiles, in general, are relatively insensitive to the variation of the exposure. The z0(E) profiles are similar in shape except for the slight difference below 7.5 eV for 25 L exposure. This further confirms the assumption that the SPB is much less sensitive to the exposure than the bond geometry. The slight outward-shift (-z direction, relative to
2.5 a.u. as indicated by broken lines) of the z0(E) curves at higher exposures increases the n(E), which reduces the work function in the area over which VLEED integrates (mm level). The shape of the z0(E) profile is a joint contribution of the occupied DOS, n(E), and the surface corrugation (local spatial DOS n(x, y)). The non-constant form of the z0(E) can be understood as due to the O-induced ‘rather local’ properties as revealed by STM and due to the non-uniform DOS in the valence band as well. The z0(E) in the z-direction is about (
2.3)–(
3.3) (atomic unit)=0.53 A˚ which coincides with the  0.45 A˚ STM grey-scale [36]. The vacating,

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

565

Fig. 11. (a and b) The exposure-resolved z0(E) profiles provide duplication of the spectra in Fig. 10(a–c). The z-axis is directed into the bulk. The 7.1 eV features on the z0(E) curves (4600 L) agree with those appearing in STS from O–Cu–O chain [47]. Absence of the feature below 7.1 eV for 25 L indicates that the O
1 dominates at this coverage. Straight lines indicate the z0 value (
2.5 a.u.) for a pure Cu(001) surface. The general outward-shift of the z0(E) profiles corresponds to a reduction in the work function [97].

ionizing and polarizing of atoms at the surface result in the strong corrugation of the surface. The small features at 7.1 eV which appeared in the z0(E) curves of 30–600 L coincide with the sharp peak at 2.1 eV below-EF probed with STS from the O–Cu–O chain region [47]. The new occupied states below 7.5 eV are identified as the contribution from non-bonding lone pairs. Hence, the absence of the lone-pair features below 7.5 eV at 25 L oxygen exposure implies that the sp orbitals of the O
1 have not yet hybridized in the O
1-derived precursor phase. It is thus clear that the sp orbitals of oxygen can never hybridize unless the two bonding orbitals of the oxygen are fully occupied. The absence of the 7.1 eV sharp features above 600 L oxygen-exposure is the annihilation of the lone pair information by the anti-bonding dipoles that are shifted outward and are highly saturated, as specified in the SPB model [98]. The violent features at 11.8–12.5 eV come from the band-gap reflection. Features surrounding the band-gap come from electron-excitation near band edges and the formation of standing waves at the boundaries of Brillouin zones. It is noted that the shapes of all the z0(E) profiles are quite similar at energies higher than 7.5 eV. The DOS at the bottom of the valence band of Cu, and even the deeper p-band, are less affected by oxidation. The DOS features for the bonding (
5 eV below EF) are not detectable with VLEED because these features are annihilated by the standing waves at the boundary of the Brillouin zone. Therefore, exercises focusing on the variation of the valence DOS are on right track. The BBB model and the new decoding method for VLEED have thus enabled the O–Cu(001) surface bonding kinetics to be decoded and consistently understood. The

566

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

four-stage Cu3O2 bond forming kinetics is quantified as follows (refer to Fig. 9 and Table 8, also a multimedia as supplementary information):  YO 430 L: The dissociated oxygen atom forms one contracting (Q1=12%) ionic bond with a Cu atom (labeled 1) on the surface. Meanwhile, metallic bonds break up and the missing row forms. The DCux reaches 0.15 A˚ and ff1O2 reaches 94.0 . The O
1 locates above the surface and forms an off-centered pyramid with its surface dipole neighbors. The missing-row atom seems to be attracted by the pairing O
1 and repelled by the dipoles, and eventually evaporates.  30 L < YO 435 L: The O
1 forms the second contracting (increase Q2 from 0 to 4%) ionic bond with a Cu atom (labeled 2) in the substrate second layer; the sp orbitals of the O
2 start to hybridize. The O
2 penetrates into the bulk; meanwhile, the angle ff1O2 expands from 94.0 to 98.0 .  35 L < YO 4200 L: The angle ff1O2 increases from 98.0 to a saturation value of 102.0 , which causes the first interlayer spacing D12 to expand by about 10%, while other parameters have little change.  YO > 200 L: The interaction between the O
2 and the lone-pair-induced Cudipole develops. Lone pairs push the Cudipole outward, and consequently, pairing dipoles form and bridge over the missing row. The DCux increases from 0.15 to a maximum 0.45 A˚ at 800 L and above. Table 8 Four-stage O–Cu(001) surface bonding kinetics [97]a Reaction stage

1 ( <30 L: BL1 formation); 2 (3035 L: BL2 and ff1O2 change); 3 (35200 L: ff1O2 expansion); 4 ( ff 200L: DCux increase)

Exposure

(L)

Q2 Bond BA12 geometry (Q1=0.12) DCux

25

30

35

0 92.5 0.125

0 94.0 0.150

0.04 98.0

50 0.04 100.0

100

200

101.0

102.0

Atomic shift
DCuz 0.1460 0.1440 (A˚)
DOx 0.1814 0.1796 DOz
0.0889
0.0447

0.1268 0.1802 0.0618

0.0938 0.1831 0.1158

0.0844 0.1852 0.1422

0.0709 0.1877 0.1682

Layer

D12

1.7522

1.7966

1.8287

1.8824

1.9086

1.9343

Bond length BL1 (A˚) BL2 BL3

1.628 1.850 1.8172

1.8326

1.776 1.8833

1.8983

1.9053

1.9121

Bond angle BA13 95.70 ( ) BA23 91.80 BA33 165.87

98.82 93.11 160.83

104.24 95.75 145.26

105.12 96.46 144.32

105.64 96.83 143.02

105.33 95.18 141.82

400

600

5800

0.250

0.355

0.450

0.1495

0.2239

0.2849

1.9262

1.9396

1.9505

105.30 99.52 139.43

104.01 101.67 135.38

103.83 103.43 135.71

a Empty space is identical in value to that in the corresponding cell of the previous column. All information is provided by the controlling variables (BA12, Q2, DCux). Error bar for bond length is 0.010 A˚ and for bond angle is 0.2 . The SPB constants: V0=10.50 eV, =
0.9703, =6.4478.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

567

Representing the joint spatial and energy DOS, the variations of the structuraldependent z0(E) profiles (Fig. 11) agree with the bonding kinetics:  The slight outward-shift of the z0(E) profile at higher exposures reduces the work function, which corresponds to the development of the anti-bonding Cudipole.  Features below 7.5 eV, particularly the small sharp peak at 7.1 eV, are derivatives of the non-bonding lone pairs of the O
2. The absence of these features at 25 L originates from the O
1 precursor, in which no lone pair has formed; while at higher oxygen exposures the lone pair information is annihilated by the protruding Cudipole characteristics.  Similarity of the fine-structure shapes at energies higher than 7.5 eV of all the profiles implies that electrons in the lower part of the valence band involve no much in the process of charge transportation. In contrast, electrons in the upper part of the valence band dominate in oxidation: holes and lone pairs form simultaneously. The DOS features of bonding (around 12 eV) are not detectable with VLEED due to the standing wave formation in the boundaries of Brillouin zones.

3.2.3.2. XRD from O–Cu(110). Feidenhans’l et al. [74] determined the atomic structures of the O–Cu(110) surface using XRD. Their results agree with the trend predicted by EMT optimisations for the minimized total binding energy of the system [3]. Both XRD and EMT approaches revealed that the oxygen adsorbates are located underneath the missing-row top layer and the zigzag O–Cu–O chains are formed (refer to Fig. 6). Referring to the inset in Fig. 5, the vertical displacement of the dipole (labeled 3) is denoted as D3z. The distance from oxygen to the buckled Cu top layer is DOz. The lateral displacement of the Cu+ row in the second layer is D1x. The z-axis is directed into the bulk. Thus the bond parameters can be converted from these data [74]: 

ðD1x ; DOz ; D3z Þ ¼ ð0:0310:005; 0:340:17; -0:370:05Þ A :

ð3:2:4Þ

As can be seen from Table 9, a slight change of the DOz varies the bond geometry. The different values of bond lengths represent quite different meanings in the physics. The value of DOz=0.34 A˚ gives a tetrahedron with four O–Cu bonds that are nearly identical in length (Column I). The value of BL3 41.85 A˚ implies that the atom labelled 3 is an ionic one. The Cu+ ion is hardly detectable by STM due to the partially emptied d-states and the reduced atomic radius (from 1.27 to 0.53 A˚). This is apparently conflicting with the STM images that show the Cu atom buckling out of the surface. Lowering the oxygen adsorbate within the error bar (0.17 A˚) to DOz=0.51 A˚ gives results in Column II. BL1 has a 6.5% contraction and BL3 extends slightly compared with the standard ionic bond length, 1.85 A˚. The bond geometry is now acceptable as the shorter bond corresponds to the ionic state of the Cu. Structural parameters in column II are acceptable while those in column I are strictly forbidden in line

568

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Table 9 Comparison of the tetrahedron bond geometries derived from the XRD from Cu(110)-(21)-O
2 and the p p VLEED from Cu(001)-( 22 2)R45 -2O
2 [93] p p ( 22 2)R45 -2O
2 Conclusion Variable Cu(110) (21)-O
2

DOz BL1 (A˚) BL2 (A˚) BL3(2) (A˚) BA12 ( ) BA33 ( ) BA13/23 ( )

I

II

III

40.34 51.85 51.85 41.84 490.0 5156.7 497.4

0.51 1.73 1.73 1.88 96.0 146.5 100.2

0.60 1.675 1.675 1.921 102.5 140.3 102.3

1.628 1.776 1.926 102.0 139.4 99.5/105.3

< 1.85 < 1.85  1.92 <104.5 140.0 102.5

with physical constraints and STM observations. The ideally suggested case, assuming the CN of the O
2 and the Cu+ as 4 (Q=0.12) and 8 (Q=0.03), is shown in column III. By adopting the value D3z=0.37+0.05 A˚, one can then insert into this frame a tetrahedron [BL1=BL2=1.32(1
0.12)+0.53(1
0.03) =1.675 A˚], in which the O
2 ion is located 0.6 A˚ beneath the dipole layer. The geometry of this tetrahedron is nearly identical to that determined for the p p Cu(001)-( 22 2)R45 -2O
2 phase. Although the patterns of reconstruction and morphology are different, the basic tetrahedron is the same in both the Cu(001)-O
2 and the Cu(110)-O
2 surfaces. 3.2.4. Summary In summary, we have obtained the quantitative information about the bond geometries and bond forming kinetics on the O–Cu(001) and the O–Cu(110) surfaces. VLEED calculations revealed that the Cu3O2 pairing-tetrahedron evolves from the CuO2 pyramid on the Cu(001) surface [Fig. 4(c and d)]. Both the bond theories and the VLEED calculations suggest that, in the precursor phase, the O
1 locates eccentrically above the fourfold hollow-site of the c(22)-2O
1 domain to form an off-centered pyramid. The off-center shift of the O
1 is DOx1.807
[1.85(1
0.12)]0.18 A˚, which is comparable to the values of 0.10– 0.13 A˚, as reported previously [38,154]. The vertical distance of the O
1 to the surface is  0.4 A˚ above the surface. The absence of the lone-pair DOS features implies that no sp-orbital hybridization occurs of oxygen in the O
1-induced precursor phase. Except for the Cu(001)-c(22)-2O
1, all the available phases on the O–Cu(001) and the O–Cu(110) surfaces are composed of the primary Cu2O tetrahedron. The parameters for the Cu2O tetrahedron are nearly the same, as summarized in Table 9. The VLEED gave quantitative data that the work function is reduced by  1.2 eV and the inner potential constant decreases from 11.56 eV for the clean Cu(001) surface to 10.50 eV upon being oxidized. The SPB parameters vary from site to site on the surface [88]. At the dipole site, z1 ffi z0,  ffi l
1. This means that the metal dipoles

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

569

enhance the SPB through the outward shift of the wave function. Dipole formation also strengthens the degree of saturation of both the real and the imaginary part of the SPB at the dipole site. The SPB of the Cu(001)
O
2 surface varies considerably from that of the clean Cu(001) surface. The z0M at dipole site is (z0M/z0(Cu)=3.37/2.50) 1.35 times and p the lM is (lM/l(Cu)=1.27/0.9) 2 times that of the pure Cu(001) surface. The values of z0M and lM quantify the protrusions in the STM image to a certain extent as the higher the electronic islands are, the denser the electrons will be there. In the missing-row site, z1 < < z0,  > > l
1, i.e., the missing-row vacancy is not occupied by ‘free electrons’ of the solid. This may quantify the depression in the STM imaging. On the Cu(001)–O
2 surface, the lowest saturation degree andpthe smallest z-scale of the SPB is (z0m/z(Cu)=1.75/2.5 ffi lm/l(Cu)=0.65/0.9) 1/ 2 times that of the clean Cu(001) surface. Therefore, electrons at the surfaces with chemisorbed oxygen are rather local. It is reasonable to describe metal surfaces with chemisorbed oxygen at higher coverage as a non-Fermi system. This is because of the lack of freely moving electrons at the surface. This mechanism should cause the non-Ohmic rectifying and higher contact resistance even though the local work function is much lower than that of the clean Cu(001) surface. It is understandable now why a standard freeelectron resistance mechanism could not work in such as strongly localized system [41]. Decoding the exposure-resolved VLEED data from the Cu(001)–O
2 surface revealed that three bond parameters dominate the four-stage bonding kinetics. In the process of oxidation, the oxygen adsorbate first forms one Goldschmidt-contraction ionic bond (1-O) to a Cu atom on the surface, and then another contracting ionic bond (2-O) follows between the oxygen and a Cu in the substrate. As a result, the oxygen adsorbate buckles into the bulk. Then, the ionic bond angle (ff1O2) increases, leading to the relaxation of layer spacing. Finally, interaction develops between the lone pairs of oxygen and the lone-pair-induced metal dipoles (3-O). During the process of oxidation the change of the SPB is not apparent except for the DOS features of the non-bonding lone pairs. From the perspective of bond forming, the origin of the bi-phase ordering on both the O–Cu(001) and the O–Cu(110) surfaces has been clear. It is interpreted that
2 the formation of O
1 and subsequently the hybridized-O gives in nature the p p
1 Cu(001)-c(22)-2O and then the Cu(001)-( 22 2)R45 -2O
2 phases. Reassembling of the primary Cu2O tetrahedron transforms the Cu(110)-(21)-O
2 into the Cu(110)-c(62)-8O
2 phase at elevated temperatures p andphigher exposures. As consequences of O-2-hybridization, the Cu(001)-( 22 2)R45 -2O
2 differs from the Cu(110)-(21)-O
2 in origin by nothing more than the fact that the [O
2 : Cudipole : O
2] string rotates itself by  45 to match the specific coordination environment. Such a chain-operation yields entirely different reconstruction patterns and surface morphologies of the two surfaces. Therefore, the phase ordering on the Cu(001)–O and the Cu(110)–O surface is simply the consequence of the Cu2O formation at different stages and under various bonding circumstances. The mechanism for the O–Cu(111) surface reconstruction should be the same because the Cu{(001), (110), (111)} surfaces differ one from another by nothing more than the crystal orientation.

570

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

3.3. O–(Rh, Pd)(110) 3.3.1. Observations It has been clear that oxygen adsorbate occupies the long-bridge hollow-site on the fcc(110) surfaces of Cu, Ni, Ag, and Pt to form a tetrahedron with its four surrounding atoms. Such a manner of occupation yields the alternative ‘O
2 : Mdipole : O
2:’ string and the ‘missing-row’ vacancies perpendicular to the close-packed direction of the fcc(110) surface. In contrast, the oxygen adsorbate was found to prefer the alternate hcp(0001) or fcc(111) facet site (see Fig. 1) and form rope-like strings along the close-packed direction of the fcc(110) surface of Rh [171–182] and Pd [246–249]. Adsorbates tend to locate at the troughs crossing a row of Rh or Pd atoms to form the zigzag O–M–O chains instead. The preference of hcp or fcc site of the oxygen is still under debate. LEED studies [250–254] revealed five patterns of (12), (13), c(24), c(26) and a ‘complex’ structure. A series of the ‘complex’ superstructures of (23) and c(24) has also been observed at 100 K and under 3 L oxygen-exposure. A tensor-LEED study [255] suggested that the reconstruction occurs in the (23) and c(24) modes. This gives rise to the corresponding (13) and (12) periodicity at the Pd(110) surface. At low temperature (170 K) and low coverage, photoelectron diffraction study [256] suggested that the oxygen adsorbate prefers short bridge or nearly short-bridge sites on the Rh(110) surface rather than the threefold sites [161]. Fig. 12 shows the STM images of the O-Rh(110) surface obtained for various values of oxygen coverage. [174,177–179] These images are essentially the same as those probed from the O–Pd(110) surface [246]. Grids may be applied to the STM photographs, in order to specify the locations of the adsorbates as represented by the small dark spots. The zigzag or ‘saw-tooth’ like protrusions form along the close-packed direction. Oxygen adsorbates rest beside a protrusion spot. The distinct features of the STM images and the corresponding interpretations can be summarized below:  Two kinds of zigzag strips of depressions appear. The depressions were assumed rows of metals that have been missed out upon reconstruction. One strip of depression corresponds to the (22)pmg-2O (a mirror and a glide symmetry) arrangement and the other to the (22)p2mg-2O phase (twofold mirror and a glide symmetry).  The strips of bright protrusions were explained as buckled Rh or Pd atoms of the first layer with lateral displacements in a zigzag fashion.  The scale difference crossing the missing-row is  0.7 A˚ while it is 0.16 A˚ along the row of protrusions.  It is interesting to note that, in Fig. 12(d), the protrusions of the rows next to the ‘missing rows’ are higher relative to that of other protruding rows. The O-induced reconstruction, in both the O–Rh(110) and O–Pd(110) surfaces, is usually assumed as the missing-row type. One out of a certain number of the Rh(Pd) rows is removed from the surface. Common to other oxygen–metal systems, the first layer spacing expands and the second contracts.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

571

Fig. 12. STM images of the O–Rh(110) surfaces [174,177]. The bright protrusions were explained as buckled metal atoms with lateral displacement while the dark stripes were assumed as missing row vacancies. Grids were applied to locate the positions of oxygen adsorbates as the dark spots. Images were recorded at conditions for (a) and (b) (2.0 V, 1 nA); (c) (0.4V, 0.25 nA) and (d) (0.35 V, 0.25 nA). Panel (d) distinguishes the z-scale difference of the protrusions next to the missing rows.

Fig. 13 illustrates the atomic structural models for the O–(Pd, Rh)(110) surfaces for different values of oxygen coverage. Indicated are the (21)p2mg-2O (1.0 ML), (22)p2mg-2O and (22)pmg-2O (0.5 ML) unit cells. These unit cells compose the complex c(22n) (n=3, 4, 5) structures. Indicated are also phases corresponding to 1/3 ML, 2/3 ML and 4/5 ML oxygen coverage. The atomic structural model suggested that:  At very low coverage, oxygen occupies a fourfold hollow site (refer to Fig. 1b) of C2v symmetry.  At YO < 0.5 ML, the oxygen adsorbate starts to induce a (12) missing-row reconstruction of the surface. Oxygen atoms occupy every other fcc(111) facet site along both sides of a metal row at the close-packed direction. Each adsorbate interacts with two atoms at the surface and one atom in the second substrate layer.  At YO=0.5 ML, the (12) missing-row reconstruction is fully developed.  At YO > 0.5 ML, the (12) missing-row reconstruction starts to degrade. The missing-rows are gradually replaced by the protruding metal rows until all the missing rows are fully recovered.  At YO=1, the O-fcc(110) surface is again unreconstructed and the oxygen adsorbates form a (21)-2O LEED pattern.

572

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 13. Missing-row type atomic structural models for the Rh(110) and Pd(110) surfaces with chemisorbed oxygen [179]. Different phases are produced depending on oxygen coverage. Oxygen occupies the apical site of a tetrahedron and forms O-metal bonds to two atoms at the surface and one atom in the second substrate layer. The number of missing rows decreases with increasing oxygen coverage. Unit cells for typical phases and the corresponding coverages are indicated. The D represents the distance between the O–O zigzagged chains across the protrusions.

It is surprising that, as the oxygen coverage increases, the number of missing rows reduces from every second (1/2) to one in n (1/n) and, finally, to zero. A reasonable explanation is yet lacking how the adsorbates can turn the atomic vacancies into real atoms that are then buckled up. LEED [173,176,179,181,182] and DFT [172,175] optimizations suggested that oxygen adsorbate prefers the hcp(0001)-facet site in the Rh(110)-(21)p2mg-2O phase. Oxygen atom sits 0.5–0.6 A˚ above the top layer and bonds to one atom in the first layer and to two in the second. The O–Rh bond lengths are 1.86–1.97 A˚ to the Rh in the top layer and 2.04–2.07 A˚ to the Rh in the second layer. Table 10 summarizes the geometrical values for the O–Rh(110) surface reconstruction. The D in Fig. 13, labeling the distance between the oxygen rows, is 2.26–2.80 A˚. Under the influence of oxygen chemisorption, the [100] row in the Table 10 O–Rh(110) surface atomic geometry (Dbulk=1.34A˚)

(21)p2mg-2O [182] (21)pg-2O (22)pg-O [176] (22)p2mg-2O (21)p2mg-2O (DFT)[172] Sub-surface O [179]

DOz

D

D12

D23

Bond length

0.60 0.6 0.54 0.71 0.66 0.50

1.13 1.16 6.99 1.13 0.70 1.10

1.33 1.36 1.34
3%
3%
3%

1.39 1.38 1.27

1.86, 2.07 1.97; 2.04 2.00; 2.05 1.99; 2.06 2.00, 1.88

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

573

second layer is distorted in a zigzag fashion by  0.1 A˚ with the Rh atomic positions shifted towards the nearest oxygen positions. However, agreement needs to be reached on the preferential site of oxygen, which of the fcc(111) or the hcp(0001) facet is preferable and what the vertical distance of the oxygen atom is. 3.3.2. Analysis In order to examine the effects of electronegativity and the scale of the lattice constant on the observations, we may compare the patterns of reconstruction of the O–(Rh, Pd)(110) surfaces with those of the O–Cu(110) surface. It is easy to understand that, reacting with the more open (Pd, Rh)(110) surfaces, oxygen adsorbates move from the C2v hollow sites to the threefold fcc(111)-facet sites rather than the hcp(0001) facet sites of the fcc(110) surface (Fig. 1). Bonding to two atoms in the second layer will create the same pattern of reconstruction occurring on the O–Cu(110) surface: a protruding row perpendicular to the close packed direction. The fcc(111)-facet sited oxygen must find a fourth atom to form the tetrahedron inside which the oxygen adsorbate locates. As illustrated in Fig. 14, the triangles indicate the primary tetrahedron (1233) in the corresponding (21)p2mg2O, (22)p2mg-2O and the (22)pmg-2O phases. The nature and kinetics of the bonding, as well as the individual valencies of surface atoms can be formulated as below.  In the precursor phase, or at very low oxygen coverage, oxygen deposits randomly into the C2v hollow site to form a B5O cluster. This cluster can be expressed as: O þ 4Bð1st layerÞ þ Bð2nd layerÞ



) O
1 þ Bþ beneath the O
1 þ 4Bdipole induced by O
1

ð3:3:1Þ

B5O forms by bonding to the B atom underneath and the O
1 polarizes its four surface neighbors.  At YO=1/3 ML, O
2 develops and the (23)-2O
2 unit cell forms. The (23) unit cell contains two quasi-tetrahedron, which can be expressed as: O2 þ 6Bð1st layerÞ þ 6Bð2nd layerÞ ) 2O
2 ðhybridÞ þ 2Bþ ð1st layerÞ þ 2Bþ ð2nd layerÞ ðB2 O bondingÞ þ 2Bdipole ðbuckled upÞ þ 4Bð1st layerÞ þ 2Bð2nd layerÞ ðthe bonding effectÞ ð3:3:2Þ Oxygen retains the bond to the B atom underneath and gets another electron from the B atom in the surface row near the protrusions. Lone pair formation induces B dipoles that form the row of protrusions. The protrusions are zigzagged because the atomic coordination determines that the tetrahedron has to rotate slightly

574

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 14. Bond model for the pmg and p2mg reconstructions of the O-(Pd, Rh)(110) surface [171] specifies the STM protrusions to be the metal dipoles. Depressions arise from the B+ ions rather than missing row vacancies. O
2 locates always in the center of a quasi-tetrahedron represented by the unit 1233. Surface bond network with involvement of H-like bond interlocks all the surface atoms and hence there are no atoms are missing during the reaction. The top left insertion illustrates how the lone pairs of the oxygen adsorbates polarize and deform the metal dipoles. Two kinds of depression rows correspond to the pmg and p2mg glide symmetry of adsorbate distribution. The number of depressed rows decreases with increasing oxygen exposure because the B+ converts to Bdipole with oxygen coverage increase.

(imagine a triangle consisting of three apexes of a tetrahedron). On the other hand, the tetrahedron rotation releases the compression between the lone pairs of an adsorbate as the atomic distance along the close packed direction may be too short for the Bdipole–Bdipole spacing (see model in Fig. 3a). Except for the dipole in the buckled-row, every other B atom in the top and the second layers near the oxygen becomes B+. Therefore, interaction between the B+ and B along the B–B+–B row (which used to be assumed as a missing-row) becomes stronger than the interaction between pure metals. The nearly free electrons along the B–B+–B row become less dense. The B atoms and B+ ions are hardly detectable with an STM, being similar to the images of missing row. Therefore, it is understandable why the invisible rows are often referred to being ‘missing row’ in previous models. If the adsorbate locates at the hcp(0001) facet site, the B–B+–B row should be composed of a Bdipole, which is the case for the O–Cu(110) surface.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

575

 At YO=0.5 ML oxygen coverage, the hybridized-O
2 gives rise to the (22)pmg-2O
2 or the (22)p2mg-2O
2 phase. The unit cells contain each a pair of quasi-tetrahedron and it can be formulated as: O2 þ 4Bð1st layerÞ þ 4Bð2nd layerÞ ) 2O
2 ðhybridÞ þ 2Bþ ð1st layerÞ þ 2Bþ ð2nd layerÞ ðB2 O bondingÞ þ 2Bdipole ðbuckled upÞ þ 2Bð2nd layerÞ ðthe bonding effectÞ

ð3:3:3Þ

 At YO=1.0 ML oxygen coverage, the (21)p2mg-2O
2 phase forms: O2 þ 2Bð1st layerÞ þ 2Bð2nd layerÞ ) 2O
2 ðhybridÞ þ 2Bþ=dipole þ 2Bþ ð2nd layerÞ ðB2 O bondingÞ

ð3:3:4Þ

At this stage, an H-like bond forms, which lowers the STM protrusions and stabilizes the surface. As the surface atomic ratio O : B=1, each adsorbate must interact with three B atoms at the surface. Therefore, each B atom becomes B+/dipole as two lone pairs are acting on it. The adsorbate drags the electron cloud of the dipoles to compensate for the lack of one atom for the tetrahedron. The formulae for all the possible phases on the O–(Rh, Pd)(110) surfaces represent the kinetics of bond formation. The oxide tetrahedron forms by evolving a B5O cluster into a B4O and then a B3O cluster. During the transition of the B5O into the B4O, the fifth B atom at the surface is released and then it is involved in another new B4O cluster formation. The lack of one atom in the B3O cluster at even higher coverage is compensated for by the formation of the H-like bond, as illustrated in Fig. 14. Dipoles provide electrons to the oxygen to form the second ionic bond. In the (22)-2O
2 phase (0.5 ML), the entire second layer is half composed of B+ ions and another half of B atoms; in the (21)p2mg-2O
2 phase (1.0 ML), the second layer is fully composed of B+ ions. The displacement of B and B+ in the second layer is not avoidable subjecting to the bond geometry. The reaction formulae indicate that the Bdipole or B+/dipole gradually replaces the + B that used to be assumed as the ‘missing row’ vacancy. Hence, the number of the invisible rows decreases with increasing oxygen coverage. This figure agrees well with experimental observations. In terms of missing-row formation, a mechanism for the mass-transport can hardly be established. It is unlikely that the oxygen adsorbate is able to turn the atomic vacancies into metal atoms that are buckled up. The current model, however, defines a feasible mechanism for the mass-transport. It is understandable that the B+ row evolves into the Bdipole or the B+/dipole row with oxygen addition. Therefore, the zigzag depressions in the STM image correspond to B+ ions (row-I and row-II in Fig. 14) rather than to missing-row vacancies. The shape-difference between the depressed row-I and row-II is due to the different glide symmetries of the oxygen adsorbates as can be seen from Fig. 14. Row-I corresponds to the (22)pmg-2O-2 symmetry while row-II to the (22)p2mg-2O
2. It was noted earlier that the H-like bond forms by dragging the electron cloud of the protruding dipoles to the bonding orbitals of the oxygen and hence the H-like bond

576

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

formation lowers the protrusions. Therefore, the slightly higher STM protrusions near the invisible row can be understood as the absence of the H-like bond near the B+ rows. It is important to note that the B4O tetrahedron bonding requires that oxygen is located inside a tetrahedron rather than at the apical site. Therefore, the bond geometry determines the relaxation of the first interlayer distance. Due to the reduced atomic sizes of metal ions in the second layer and the strong interaction between the second B+ layer and the B in the third layer, the second interlayer distance contracts. In this sense, the bond model is able to account for the observed layer spacing relaxation and mass transportation as well as the surface morphologies of various phases. It is obvious that the difference in the host atomic size and host electronegativity determines the site specificity of the oxygen and the orientation of the tetrahedron. This delineates the (Rh, Pd)(110)–O surface from the Cu(110)–O surface in terms of reconstruction patterns, although the basic oxide tetrahedron is common. 3.4. O–(Co, Ru)(1010) 3.4.1. Observations The hcp(1010) plane with close-packed rows separated by the c-axis distance is an analog to the fcc(110) surface with slight difference in layer spacing (as illustrated in Fig. 1). In contrast to the (21) missing-row structures formed on the O–(Cu, Ni, Ag, Pt)(110) surfaces, three intriguing superstructures have been observed in sequence on the O–Co(1010) surface [42,43,183–185] and two phases on the O–Ru(1010) surface [110,186,187]. From the standpoint of microscopy and crystallography, O–(Co, Ru)(1010) multiphase ordering has been well identified. Discrepancies still exist on the vertical positions and the site specificity of the oxygen adsorbates. In an earlier LEED study, Schwarz et al. [184] found that the Co(1010)(21)p2mg-2O phase is developed from the p(21)-O or the c(24)-4O phase by increasing the oxygen coverage at higher temperature. Based on the STM observations, Koch et al. [42,43] proposed several models for the reconstructed O–Co(1010) phases. LEED calculations [185] suggested that in the final (21)p2mg-2O phase, oxygen occupies the threefold-coordinated fcc(111) facet site and bonds to two Co or Ru atoms in the first atomic layer and one Co or Ru atom in the second, this being the same as that occurring at the O–(Rh, Pd)(110) surface. Oxygen adsorbate rests above the top layer of the otherwise unreconstructed surface. The oxygen adsorbates prefer locations between two neighboring metal rows and form a zigzag O–O chain along the close-packed direction. Contrastively, oxygen adsorbates prefer the same fcc(111) facet sites but locate beside one specific (Rh, Pd) metal row to form the zigzag O–M–O row at lower coverages. This is the major difference between O–(Rh, Pd)(110) and O–(Co, Ru)(1010) in observation. Furthermore, chemisorption of oxygen causes a significant expansion ( 25%) of the first Co interlayer spacing and a slight contraction (  5%) of the second Co interlayer spacing with respect to those of the bulk. A LEED, DFT and HREELS study conducted by Schwegmann et al. [63] revealed that the O–Ru(1010) surface shares the same

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

577

reconstruction patterns as the last two phases of O–Co(1010) despite the slightly different details concerning atomic positions and layer spacings. Tight-binding calculations yielded the apparent O-derived DOS features in the valence band and above of Ru, which are quite the same as those for the N–Ru(0001) surface calculated by the same group [257]. The DOS features for both the N–Ru(0001) and O–Ru(1010) coincide with the current model specifications (Section 2.3) of bonding, non-bonding, holes and anti-bonding states. The O–Co(1010) tri-phase ordering and the structural models are summarized below:  The disordered p(21)-O forms upon flashing the c(24)-4O phase to 450 K. It can be seen from Fig. 15 (a) that the STM image exhibits a checkered pattern, randomly filled by black and white rectangles. The dimensions of the rectangles are 5.0 A˚ ( 2aCo) by 4.0 A˚ (  cCo) along the [1210] and the [0001] directions, respectively. The gray scale is about 1.1 A˚. It is reasonable to

Fig. 15. (a) STM image [43] and (b) the corresponding bond configuration [183] for the Co(1010)-p(21)2O
1 precursor phase with randomly filled checkered domains. The pairing O
1–O
1 dimer rests on the Q top of surface atoms forming a ‘ -shape’ dimer bond along the close-packed direction.

578

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

consider this less-ordered p(21)-O phase as the precursor state because we have found, using VLEED, that the annealing supplies a disturbance rather than a driving force to enhance the reaction and that the O–Cu bonding processes are reversible [99].  The c(24)-4O phase forms when the clean Co(1010) surface is exposed to 2.5 L oxygen at 300 K. STM imaging revealed that the ordered c(24)-4O phase forms uniformly over large areas [see Fig. 16(a)]. Rows of white oval bumps with double Co periodicity (2aCo  5.0 A˚) are present in the direction of the close-packed Co rows. The rows of bumps are separated by 8.0 A˚ ( 2cCo). The bright bumps,  1.1 A˚ in height, are resolved as a honeycomb-like pattern separated by the zigzag depressions. From the STM image, Koch et

Fig. 16. (a) STM image of the Co(1010)-(24)-4O
2 surface [43]; (b) the corresponding bond configuration [183]; and, (c) the hard-sphere model for the Ru(1010)-(24)-4O
2 structure [63]. A tetrahedron is framed by a dotted square. The lack of one (Co, Ru) atom for the tetrahedron is compensated by a virtual bond between O
2 and the electron cloud labeled 2, which sharpens the tip of the honeycomb-like bumps in the STM images.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

579

al. suggested that oxygen occupies the hcp(0001) facet site and reacts with one Co atom in the top layer.  The (21)p2mg-2O phase was obtained by dosing oxygen (5  10 L) at room temperature to either the c(24)-4O or the (21)-O phase. STM revealed a regular array of protrusions, being similar to that of clean metal surfaces in the scale of 40.3 A˚. All the Co atoms of the topmost plane are at nearly the same levels of height. The model proposed for this phase interlocks the model for the c(24)-4O phase with oxygen preferring the hcp(0001) facet site. However, LEED studies by Gierer et al. [185] have ruled out the hcp(0001) facet site preference of the adsorbate. It is suggested that the adsorbate atoms are located above the fcc(111) facet site instead, agreeing with the model proposed by Comelli et al. [180] for the O–(Rh, Pd)(110) surfaces. The optimal atomic geometry is that oxygen atoms reside in the fcc(111) facet sites 0.74  0.05 A˚ above the first Co layer. The lateral distance between oxygen and the densely packed Co rows is 1.13  0.10 A˚ (D/2 in Fig. 13). The O–Co bond lengths are estimated to be 1.83  0.10 A˚(1) (to the Co atom in the first layer) and 1.99  0.10 A˚(2) (to the Co atoms underneath the oxygen). The oxygen-derived structure of the Co(1010) surface is one where the first Co interlayer expands from 0.62 A˚ to 0.90 A˚, which amounts to 25% with respect to the bulk value, 0.72 A˚. Comparatively, the O–Ru(1010) bi-phase ordering and the structural models are nearly the same:  The c(24)-4O and the (21)p2mg-2O phases form in sequence on the O–Ru(1010) surface at room temperature by oxygen exposure of 0.7 L and 2.5 L, respectively. This is easier to achieve than the formation of these phases on the O–Co(1010) surface.  According to the LEED and DFT data, [63] the oxygen adsorbate locates at the apical site of a tetrahedron and interacts with two Ru atoms in the top layer and one in the second layer. Oxygen atoms reside in the threefold hollow sites 1.02 (LEED)  1.05 (DFT) A˚ above the first Ru layer in the first phase, with a lateral distance to the densely packed Ru rows of 1.18  1.19 A˚ (/2). The O–Ru bond lengths are 2.09(1) A˚ and 2.10(2) A˚. In the second phase, oxygen sites 0.96 (LEED)  1.06 (DFT) A˚ above the first layer. All the three O–Ru bonds are identical in length, 2.03 A˚. The O–Ru distance (D/2) becomes 1.13 (LEED) –1.19 (DFT) A˚. Oxygen adsorption expands slightly the Ru(1010) first interlayer by  4%. 3.4.2. Analysis 3.4.2.1. Co(1010)-p(21)-O
1 : pairing-O
1 (< 0.5 ML). Fig. 15 shows the bond model for the STM image of the Co(1010)-p(21)-O precursor phase. The oxygen adsorbates are usually imaged with STM as depressions, even if the oxygen atoms are located above the surface, because the O-2p state is lower than the EF of a metal. Hence, the rectangular STM depressions can be identified as the pairing O
1–O
1

580

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Q dimers. The O
1–O
1 dimer rests atop of two Co atoms forming the ‘ ’ shaped Co2O2 bond along the close-packed direction. The O
1 catches one electron from the Co atom underneath and shares one electron with the other O
1. The O
1 polarizes its rest neighbors, which are responsible for the STM protrusions of this precursor phase. The second layer is not affected by the reaction at this stage, as there is no charge transport between the oxygen adsorbate and Co atom in the second layer. The dark or bright rectangular domains (5.04.0 A˚ in two-dimensions) are in the scale of the regular lattice. This result may be inferred by carefully scaling the STM images. The reaction at this stage may be formulated as follows: O2 þ 8Co ðsurfaceÞ ) 2O
1 þ 2Coþ þ 6Codipole or 2O2 þ 12Co ðsurfaceÞ  ) 2 2O
1 þ 2Coþ þ 4Codipole ; etc:

ð3:4:1Þ

Unlike the nanometric Cu(001)-c(22)-O
1 precursor state, in which oxygen forms an off-centered pyramid (4(O
1 + Cu+1) or 2O
1 + Cu+2) with its surface neighbors, the oxygen adsorbates in the Co(1010)-p(21)-O
1 phase prefer the atop atomic positions. In contrast, in the O-(Rh, Pd)(110) precursor states, O
1 occupies the C2v-hollow site and bonds to the atom underneath first. Obviously, the O
1 performs quite differently at these surfaces. This can be attributed to the differences in basic conditions as discussed in Section 2.1, i.e., the scale and geometry of the host lattice and the electronegativity of the host atom. 3.4.2.2. (Co, Ru)(1010)-c(24)-4O
2: hybridized-O
2 (0.5 ML). With increasing oxygen exposure, the O
1 evolves into O
2 and the oxide tetrahedron forms through a process of re-bonding, as shown in Fig. 16. The basic quasi-tetrahedron denoted as (1233) is expressed as (B=Ru, Co): O þ B ðsubstrateÞ þ 2B ðsurfaceÞ

þ ) O
2 ðhybridÞ þ Bþ ðsubstrateÞ þ 2Bdipole þ 2Bdipole In a c(24)-4O
2 unit cell, 2O2 þ 8B ðsurfaceÞ þ 8B ðsubstrateÞ h i

þ ) 4 O
2 ðhybridÞ þ Bþ ðsubstrateÞ þ 2Bdipole þB ðsubstrateÞ ;

ð3:4:2Þ

where the (2Bdipole)+ represents the fact that a virtual bond forms between the O
2 and the electron cloud of the two dipoles. The virtual bond is not a real one but it compensates for the lack of one atom in the oxide tetrahedron formation.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

581

The honeycomb-like or oval-shaped protrusions in the STM image are composed of four Codipole that are headed towards the center of the ‘oval’. In the Co2O tetrahedron, the lack of one Co atom for the bond is compensated by the polarized electron-cloud (as denoted 2) as do the (Rh, Pd)(110)-c(2n2)pmg-2O
2 phases. This process leads to a ‘virtual bond’ between the O
2 and the pairing dipoles, which sharpens the ‘tip’ of the ‘honeycomb’ protrusion, as can be observed from Fig. 16. As all the surface atoms are interlocked by the bond network, no atoms are missing there, being the same as that occurred in the O–(Pd, Rh)(110)-(2n2)-2O
2 surfaces. In addition, every other close-packed row in the second layer is composed of Co+. Although the STM images for the Ru(1010)-c(24)-4O
2 phase are lacking, the reconstruction pattern determined by LEED and DFT is the same as the Co(1010)-c(24)-4O
2 surface, as can be compared with models in Figs. 15 and 16. Therefore, the specification here should hold for both surfaces. The contribution from the electron clouds of the dipoles to the oxide tetrahedron formation in the (Co, Ru)(1010)-(24)-4O
2, the (Co, Ru)(1010)-(21)p2mg-2O
2 and the Cu(110)-c(62)-8O
2 phases evidences that the sp orbitals hybridization of an O
2 is independent of its bonding constituents. Oxygen can bond to any component whether it is an atom or electron-cloud of a single or a cluster of dipoles, which is able to supply electrons to fill the hybridized orbital of the oxide tetrahedron. It is even interesting that the minor geometrical difference between the (Pd, Rh)(110), (Co, Ru)(1010) and the (Cu, Ni, Ag, Pt)(110) causes entirely different orientations of the ‘O
2 : Mdipole : O
2’ chains. Contrary to the single O–Cu–O chain perpendicular to the close-packed direction of the (Cu, Ni, Ag, Pt)(110) surfaces, the pairing O–O zigzag chains in Fig. 16 run along the close-packed direction of Ru and Co. The fact that the c-axis lattice 4.06/4.33 A˚ of Co/Ru is too long for the distance of Bdipole–Bdipole and the a-axis lattice 2.51/2.68 A˚ may be too short for the ionic bonds forces the oxygen to create a new environment for the Co2O or Ru2O tetrahedron. This leads to the entirely different orientation of the O–(Co, Ru)–O chain and the shapes of the protrusions. As discussed, in the (Pd, Rh)(110)c(22)-2O
2 surface, oxygen adsorbates prefer to locating in the troughs beside a specific metal row, which yields the alternating ionic-row and buckled dipole-row observed as the zigzag O–M–O chain. However, oxygen atoms prefer positions, located in the (Co, Ru)(1010)-c(24)-4O
2 phase, between two nearest metal rows. The latter leads to the ‘honeycomb’ pairing-dipole rows on the (Co, Ru)(1010) surface rather than the ‘saw-tooth’ like single protruding rows on the (Rh, Pd)(110) surface. It is obvious that the lattice geometry determines the site specificity of the adsorbate, which results in the different reconstruction patterns. The variation of the reconstruction patterns originates from nothing more than the specific site of the adsorbate. 3.4.2.3. (Co, Ru)(1010)-(21)p2mg-2O
2: H-like bond dominates (1.0 ML). The (Co, Ru)(1010)-(21)p2mg-2O
2 phase, as shown in Fig. 17, interlocks with the c(24)-2O
2 phase by adding the zigzagged O–O chains at positions between the pairing dipole rows. The chemical reaction is formulated as:

582

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 17. (a) STM image of the Co(1010)-(21)p2mg-2O-2 surface [43]; (b) the hard-sphere model for Ru(1010)-(21)p2mg-2O-2 surface [63]; (c) the bond configuration [183] for the reconstruction. Shown in the middle are the lone pairs of the oxygen adsorbate atoms that squeeze the electron cloud of a Co or Ru atom, which leads to the corrugated morphology. The STM protrusions in this case correspond to dense polarized electrons (
) rather than ion core (+) positions. As the bond network interlocks all the surface atoms, no atoms are missing. The H-like bond formation restores significantly the decreased work function.

O2 þ 2BðsurfaceÞ þ 2BðsubstrateÞ  ) 2 O
2 ðhybridÞ þ Bþ ðsubstrateÞ þ Bþ=dipole

ð3:4:3Þ

In the present phase, each Co and Ru atom at surface has three O
2 neighbors due to the 1.0 ML coverage. Every Co or Ru atom at the surface becomes the ‘+/ dipole’. Therefore, H-like bonds dominate at the surface, which lowers the STM protrusions and narrows the anti-bonding band of the surface. It has been revealed using UPS63 that the oxygen-reduced work function of the Ru(1010)–O surface is

583

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

restored by 0.49 eV and 1.12 eV, respectively, as the c(24)-4O
2 phase changes to the (21)p2mg-2O
2 phase. The effect of H-like bonds becomes more apparent on the STM protrusions and the work-function recovery when the oxygen coverage increases from 0.5 ML to 1.0 ML. This can be compared using the STM images and the structural models in Figs. 16 and 17. The scale of the image in the latter reduces to a level similar to that for clean metals (0.15–0.3 A˚). All the Co atoms in the second layer become Co+ ions with reduced radii and lowered energy states. The interaction between the Co+ second layer and the third metallic Co layer is obviously stronger than it is between two Co metallic layers. Therefore, it is readily understood again that the first interlayer distance expands by an amount that depends on the bond geometry and the second interlayer spacing contracts driven by the enhanced Co+–Co interlayer interaction. This mechanism should hold for the Ru(1010)–O surface though LEED and DFT revealed a different amount of relaxation from that for the Co(1010)–O surface. However, we should note that the geometrical solution is not unique, and it is subject to the parameterization in data processing. Fig. 17 also shows how the non-bonding lone pairs of the two O
2 ions squeeze and deform the electron-cloud of one Co or Ru atom. This configuration accounts for the STM protrusions not only of the current Co(1010)-O
2 phases but also of the (Rh, Pd)(110)-O
2. At these surfaces, STM protrusions correspond to the dense electron-cloud (labeled ‘
’) rather than the ion core positions as labeled with ‘+’. As indicated, the slight protrusion in Fig. 17 locates between two ion cores. Therefore, it is not realistic to derive atomic structural information simply from the locations of the protrusions. Obviously, only the current bond model can explain consistently the dislocation of surface atoms and the alteration of atomic valencies. Comparing the reconstruction which has occurred to the (Co, Ru)(1010)(21)p2mg-2O
2 surfaces with that to the (Rh, Pd)(110)-(21)p2mg-2O
2 surfaces, it can be found that both systems have the same adsorbate arrangement. However, they yield entirely different patterns in the STM imaging. The former presents zigzag strips, while the latter gives a regular array of depressions and protrusions. One may attribute such a difference to the  values in Figs. 13 and 17: 

For the Rhð110Þ-ð21Þp2mg-2O
2 : D1 ¼ 3:81
D2 ¼ 3:81-21:13 ¼: 1:55A :     For the Co 101
0 -ð21Þp2mg-2O
2 : D1 ¼: 2:26A and D2 ¼ 4:06
D1 ¼:1:80A : An alternation of the D1 and D2 values and a small geometrical difference (0.25 A˚) determines the site-specificity of oxygen and gives quite a remarkable difference in the form of the STM images. Knowledge about the O–(Co, Ru)(1010) surface reaction has thus been developed in terms of bond forming. The disordered Co(1010)-p(21)-O precursor is identified Q as the derivative of the O
1. The ‘ ’-shaped O
1–O
1 dimer rests atop of surface Co atoms along the close-packed direction. Meanwhile, O
1 polarizes its rest neighbors leading to the protruding domain boundaries. At higher oxygen coverage of the (Ru, Co)(1010) surface, oxygen performs quite the same, as it does in the (Rh,

584

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Pd)(110) surfaces. O
2 locates at the fcc(111) facet site and forms a tetrahedron with two Co or Ru atoms at the surface and one Co or Ru atom in the second layer. The lacking of one atom for the tetrahedron formation is compensated by dragging the electron cloud of the dipoles. For both the Ru and Co, a c(24)-4O
2 phase forms and then a (21)p2mg-2O
2 phase follows. The H-like bond lowers substantially the STM protrusions and narrows the anti-bonding sub-band, which recovers the reduced work function significantly, as observed. 3.5. O–Rh(111) and O–Ru(0001) 3.5.1. Observations Both the O–Rh(111) and the O–Ru(0001) surfaces share considerable similarities in the patterns of O-induced reconstruction [194]. For instance, three phases of p(22)-O, c(22)-2O and p(11)-O form sequentially on these two surfaces. The Ru(0001)-p(11)-O phase can form under N2O pre-treatment [55,85] while the Rh(111)-p(11)-O phase can be obtained by low-energy oxygen-ion-beam irradiation [53]. Common to all other O-chemisorbed systems, the first layer spacing expands and the second contracts. In the p(22)-O precursor phase (0 0.25 ML), the R (=Ru, Rh) atoms buckle up radially away from (or towards, as disputed) the adsorbate and the original C3v symmetry of the unit cell remains. In the c(22)-2O phase (0 0.5 ML), pairing-row forms at the surface. In the p(11)-O phase (0=1.0 ML), the surface becomes unreconstructed. However, discrepancies yet remain regarding the vertical position and the site specificity of the adsorbate. DFT [55,85,194] and LEED optimisations [190,195,200,258,259] suggested that oxygen adsorbates are located  1.2 A˚ above the top layer of the surface and the oxygen site-specificity follows the rules of homo-epitaxial growth of the regular fcc(111) or hcp(0001) crystal lattices. The adsorbate prefers the apical site of a tetrahedron and remains identical bond length to the surface atoms throughout the course of reaction. The bond length, 2.00–2.10 A˚, equals approximately to the sum of the atomic radii of oxygen and R atoms (values are given in Table 2). A recent DFT calculation [260] on the incorporation of oxygen into the basal plane of the late 4d transition metals (TMs) of Ru, Rh, Pd to Ag suggested that occupation of subsurface sites is always connected with a significant distortion of the host lattice, rendering it initially less favorable than on-surface chemisorption, and that O always favors adsorption in the hollow sites, which represent a continuation of the bulk stacking sequence, i.e., hcp sites on Ru(0001) and fcc sites on Rh(111), Pd(111), and Ag(111). The combination on-surface O in fcc sites and subsurface O in tetra-I sites to be either most stable or energetically very close to the most stable geometry. The tetra-I site for Ru would only allow for an O-metal bond length of 1.65 A˚. This situation gradually becomes better for the other elements, yet for Ag, which has the largest lattice constant, this value is with 1.80 A˚ still significantly too short. Thus, subsurface O incorporation always induces a substantial local expansion of the metallic lattice. With increasing on-surface coverage, the repulsive interaction among the more densely packed adsorbates decreases the preference for on-surface adsorption; eventually O penetration may then become more favorable than a continued filling

585

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

of the on-surface sites. It is concluded that O incorporation into the subsurface region commences at progressively lower coverages for the late 4d TM sequence from Ru to Ag. In contrast, STM studies of the O–Ru(0001) surface by Meinel et al. [56] and Over et al. [262] implied that a sub-surface oxygen layer exists, and the bond nature of the different phases is entirely different. In a velocity-distribution spectroscopic study, Gibson et al. [193] suggested that both the sub-surface and on-surface oxygen coexist and they share common kinetics. Ganduglia and Scheffler [194] proposed that, at higher coverage, oxygen might occupy either the fcc(111) or the hcp(0001) hollow site of the Rh(111) surface due to the small difference in binding energy between these two sites. Table 11 gives information about the O–Ru(0001) and the O–Rh(111) surface reconstruction determined using LEED and DFT. In the DFT studies, Stampfl et al. [85] found that an oscillation takes place of both the work-function-change () and the dipole moment with increasing oxygen coverage onto the Ru(0001) surface. The dipole moment reaches its maximum at 0.5 ML oxygen coverage and then drops down swiftly. Madey et al. [263] observed that  decreases from its maximal 1.2 eV at 0.75 ML with increasing oxygen. In observing oxygen motion on the Ru(0001) surface with an STM, Renisch et al. [264] found it is essential to include the lateral interaction in modeling the collective phenomena such as surface diffusion or reaction. Table 11 Structural information for O–Rh(111) (Dbulk=2.20 A˚) and O–Ru(0001) surface reconstruction (units in A˚) O-position (Ref.) Rh(111)-fcc (21) (LEED) [190] (22) (21) (11)(DFT) [194] (22) (21) (11) (LEED) [54] Ru(0001)-hcp (21) (LEED) [200] (22) (LEED) [82] (22) (21) (11) (DFT/LEED) [55,85] (21) (0.6 ML) (11) (0.9 ML) (11) (1.6 ML) MEIS [261]


DOz

DOs D12

D23

Bond length D//

1.22 1.24 1.18 1.16 1.24 1.23 1.16

0.05 2.23 2.11 0.03 2.19 2.37 2.25 0.03 2.24 2.37

2.21 2.21 2.19 2.25 2.19 2.19 2.25

1.92; 2.01

1.25

2.00 1.95

0.03 radial
0.07;
0.01 Pairing-row 0.02 radial -0.06; -0.01 Pairing-chain

2.00 1.95 2.02

1.220.02 1.3 1.26 1.25

2.26 2.10 2.14 2.22(+2.7%)

2.25 2.03 0.06 2.16 2.03 2.18 2.13(
0.9%)

1.23 1.19 1.16

2.14 2.17 2.22

2.15 2.11 2.10

Pairing-row 0.12 0.09 Radial Pairing Buckle-in

Z-axis is directed into the bulk. DOs stands for the lateral dislocation of oxygen. D// is the lateral dislocation of R atoms.

586

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 18 presents the STM images for the p(22)-O and c(22)-2O phases at the Ru(0001) surface. [56,199]. Fig. 19 also gives the reconstruction models for the O–Rh(111) [54,194] and O–Ru(0001) surfaces [55,85] determined by LEED and DFT. The on-surface oxygen adsorbate derives the ‘radial’ and ‘pairing-row’ structure patterns at the surfaces. Similar STM patterns have been observed from O–Au(111) [265] and Ag(111) [266] surface after prolonged annealing in oxygen (1bar) at 800  C. These observations form the up-to-date knowledge about the O-hcp(0001) and the O-fcc(111) surface reconstruction. 3.5.2. Analysis It is known that the electronic configuration for Ru is 4d75s1and for Rh it is 4d85s1. The electronegativity of both Ru and Rh is of the same value of 2.2. Their atomic radii [a=2.672(Ru) and 2.684(Rh) A˚] are similar. As shown in Fig. 1, the top two

Fig. 18. STM images of Ru(0001) surface with chemisorbed oxygen at 400 K. The images correspond to (a) 0.20, (b) 0.25; and (c) 0.5 ML oxygen exposures, respectively. The p(22)-O (radial) and the p(21)-O (pairing row) structures are fully developed at 0.25 and 0.5 ML [56].

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

587

Fig. 19. R4O cluster bonding [189,197] and the reconstruction models for the O–Rh(111) [54,194] and O– Ru(0001) surfaces of C3v symmetry [55,85]. (a) At YO 41/4 ML coverage, O
1 locates in the center of a tetrahedron and forms one bond with atom labeled 1. The O
1 induces and pushes the dipoles labeled 2 radially away and the C3v symmetry remains, producing the clusters of STM protrusion. (b) At YO 41/2 ML, O
1 evolves into O
2 by forming one more bond with a surface R atom labeled 1. Dipoles (labeled 2) are sustained then by the lone pairs. The dipole and ionic row move closer towards the sites without adsorbates, which generates the pairing STM protrusion-depression patterns. (c) At YO 41.0 ML, H-like bonds dominate at the surface. Two lone pairs polarize each surface R atom (1/2) which donates meanwhile one electron to the adsorbate. Formation of the H-like bonds restores the reduced work function and lowers the STM protrusions. The entire surface network is stabilized and becomes unreconstructed in crystallography.

588

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

layers of both hcp(0001) and fcc(111) surfaces share the same lattice geometry, packing in an AB order. The tetrahedral hcp(0001) site is more favorable than the fcc(111) hollow site to facilitate a R4O tetrahedron according to the current bond model. The nature and kinetics of the R4O cluster bonding and its consequences on the atomic valency and surface morphology for the two systems can be formulated as follows. 3.5.2.1. O
1 effect (YO=0.25 ML): radial reconstruction with C3v symmetry. For the p(22)-O-1 precursor phase, the reaction can be expressed as: O þ 4R ð1st layerÞ þ 4R ð2nd layerÞ ) O
1 ðsub-surfaceÞ þ Rþ ð2nd layerÞ

þ 3Rdipole buckled away of the O
1

ð3:5:1Þ

þ R ð1st layerÞ þ 3R ð2nd layerÞ Instead of locating above the top layer, oxygen atom tends to sink into the center of one of the four tetrahedral sites to form an R4O cluster [Fig. 19(a)]. Oxygen forms one bond with the R (labeled 1) underneath. The O
1 then polarizes and pushes its three surface neighbors (labeled 2) radially outwards. Therefore, the C3v symmetry remains. The clustered dipoles are responsible for the STM protrusions. At the surface, there is still a metal atom in a unit cell as it can be seen. 3.5.2.2. O
2 effect (YO=0.5 ML): paring row reconstruction. The p(21)-O
2 or c(22)-2O
2 phase can be decomposed as: O2 þ 4R ð1st layerÞ þ 4R ð2nd layerÞ  ) 2O
2 ðhybridÞ þ 2Rþ ð1st layerÞ þ 4Rdipole ð1st layerÞ ðpairing chainÞ þ 2Rþ ð2nd layerÞ þ 2R ð2nd layerÞ At this stage, oxygen forms the second bond with one (labeled 1) of its three surface neighbors and then the sp orbitals of oxygen start to hybridize [Fig. 19(b)]. The number of dipoles of the same R4O tetrahedron reduces from three to two. The lone pairs of oxygen replace the role of O
1 in sustaining the Rdipole (labeled 2). On the surface, the atoms with altered valencies form the alternative rows of protrusion (Rdipole) and depression (R+), and the original C3v symmetry of the tetrahedron breaks, as predicted by the DFT and LEED optimizations and detected with STM imaging. The valence of the R+ underneath the O
2 remains. The mechanism for the surface relaxation is common to the situations addressed in previous sections, that is, the bond geometry determines the first layer expansion and the altered atomic valency in the second layer shortens the second interlayer spacing.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

589

3.5.2.3. O
2 effect (YO=1.0 ML): H-like bond dominates. The p(11)-O-2 phase on the C3v surface can be formulated as: 2O2 þ 4R ð1st layerÞ þ 4R ð2nd layerÞ ) 4O
2 ðhybridÞ þ 4Rþ=dipole ðsurfaceÞ þ 4Rþ ð2nd layerÞ

ð3:5:3Þ

As shown in Fig. 19(c), oxygen adsorbates have occupied all the tetrahedral sites on the surface. Each adsorbate needs one R atom for the bond and two to be polarized at the surface. Because the atomic ratio O : R=1, each of the surface R atoms has to interact with three oxygen neighbors through one ionic bond and two lone pairs. Therefore, all the R+ and Rdipole at the surface turn to be R+/dipole (labeled 1/2) and hence H-like bonds dominate at the surface. In such a way, the surface-dipole moment and the dipole-related  are weakened substantially due to the H-like bond formation, and the surface becomes unreconstructed in crystallography, agreeing with the DFT calculations and work function measurements, as mentioned above. This trend is the same as that occurring to the (Co, Ru)(1010)O
2 surfaces and the (Rh, Pd)(110)-O
2 surfaces at higher oxygen coverage. The analysis strongly supports the model of sub-surface oxygen formation though most of the numerical optimizations favor an on-surface oxygen mechanism (see Table 11) for oxygen chemisorption on these C3v surfaces. One cannot however exclude the possibility of multiple numerical solutions due to the correlation among the parameters used in optimization and the initial conditions being taken, as discussed earlier. Nevertheless, existing grounds exclude the possibility that three identical bonds form and remain unchanged throughout the course of reaction. If the oxygen is located above the top layer and if the oxygen bonds to its three surface neighbors identically, one could explain neither the interlayer relaxation nor the STM protrusions. The on-surface oxygen mechanism should shrink the D12 instead, as the R+ reduces its size considerably and the R+ produces STM depressions under normal tip conditions. Actually, the adsorbate, whether it is O
1 or O
2, is retained always at the nearly central position of the oxide tetrahedron throughout the course of reaction with the C3v surfaces. The striking significance of the precursor is that the O
1 polarizes all of its surface neighbors that maintain the original C3v symmetry. O
2 forms two ionic bonds with the host atom in the p(21)-O-2 phase or electron clouds in the p(11)-O
2 phase, which breaks the C3v symmetry. The O
2 polarizes its two atomic neighbors through the lone pair interaction. The O
2 should be slightly off-centered inside the tetrahedron by an amount, which might be too small to be detectable. The oxidation alters the valencies of O into the O
1 and O
2, and R atoms into R+, Rdipole and R+/dipole with the measurable variation of  and the surface dipole moment. The real process of oxidation is thus suggested to be that electron transport dominates and oxygen adsorbates reside inside the bulk rather than float atop the surface. This is the case for thermal oxidation of the diamond {111} plane, which has been confirmed to go through the {111} channels throughout the course of oxidation (Section 9.1) [267]. In summary, the bond model suggests that oxygen adsorbates sink into and remain at the hcp(0001) hollow sites throughout the course of reaction with the C3v

590

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

surfaces. The rule of homo-epitaxial growth of the fcc(111) or the hcp(0001) surface seems unlikely. During the process of reaction, the R4O cluster configuration remains the same but the valencies of atoms at the surface change continually and substantially. The O
1 reconstructs the surface in a radial mode and then the O
2 produces the pairing-row pattern. Overdosing with oxygen (Y 50.5 ML) yields the p(11)-O
2 phase in which H-like bonds are dominant, which stabilize the surface. The H-like bond formation interlocks all the surface atoms, which may form a barrier for surface diffusion [268]. This conclusion should be valid for other fcc(111) and hcp(0001) surfaces with chemisorbed oxygen. For instance, a first principle approximation of the well-defined O–Al(111) interaction [75] indicates that the adsorbate atoms prefer the hcp tetrahedral sites, 1.92 A˚ below the topmost Al layer which has relaxed by 25%. A DFT calculation of Ganduglia et al. [269] suggests that oxygen switches from the on-surface fcc site to the subsurface hcp sites of the Rh(111) plane, and indicates that at even higher coverages oxygen incorporation is followed by oxygen agglomeration in two-dimensional sub-surface islands directly below the first metal layer. Inside these islands, the metastable hcp/octahedral (onsurface/sub-surface) site combination will undergo a barrierless displacement, introducing a stacking fault of the first metal layer with respect to the underlying substrate and leading to a stable fcc/tetrahedral site occupation. The subsurface oxygen atoms in tetrahedral sites are fourfold coordinated to metal atoms. It has been suggested that these elementary steps, namely, oxygen incorporation, aggregation into sub-surface islands and destabilization of the metal surface may be more general and precede the formation of a surface oxide at close-packed transition metal surfaces. A DFT calculation by Reuter et al. [270] predicted that the oxidation of the Ru(0001) surface proceeds via the accumulation of subsurface oxygen in twodimensional islands between the first and second substrate layers. This leads locally to a decoupling of an O–Ru–O trilayer from the underlying metal. Continued oxidation results in the formation and stacking of these trilayers, which unfold into the RuO2(110) rutile structure once a critical film thickness is exceeded. Along this oxidation pathway, they identified various metastable configurations in which oxygen occupies the octahedral and tetrahedral sites, respectively. These configurations are found to be rather close in energy, indicating a likely lively dynamics between them at elevated temperatures. 3.6. O–Rh(001) and (N, C)–Ni(001) 3.6.1. Observations With increasing oxygen exposure of the Rh(001) surface, three outstanding phases form sequentially, as identified using LEED, [157,164,271], STM, SPA-LEED and PES. [52,166]. The disordered p(22)-O phase (Y=1/4 ML) forms first and then is followed by the c(22)-2O (Y=1/2 ML) radial reconstruction and, finally, the c(22)p4g-2O (Y=1/2 ML) clockwise-and-anticlockwise rotation of the unit cells, or clock reconstruction. These reconstruction patterns are the same as those occurring on the Ni(001) surfaces with chemisorbed carbon and nitrogen. Baraldi et al. [161– 163] suggested that this sequence of phase transitions belongs to the Ising uni-

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

591

versality, or order–disorder transition. Fig. 20 shows the STM images of the two ordered phases for the O–Rh(001) surface. The oxygen adsorbate fills in the next nearest fourfold hollow-site of the Rh(001) surface, in the same way as does oxygen on the Cu(001) surface. It can be seen that both the c(22)-2O and the c(22)p4g-2O

Fig. 20. STM images [52] and the corresponding bond models [167] for (a) the Rh(001)-c(22)-2O
1 radial and (b) the c(22)p4g-2O
2 phases. During the reaction, a Rh5O pyramid evolves into the Rh4O tetrahedron defining half of the surface atoms to be Rhdipole (labeled 2) and another half Rh+/dipole (labeled 1/2). The electrostatic forces create rhombi-chains along the <11> directions. Here, (c) the ‘centered pyramid’ [272] and (d) the ‘off-centered rhombus’ [254] models are compared.

592

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

phases require the same local oxygen coverage of 1/2 ML. It was noted [52] that formation of the two ordered phases depends not on the local oxygen coverage but rather the overall oxygen exposure. Direct LEED studies [272], revealed, however, that oxygen resides in hollow sites of the Ni(001) surface; the first substrate layer distance is expanded and the second substrate layer reconstructs into a buckled layer. It is suggested that [273] bond form between oxygen and the second-layer nickel atom below it. No apparent rotation was identified [274]. Fig. 21 shows the STM images of the ordered Ni(001)-(22)p4g-2C [287] and Ni(001)-(22)p4g-2N [286] surfaces that exhibit the same type of ‘clock’ reconstruction as that occurring on the O–Rh(001) surface albeit the slight difference in the rotation angles. However, the calculated STM image shows the C atoms as depressions while in the experiment, a small protrusion is suggested [275]. It is interesting to note that the Ni(100)-(22)p4g-2N STM image exhibits two orientations of the depressions. With slight deviation, one is along the [10] direction and the

Fig. 21. STM images [286,287] and the corresponding atomic and bond model (insertions) for Ni(001)(22)p4g-(C
4, N
3) clock reconstruction. Individual atomic valencies and the orientation of the depressions are indicated. Although they appeared quite similar to that of the Rh(001)-(22)p4g-O
2 phase, the surface atomic valencies and the driving force are different.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

593

other closes to the [01] direction. This provides the opportunity for one to identify the atomic valencies at the surface. In 1979, Onuferko et al. [276] proposed a centered-pyramid model [Fig. 20(c)] for the ‘p4g’ reconstruction. The model indicates that a ‘stress release’ mechanism provides forces driving the ‘radial’ reconstruction phase to transit into the ‘clock’ rotation. The oxygen adsorbate was determined to locate DOz > 0.6 A˚156 above the C4v hollow site and form a centered-pyramid with the four Rh atoms at the surface. It was interpreted that the compressive stress of the pyramid in the Rh(001)-c(22)-2O surface is released through the rotation of the Rh4O pyramid. The release of stress provides the forces that drive the reconstruction while the basic R4O pyramid remains during the process of reconstruction despite a slight change of the vertical position of the oxygen. This model applied well to all the (N, C)–Ni(001) and the O–Rh(001) surface reconstructions. In 1998, Alfe et al. [160,161,277] suggested an alternative (Fig. 20d) for the Rh(001)-c(22)p4g-2O second phase. Instead of forming a pyramid with the four surface neighbors in the C4v hollow site, the adsorbate was assumed to prefer the site eccentrically above the ‘rhombus’. It is suggested that the adsorbate jumps back and forth along the longer axis of the rhombus. Norris et al. [278] derived from XRD measurement that, in the second ‘p4g’ phase, the Rh atom is displaced by 0.19  0.02 A˚ in the plane along the < 11 > direction. The oxygen atom is situated in the rhombus sites with an in-plane shift of 0.20  0.05 A˚ on either side of the center. The ‘off-centered-rhombus’ model leads to the conclusion [161] that for all the Rh{(001), (110), (111)} surfaces at medium-high oxygen coverage, oxygen adsorbate tends to locates at an apical site of a tetrahedron and to form three identical bonds with the Rh atoms. So formed Rh3O tetrahedron was taken to be a C3v point-group symmetry and the lengths of all the three O–Rh bonds in the Rh{(001), (110), (111)} surfaces to be in the range of 2.00–2.06 A˚. The Rh–O–Rh bond angle was suggested to be around 90 . DFT calculations suggested that the ‘off-centered rhombus’ model is less favorable than the ‘centered-pyramid’ model for the (C, N)–Ni(001) surfaces [277]. Table 12 summarizes the geometrical information for the ‘p4g’ reconstruction for the Rh(001)-c(22)p4g-2O and the Ni(001)-c(22)p4g-2(C, N) surfaces. Kirsch and Harris [279] calculated surface reconstructions of Ni(001) surface induced by C, N and O adsorption and suggested that C and N atoms prefer the nearly coplanar sites with the top Ni surface and induce the ‘‘clock’’ reconstruction of the surface while O atoms prefer sites slightly above the Ni(100) surface plane and have little effect on the overall surface structure. The local environments of the C, N, and O atoms on these surfaces are similar to their environments in a series of late transition metal carbonyl clusters, suggesting that some of the same electronic factors may play a role in favouring the different structures. Results of the calculations suggest that when adsorbates occupy coplanar sites on Ni(100), much of the Ni–Ni bonding within the surface layer and between the surface- and second-layers is disrupted. On the C- and N-covered surfaces the disruption is more than compensated for by the formation of strong adsorbate–Ni bonds and by new Ni–Ni surface bonds resulting from the clock reconstruction. When O is forced into a coplanar site, however, both the higher electron count and increased electronegativity of the O

594

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Table 12 Atomic geometry of the Rh(001)-c(22)p4g-2O surface (D12=1.902 A˚; z is directed into the bulk) and the Ni(001)-c(22)p4g-2(C, N) surfaces Model [Ref.]

DOz

DOs

D <11 >

Off-centered rhombus model [254] for O–Rh(001)

O-Rh(001) (LEED) [161] (DFT) [160] DFT [277] SXRD [278]


1.01 0.05 0.290.15 0.2 0.07

Centered pyramid model [276] for all the ‘p4g’

N-Ni(001) (IES) [156] SEAFS [280,281] PED [282] SXRD [283] C–Ni(001), PED [282] LEED [276] SEXAFS [284] LEED [285]


0.6 0.1

D12

D23

1.940.05

1.870.03


1.02 0.08 0.20 1.99
1.00 0.04 0.21 0.35 1.922
0.64 0.06 0.200.05 0.19 0.02 1.960.05 0.2 0.1 0.77 0.10 0.55 0.20 0.30 0.01 0.55 0.20 0.35 0.05


0.25 0.05
0.20 0.10
0.30 0.05
0.30 0.12
0.2 0. 2
0.31 0.05

Rhombus chain Rh2O [100] +STM [52] > 0 model [100] for Ni3N [100] + STM [286] > 0 all the ‘p4g’ Ni4C [100] +STM [287] > 0

+0.150.10 +0.170.01 +0.150.10 +0.200.05

0.45 0.07 +0.190.04 >0 >0 >0

0.3 0.37 0.64

atoms lead to severe disruption of the surface bonding and weak Ni–O bonds. When O atoms sit above the surface, they form more polar Ni–O bonds, contribute less electron density to the Ni surface bands, and cause less disruption to Ni–Ni surface bonds. These results suggest that, similar to the organometallic clusters, the site preferences of C, N, and O atoms are directly related to their electron count, and in turn to the relative occupation of both Ni–Ni and X–Ni (X=C, N, O) antibonding bands. 3.6.2. Analysis We now show that the O–Rh(001) radial and the subsequent clock reconstruction results from the formation of the oxide tetrahedron, in which the O
1 transits to an O
2 with sp-orbital hybridization. The sp-orbital hybridization also holds for the C and N when they react with the Ni(001) surface. It is derived that the electrostatic force due to charge redistribution drives the p4g reconstruction [100,288,289]. The balance of bond tension against the electrostatic force along the < 11> direction stabilizes the clock rotation on the O–Rh(001) and N–Ni(001) surfaces. However, equilibrium of electrostatic repulsion along the < 11> direction and a response of bond compression stabilize the C–Ni(001) clock rotation. Although the patterns of reconstruction and morphology for the O–Rh(001) and the (C, N)–Ni(001) surfaces are nearly the same, the surface atomic valencies and the driving forces are different due to different valencies of the C
4, N
3 and O
2. The O–Rh bond experiences tension rather than compression.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

595

With the bond model, the Rh(001)-c(22)-2O
1 radial and the subsequent Rh(001)-(22)p4g-2O
2 clock reconstruction can be formulated as follows [Fig. 20(a and b)]: O2 ðadsorbateÞ þ 4Rh ðsurfaceÞ þ 4Rh ð2nd layerÞ ) 2O
1 ðsub-surfaceÞ þ 2Rh þ ð2nd layerÞ

þ 2Rh ð2nd layerÞ þ 4Rhdipole O
1 -induced

ð3:6:1Þ

this O
1-induced phase transits into the Rh(001)-(22)p4g-2O
2 second phase (see the corresponding deformed unit cell in Fig. 20b): ) 2O
2 ðhybridÞ þ 2Rhþ ð2nd layerÞ þ 2Rhþ=dipole

þ 2Rh ð2nd layerÞ þ 2Rhdipole O
2 -induced

ð3:6:2Þ

Similarly, the Ni(001)-(22)p4g-2(N
3, C
4) surfaces can be formulated as (Fig. 21) [100]: N2 ðadsorbateÞ þ 4Ni ðsurfaceÞ þ 4Ni ð2nd layerÞ ) 2N
3 ðhybridÞ þ 2Niþ ðsurfaceÞ þ 2Niþ=dipole ðsurfaceÞ þ 2Niþ ð2nd layerÞ þ 2Nið2nd layerÞ

ð3:6:3Þ

and, C2 ðadsorbateÞ þ 4Ni ðsurfaceÞ þ 4Ni ð2nd layerÞ ) 2C
4 ðhybridÞ þ 2Niþ ðsurfaceÞ þ 2Ni2þ ðsurfaceÞ þ 2Niþ ð2nd layerÞ þ 2Ni ð2nd layerÞ

ð3:6:4Þ

It is to be noted that from the rhombi-chain formation point of view, the p4g STM images induced by C
4, N
3 and O
2 are substantially the same despite the rotation angles. However, Ni(001)-(22)p4g-N
3 STM image shows two apparent orientations of the alternative thick lines of depressions, with slight deviation from the [10] direction and the [01] direction. Linking the Ni+ ions labeled 1 in N–Ni(001) surface [see Fig. 21(b)], one can find that the 1–1 bridge matches ideally the orientations of the STM depressions. This verifies that the radius of a Ni+ ion is much smaller than that of a Ni+/dipole. Therefore, applying the model of bonding to the STM images enables the individual atomic valencies of the surface Ni atoms to be identified. Table 13 summarizes information for the C
4, N
3 and O
2 derived ‘p4g’ phases on the Ni(001) and Rh(001) surfaces. For instance, oxygen sinks into the C4v hollow site in the Rh(001) surface and bonds to the Rh atom underneath. An Rh5O cluster forms in the c(22)-2O
1 precursor phase. The O
1 polarizes and pushes the electron cloud of the surface atoms radially away from the central adsorbate. In the second phase, the c(22) cell deforms into two rhombi (without adsorbate) and two

596

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

squares with adsorbates inside, as can be seen from Fig. 20(b). The Rh5O pyramid will evolve into an Rh4O tetrahedron with an off-centered shift of the adsorbate in the hollow and one Rh atom at surface was released from the original Rh5O pyramid. As the O
2 has already bonded to one Rh underneath, the tetrahedron defines one Rh+ (labeled 1) and two lone-pair-induced Rhdipole dipoles (labeled 2) of the four nearest surface neighbors. The Rh5O ! Rh4O transition gives rise to the overall ‘p4g’ reconstruction. As can be seen from the primary unit cell containing the adsorbate [Fig. 20(b)], three of the four surface neighbors are labeled with 1, 2 and 2, respectively. Because the surface atomic ratio O : Rh=1 : 2 and each oxygen bonds to one atom at surface and needs two atoms to be polarized, half of the overall surface atoms are thus defined as Rhdipole and another half as Rh+/dipole. The Rh+/dipole contributes to the H-like bond. One can find that atoms labeled 1 change their positions in a clockwise fashion if one counts the O-occupied hollows along the < 11> direction (gray thick lines). The adsorbate dislocates eccentrically in direction in a periodic p way. p From this point of view, it would be essential and complete to consider a c(4 24 2)R45 -16O
2 complex unit cell in practice due to the periodicity of the off-centered adsorbate positions. Strikingly, the ‘rhombi’ hollows without adsorbates form chains along the < 11 > direction. The above argument is also applicable to the (N
3, C
4)-Ni(001) surfaces. From the atomic structure point of view, the current description favors both the existing models to a certain extent. Oxygen prefers the site inside (sub-surface) the C4v hollow (Onuferko’s model) and eccentrically (Alfe’s model) in a periodic way (detected as oscillation). It is understandable that periodicity can be detected as oscillation. Most importantly, specification of the valencies of the adsorbates gives great detail about the electronic structures, driving forces and bond stresses, of the Rh(001)–O and the Ni(001)–(C, N) surfaces [100]. Although the STM and LEED signatures appeared the same for these three ‘p4g’ reconstructed surfaces derived by C, N and O, the underlying mechanisms are quite different. Therefore, the reaction is a process in which charge transportation dominates. What one often observes with STM and LEED is simply a portion of the consequences. For more details regarding the sp-orbital hybrid bonding of C, N, and O to the fcc(001) surfaces of Rh and Ni please refer to a recent report [100].

Table 13 Summary of the geometrical change, driving force, Fi, and bond tension, T, of the ‘p4g’ clock rotation (2) derived from the STM images with the current model [100] (22)p4g

Rh(100)-O-2

Ni(100)-N-3

Ni(100)-C-4

Rotation angle F ( ) S <11 > (A˚) Bond strain L/L (%) Electrostatic forces Fi Bond stress Fb (dyn) Surface atomic states

0.0; 9.0 0.00; 0.30 1.2 7.99; 11.09 35.16 Dipole; +/dipole

0.0; 12.0 0.00; 0.37 2.2 8.21; 14.59 35.95 +; +/dipole

0.0; 20.0 0.00; 0.64 6.4
3.11;
91.17
133.00 +; 2+

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

597

3.6.3. Quantification: driving force and bond stress One can estimate the lateral displacements of the Rh atoms by measuring the sharp angle  of a rhombus in the STM image [Fig. 20(b)]. The average value of  is measured as 72 . The unit cell containing O
2 adsorbate rotates, hence, 2=(90
)/ 2=9 , which displaces the Rh atom along the < 11 > direction by p S <11 > = 2Rtg2=0.30 A˚, where R=1.342 A˚ is the atomic radius of Rh. Table 13 summarizes the geometrical change, driving force, Fi, and bond stress, T, of the ‘p4g’ clock rotation (2) derived from the C
4, N
3 and O
2 induced ‘p4g’ reconstruction on the fcc(001) surface of Rh and Ni [100]. Inspecting Figs. 20(b) and 21 one can find that the surface network is composed of one-dimensional ‘- 2 - (1/2) – (1/2) - 2 – 2 - (1/2) – (1/2)-’ rhombi-chains along the < 11> directions. Label 1 and 2 represents the valencies of ‘+’ and ‘dipole’, respectively. The electrostatic charges of the Rh+/dipole (1/2) and the Rhdipole (2) are not equal and the Rh+/dipole is slightly positive compared to the Rhdipole that has a negative feature. The intensity of interaction is in such an order: 2 - 2  (1/2) - (1/2) > 0 > (1/2) - 2. The repulsion between the 2 - 2 or (1/2) - (1/2) and the slight attraction between the 2 - (1/2) determines that the distance of 2 - 2 or (1/2) - (1/2) is longer than the distance of (1/2) - 2. Rhombus formation displaces the Rh atoms along the < 11> direction, and consequently, leads to the overall rhombi-chain network at the surface. The alternate attraction and repulsion along the chain will squeeze the 2-(1/2) closer without otherwise a response of bond tension to equilibrate the electrostatic force along the chain. From Fig. 22(a and b), it is seen that the bond expands by an amount L/L=1/cosF
1=1.2%, which is negligibly small and, therefore, a mechanism of bond-tension-increase rather than the mechanism of bond-compression-release dominates in the Rh(001)p4g-(22)-2O
2 or the Rh(001)p p 
2 (4 24 2)R45 -16O phase transition. Without knowing the exact dipole moment, one may assume that Coulomb interaction dominates along the rhombi-chain that contains infinite number of atoms (n 5  100 is sufficient for calculation). The Coulomb potential Vi and the electrostatic force Fi acting on the ith atom in the rhombi-chain are:   1 X qj 1 X 1 1 j
Vi ¼ ¼ ð
1Þ 4"0 i6¼j rij 4"0 i6¼j ð2j
1Þa
2s ð2j
1Þa þ 2s Fi ¼
qi

@Vi @r

ð3:6:6Þ

The relative charges for Rhdipole and Rh+/dipole are defined as qi=
"e and "e, respectively, by introducing an effective charge factor "="c + ' ("i
"c)/2 which takes the valence screening effect into account. "c=0.5 ('=0) and "i=1.0 ('=2) correspond to covalent and ionic states, respectively. For the current O–Rh system, 
=3.5
2.2=1.3, "=0.825. As a component of the electrostatic force Fi, the Fb in Fig. 22(b) balances the bond tension T and hence the clock rotation of the unit cell:

598

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 22. (a) The non-uniform ‘- 2 - 1/2 – 1/2 - 2 -’ rhombi-chain is extracted from Figs. 20(b) and 21 to estimate (b) the atomic dislocation and (c) the driving force and bond stress. Labels 2 and 1/2 stand for different valencies (refer to Table 13) for O
2, N
3 and C
4 induced chains. The clock rotation angle F is p derived from the STM rhombus sharp-angle Y. F=(90
Y)/2; S= 2R tgF; L/L=1/cosF
1; Fi is decomposed as Fb. (c) Shows the rotation angle dependence of the driving force Fi. Rotation is stabilized at 9 by T=Fb. If F > 9 , T > Fb; otherwise, T
Fi Fb ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2½1 þ cosð2ð90
FÞÞ

ð3:6:7Þ

When the Rh4O tetrahedron rotates from 0 to 9 (see Table 13), S <11 > shifts from 0 to 0.3 A˚ and the electrostatic force Fi increases from 8 to 11 dyn. The tensile bond stress, T, increases to 35 dyn at F=9 . It is easy to understand that, at F < 9 , T < Fb, while at F > 9 , T > Fb. Therefore, the coupling of the alternative electrostatic attraction and repulsion along the rhombus chain with the response of bond tension stabilizes the rotation. The bond tension might be over-estimated because the dipole potential should take the r
6 ij form rather than the simple Coulomb potential. However, this can be precisely determined provided there is a known dipole moment. Nevertheless, we have justified that the driving force behind the Rh(001)-O
2 clock rotation comes from the electrostatic interaction along the rhombi-chain and further rotation of the tetrahedron is constrained by the response of a bond tension. This bond-tensionincrease mechanism also holds for the Ni(001)-N
3 surface. The Ni(001)-C
4 rota-

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

599

tion is driven by the nonequivalent electrostatic repulsion in the < 11 > direction and the rotation is balanced by the response of bond-compression-increase, instead [100]. 3.7. O–(Ag, V)(001) 3.7.1. O–Ag(001) 3.7.1.1. Observations. Using a combination of LEED and HREELS, Fang [290] found that oxygen could induce two different phases on the Ag(001) surface. At low temperature, the system displays a c(22) LEED pattern with an EELS peak at 37 meV. By increasing the temperature from 180 to 300 K, a transition to a (11) LEED pattern happens, and the EELS peak shifts from 37 meV to a lower value of 30 meV. The EELS peak shift indicates that the stretch interaction of oxygenadsorbate is weakened. This phase transition was found to be reversible. It is interpreted that the high-temperature (11) structure correspond to full oxygen coverage (0=1 ML), while the low-temperature c(22) structure arise from a lower coverage (0=0.5 ML), with a substantial amount of oxygen adsorbed in subsurface. Recently, Rocca et al. [291] reported a more complete investigation of O–Ag(001) surface, using the techniques of LEED, HREELS, XPS, and XPD. Measurements confirmed the existence of the phase transition observed by Fang, but leads to a different structural model. The estimated oxygen coverage is never greater than 0.4 ML. By fitting the surface geometry to high-temperature XPD results, it was predicted that atomic oxygen would sit in the fourfold hollow site, while low temp at p perature the fit of XPD data suggests that the substrate undergoes a ( 22 2)R45 missing-row reconstruction, being the same to the O-Cu(001) second phase with the Cu3O2 pairing-tetrahedron configuration. In the second phase, oxygen atoms sit in the hollow sites near the Ag missing rows of the substrate, thus giving rise to a c(22) LEED pattern. A first principle DFT calculation by Cipriani et al. [292] examined the phase stability at different oxygen coverage. It was found that the missing-row reconstruction is almost degenerate in energy with the non-reconstructed c(22) structure, both of which share the same LEED pattern. The geometrical details of the DFT calculation structure slightly differ from those inferred from XPD measurements [290]. Both the local density approximation (LDA) and the generalized gradient approximation (GGA) predicted that the surface fourfoldcoordinated hollow site and the sub-surface fivefold-coordinated hollow site are most stable, practically degenerated, and separated by an energy barrier of  25 meV or less. It is striking to note that the first Ag interlayer spacing expands by up to 30% upon oxygen adsorption [292]. According to the XPD results [291], oxygen adsorbates displace laterally by 0.36 A˚ toward the missing row and shift vertically by 0.15 A˚ down below the top Ag atoms. The Ag atoms close to the missing row are shifted up by 0.3 A˚ and laterally by a negligible amount. In the calculated structure, oxygen atoms sit  0.3 A˚ above the Ag atoms that moves  0.2 A˚ close to the missing row and  0.1 A˚ below the other surface atoms. Oxygen atoms do not relax laterally. However, the origin of the structural discrepancy is unclear. Nonetheless,

600

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

p p the overall missing row structure coincides with the Cu(001)-( 22 2)R45 -2O
2 phase which is dominated by the Cu3O2 structure where oxygen atoms move away from the missing row and go below the Cu atoms near the missing row for the tetrahedron formation. The Cu (dipoles) are shifted up and relax laterally towards the missing row by  0.25 A˚ to form the ‘dumb bell’ protrusion cross over the missing row. As discussed previously, accurate determination of the static position of atoms is not practical as the reaction is a kinetic process. In addition, errors are often involved in measurement and approximations in theoretical determination as well. The significant difference between the O–Cu(001) and the O–Ag(001) is the critical temperature for phase transition. Missing row forms at the O–Cu(001) surface at  700 K while it forms at the O–Ag(001) surface below 300 K. XPS profiles [291] from O–Ag(001) show that the O 1s lower binding energy (528.3 eV) component increases its intensity with temperature rise at an expenses of decreasing the high binding energy (530.9 eV) peak that dominates at T4300 K. This is also in line with the 37!30 meV EELS peak transition. The transition of both the XPS binding energy and the EELS stretch energy indicates that the O–Ag(001) surface undergoes a phase transition from stable to quasi-stable at  300 K or lower with temperature rise. At 246 K, the stable phase forms gradually with aging time. The stable phase formation completes after a 3200 s aging at 246 K. With oxygen adsorption, an additional DOS feature at
2.0 to
3.0 eV emerges [291,293] which agrees with those observed from the O–Ag(110) (
1.5 
3.3 eV) [294] and O–Ag(111) (
2.0 eV) [295] surfaces. These features were attributed to the contribution from O 2p states [291] while we have ascribed them as contribution from non-bonding lone pairs of O
2. 3.7.1.2. Analysis. It is interesting to note that the O–Ag(001) bi-phase structures are partly analogies of the O–Rh(001) and the O–Cu(001) surfaces. The fivefold coordinated oxygen is the same as that in the Rh(001)-c(22)-2O
1 first phase and the low temperature missing-row structure is the same as the Cu(001)p p ( 22 2)R45 -2O
2 second phase. Formulae of reaction and the surface atomic valencies of these two phases may refer to the conventions of the corresponding Rh(001)-O-1 and Cu(001)-O-2 phases (Sections 3.3.2 and 3.2.2, respectively). Neither like the transition from CuO2 paring-pyramid to the Cu3O2 pairing tetrahedron at the Cu(001) surface nor the transition from the Rh5O to the rotated Rh4O at the Rh(001) surface, the O–Ag(001) surface transits from the Ag5O (O–Rh(001) first phase) to the Ag3O2 [O–Cu(001] second phase] at much lower critical temperature. This intriguing fashion of low-temperature phase transition may originate from the low electronegativity (1.9) and the large atomic size (rAg=1.442 A˚) of Ag atom (see Table 1). At the low temperature, O
2 is more stable than the O
1 according to the XPS and EELS spectral features. The valence DOS features agree with that detected from other oxide surfaces. These features can be unambiguously specified as the contribution from the nonbonding lone pairs of oxygen upon the sp-orbital hybridization. The transition of XPS O 1s peak energy and the EELS stretch vibration are indication of oxygen dehybridization, which happens to the O–Cu(001) at about 700 K [209].

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

601

3.7.2. O–V(001) 3.7.2.1. Observations. With the STM, LEED, and DFT approaches, Koller et al. [296] investigated the O-induced V(001) surface reconstruction. The STM images in Fig. 23 show apparently the bi-phase structures beside the visible dark lines arranged regularly at the surface along the < 100> directions. The distance between the dark lines is usually 4–6 vanadium < 100 > lattice constants, and, therefore, a (15) reconstruction mode was suggested. In the simulated STM images, the fourfold hollow sites occupied by oxygen atoms appeared as dark spots while the unoccupied hollow sites remain bright. The first phase within the domain is the same as the Rh(001)-c(22)-2O
1 phase [see Fig. 20(a)]. Oxygen occupies the next nearest fourfold hollow-site and induces the radial reconstruction, as indicated in the STM image (b). AES detected oxygen coverage as 2/3 ML. With increasing oxygen coverage to  0.73 ML, the second phase appears, in which the dark lines remain. The short-ordered, zigzagged ‘O–O’ chains run along the < 11> direction. The bright spot is composed of four atoms. Ab initio calculations at very low oxygen coverage revealed that the C4v-hollow sited oxygen along the dark lines is energetically most favorable (
4.89 eV, Fig. 24a) and is followed by the twofoldcoordinated bridge site (
3.17 eV) and then the site atop a vanadium atom (
2.97 eV). The optimal structures at higher oxygen coverage are given in Fig. 24(b, c) with corresponding binding energies of
5.26 eV and
5.20 eV, respectively. One should note that the oxygen atoms in the dark lines now prefer the bridge sites in the two phases involving the reconstructed domains compared with the initial phase of oxidation. LEED optimisation [296] with a total number of 26 independent parameters leads to a structural model for the second phase (Fig. 24d, with R-factor value of 0.17), which is inconsistent with the models derived from STM imaging and ab initio calculations. The oxygen atoms in the bridge sites reside 0.12–0.18 A˚ above the top vanadium layer, while the oxygen atoms in the C4v-hollow sites are  0.5 A˚ above the top layer. The structural discrepancy determined by different methods was attributed to: (i) the ab initio data are for T=0 K, whereas the LEED was taken at room temperature; (ii) the involvement of the (14) and (16) superstructures may influence the LEED measurements. Nevertheless, both the LEED and DFT optimization revealed an expansion of the first layer spacing (+4.2%) and a contraction (
3.2%) of the second with respect to the clean V(001) surface that contracts by
7
8.5% (determined by LEED) or
15% (by DFT). 3.7.2.2. Analysis. Disregarding the accuracy in the vertical positions of the oxygen atoms in the dark lines, the bi-phase structure for the O–V(001) surface can be easily formulated. The formula for the first O
1-derived phase may refer to that for the Rh(001)-c(22)-2O
1(Y=1/2 ML) surface which showed the radial reconstruction (Section 3.6.2). The oxygen forms a bond with a Vanadium atom underneath and then polarizes the surface neighbors, which form the V5O structure. With the increase of oxygen the O
1 evolves into the O
2 that gives p coverage, p  rise to the short-ordered ( 23 2)R45 -4O
2 phase, which can be formulated as follows:

602

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 23. (a) and (c) show the STM images of the O–V(001) bi-phase structure [296] and the corresponding models. (b) and (d) denote the corresponding models for the V(001)-c(22)-2O
1 and the p p ( 23 2)R45 -4O
2 phases. The first phase (a) is the same as that of the Rh(001)-O
1 surface which is composed of V5O. The second phase (c) consists of the V4O tetrahedron. In the second, the zigzagged O– O chain forms along the <11> direction.

p p  V(001)-( 23 2)R45 - 4O
2 (Y=2/3 ML): 2O2 þ 6V ðsurfaceÞ þ 6V ð2nd layerÞ ) 4O
2 þ 4Vþ ð2nd layerÞ þ 4Vþ=dipole þ 2V ð1st layerÞ þ 2V ð2nd layerÞ p p A V(001)-( 23 2)R45 -4O
2 complex unit cell is shown in Fig. 24(e). The bright protrusions correspond to V+/dipole. Compared with the patterns of reconstructions on the fcc(001) surface of Cu and Rh, the V(001)–O surface exhibits a quite different pattern of reconstruction. There are two outstanding issues needing clarification. One is the inconsistency between structures determined by LEED and by DFT/STM observations; the other

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

603

Fig. 24. Comparison of the structural models. Ab initio optimization of (a) the C4v-hollow sited oxygen in the dark lines and the arrangement of oxygen for (b) the first and the (c) second phase as observed with STM. Panel (d) shows the LEED optimal structural model for the reconstructed surface at higher oxygen p p coverage. Panel (e) is the current bond model suggesting a short-ordered V(001)-( 23 2)R45 -4O
2 (2/3 ML) reconstruction.

is the oxygen coverage. AES detected higher oxygen coverage (0.67 and  0.73 ML for the two phases, respectively) but the model gives only 1/2 and 2/3 ML ideally for the two corresponding phases. The structural discrepancy may arise from the huge parameter space that warrants non-unique solutions [99]. The agreement between the STM and the DFT results supports each other, which may indicate the essentiality of searching for other possible numerical solutions in LEED optimization, such as sub-surface oxygen. Modeling analysis conducted here insofar indicates that the oxygen atom always tends to be located inside the tetrahedron. On the other hand, the V atoms in the center of the bright protrusions are expected to be missing as these V atoms interact with no oxygen atoms but they experience strong repulsion

604

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

from the neighboring V+/dipole. The expected missing V atoms may reduce the number of V atoms at the surface layers. In addition, the density of oxygen in the dark lines may be higher than the oxygen atoms inside the reconstructed domains. This may further increase the surface atomic ratio of O : V. These two factors, would account for the difference in the oxygen coverage between the AES measurement and the model specification. Further confirmation of the missing row formation at the V(001)-O surface is needed. It is interesting to compare the O–(Cu, Rh, Ag, V)-(001) surface reconstruction. The different atomic sizes and different values of electronegativity give rise to the entirely different patterns of reconstruction on the four surfaces of the same lattice geometry. The process of tetrahedron bond formation is the same despite the different transition temperatures and the corresponding patterns of observations.

4. STS and PES: valence DOS 4.1. Signature generality Chemisorption is a process in which chemical bond forms and the valence electrons transport among the bonding constituent atoms. Processes of charge transport modify the valence band structure and introduce additional DOS features. The derived DOS features are detectable using STS (around EF), PES (E< EF), and IPES (inverse PES, E> EF). The database for the oxygen-derived DOS for metals has been well established. However, identification of the DOS features in the valence band and above is still under debate and a clear definition of the DOS features is necessary. 4.1.1. STS STS measurements can be carried out by recording the dIt/dVt
Vt or d(lnIt)/ d(lnVt)
Vt curves, at a constant tip current, It, and various tip voltages, Vt. The features of an STS spectrum are associated with the on-site DOS of a few atoms at a surface [297,298]. Features below EF (tip negative bias) correspond to the occupied DOS at the surface while features above EF (tip positive) represent the allowed, yet unoccupied, DOS of the sample surface [16]. Fig. 25 shows the STS spectra of a Cu(110) surface [47] and of an Nb(110) surface [304] with and without chemisorbed oxygen. In the first panel, spectrum A was taken from the clean Cu(110) surface while B and C were taken from, respectively, the site above the bright spot (dipole) and the site between two bright spots along a ‘O
2 : Cudipole : O
2’ chain at the Cu(110)-(21)-O
2 surface (Fig. 6 in Section 3.2). On the clean surface, empty DOS of 0.8–1.8 eV above EF are resolved and no extra DOS structures are found below EF. The STS spectra recorded from the Cu(110)-(21)-O
2 islands reveal that the original empty-DOS above EF are partially occupied by electrons upon chemisorption, which result in a slight shift of the empty DOS to higher energy. Additional DOS features are generated around
2.1 eV below the EF. The sharp features around
1.4 eV have been detected with ARPES [233] and with the de-excitation spectroscopy of metastable atoms [300]. The

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

605

Fig. 25. STS profiles of (a) a Cu(110) surface [47] and of (b) a Nb(110) surface [304] with and without chemisorbed oxygen. Spectra in panel (a) were obtained (A) at a metallic region, (B) on top of, and, (C) between protrusions of the ‘O
2 : Cudipole : O
2 :’ chain [299] on the Cu(110)-(21)-O
2 surface. Insert in panel (b) shows the STM image of O–Nb(110) surface with O–Nb chains and triangle-shaped atomic vacancies.

Cu–3d DOS are between
2 and
5 eV [35,233,301], and the O–Cu bonding derivatives are around the 2p-level of oxygen,
5.6 to
7.8 eV below EF [302]. Both Cu–3d and O–Cu bond DOS features are outside the energy range of the STS (EF  2.5 eV). Recently, Uehara et al. [303] measured from the O–Cu(110) surface a 2.3 eV peak using the STM light-emission spectra. This feature was associated with the electron transition from the Cu 4s–4p hybridized orbitals of Cu to the empty O-2p orbitals. It is to be noted that the STS features of spectrum C (taken from between the bright spots) is more pronounced than that of spectrum B. Taking the tip-size effect

606

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

of an STS (with  2.5 A˚ lateral uncertainty) and the constant current mode into account, the origin of the intensity difference between spectra B and C can easily be understood. Spectrum C is taken from the depression (inward curvature) that is atop an oxygen adsorbate. The above-EF feature is dominated by the two neighboring dipole protrusions while the below-EF information comes from the lone pair of the O
2 underneath. Thus, it is understandable that the above-EF features of profile C are stronger than those of profile B as profile C is collected mainly from a dipole (from outward curvature) site. The tip above the dipole collects information of both the lone pair and the dipole but the information is relatively weaker because of the positive curvature at the dipole site. So the intensity of an STS spectrum is determined by (i) the tip size, (ii) the on-site curvature and, (iii) the energy states of O
2 adsorbate ( < EF, lone pair electrons) and Cudipole (> EF, polarized electron). Therefore, the origin for the new peaks at
2.1 eV and the shift of the surface states at 0.8–1.8 eV become clear now with the current BBB modeling specifications [299]. STS spectra in a broader energy range (
8, 8 eV) from the O–Nd(110) surface (Fig. 25) show DOS features near to the EF of O–Nb(110) is not that apparent as from the O–Cu–O chain [304]. They are rather weak compared with the intense resonant features. The resonant peaks at positive bias (unoccupied states) were assigned to a tunneling via quantized states in a potential well induced by the combination of image states and the applied electrical field [305,306]. Since the lowest image state is energetically tied to the vacuum level, the position of the first resonance can be used for an estimation of the work function [306]. The successive higher resonance at 6.4 and 7.7 eV are in accordance with resonance found for other transition metals such as Cu/Mo(110) [305] and Ni(001) surfaces [306,307]. These resonances are shifted to lower energies on the O–Nb surfaces, independent of the tip positions, i.e., on or away from the O–Nb chains. The resonance shift was related to the variation of work function upon oxidation [304]. During oxygen chemisorption at room temperature an initial decrease by =
0.45 eV for less-than-monolayer coverage was observed followed by an increase of 0.8 eV beyond the value for the clean surface for higher coverages [308]. The shift of the second peak at 5.6 eV coincides well with the work function increase and the rest two peaks are attributed to satellites of resonance. The occupied DOS features on the O–Nb chain are at
5.8 eV and at
6.2 eV at the STM triangular vacancy position compared to the DOS at
5.0 eV for clean Nb(110) surface. The energy shift of the occupied states was attributed to the overlap of O 2p and Nb 4d states with convolution of the reconstruction effect [304]. 4.1.2. PES, IPES and XPS Fig. 26 shows the ARPES spectra from the O–Cu(110) surface. Three apparent features were recognized and they are interpreted as follows [59]. The assignment of py and pz comes from two sources. Based on the geometrical structure it was assumed that the strongest Cu–O interaction is along the O–Cu–O chains and thus one will make the assignment py to the structure with the largest dispersion. Secondly, this feature is only observed at large incident angles # (angle between the incident beam and the surface normal), which suggests that one is observing an orbital in the plane, and this could only be the py orbital (along the O–Cu–O chain)

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

607

Fig. 26. ARPES spectra for O–Cu(110) [59] surfaces. Two additional features around
2 eV and
6 eV are resolved coinciding with the features resolved with STS.

because of the polarisation dependence. The assignment of the pz structure (the asterisk indicates an anti-bonding level) stems from the observed polarisation dependence and the comparison with the dispersion of the pz orbital. Hu¨fner [59] explained that the O–p bands located below the Cu–d band corresponds to the O–Cu bonding band, and those above the Cu–d band are the occupied O–Cu ‘anti-bonding’ bands. The oxygen 2py band shows the largest dispersion, as expected from the geometrical arrangement, which was suggested as indicative of strong bonding along the O–Cu–O strings. A set of PES profiles from an O–Cu(001) [309] surface shows three new DOS features around
1.45,
3.25 and
5.35 eV within the valence band region (above
7.04 eV) compared to that of a clean Cu(001) surface. The states around
1.45 eV (which coincide with the
1.4 eV STS feature in Fig. 25) were found to be antiresonant, i.e., the intensity has no apparent change with varying incident beam energy. The anti-resonant DOS feature is believed to be the character of electrons that are strongly confined in one-dimension such as molecular chains [310]. Thus,

608

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

the anti-resonant DOS features around
1.45 eV can be related to the electron lone pairs that zigzag the ‘O
2:Cudipole  :O
2’ strings at the Cu(001)-O
2 surface.
In the He-II h! ¼ 21:22 eV PES studies, Belash et al. [301] found that, with increasing number of oxygen atoms on a polycrystalline Cu surface, oxidation takes place in three steps: In the first step (the lowest, 12 L, exposures), oxidation begins, which immediately leads to the rise of a shoulder at
1.5 eV and a small peak at
6.0 eV in the PES spectra. The emergence of the new DOS features is at the expense of a sharp fall of the DOS features between
3.0 eV and EF. At the second step (12  1000 L), a further increase in the oxygen exposure leads to the increase of both the
6.0 eV DOS feature and the sharp-fall at EF > E >
3.0 eV. The DOS at EF falls to zero and a band-gap of  1.0 eV width is produced. At the third step with an oxygen exposure of 5103 L, a surface compound forms, which displays semiconductive properties. Its electronic structure is quite the same as that of the bulk Cu2O. These reaction steps agree well with the O–Cu(001) surface bonding kinetics as determined with VLEED (please refer to Section 3.2.3). A set of ARPES spectra, as shown in Fig. 27(a and b), from the O–Pd(110) surface [60] displays that two adjacent clusters of DOS populate below the EF of the Pd(110) surface. One is around
2.0 eV and the other is around
5.0 eV. These signatures arise at the expense of the weakening of the DOS close to the EF. PES investigations on the O–Rh(001) surface [157,159,166] showed that oxygen induces significant change in the energy states around
2 to
6 eV below EF. Fig. 27(c and d) [166] compares the ARPES spectra from the c(22)-O
1 (radial reconstruction) and (22)p4g-O
2 phase on the Rh(001) surface. DOS for holes below the EF can be resolved from the PES profiles of both phases. Additional O-derivatives can be identified at around
5 eV for the first (22)-O
1 phase. The feature around
5 eV shifts down a little with an additional peak at about
2.0 eV at the some azimuth angles for the O
2-induced ‘p4g’ phase. The PES features of the O–Pd(110) and O–Rh(001) surfaces are substantially the same as those observed from the O–Rh(110) [173] and the O–Cu{(001),(110)} [309] surfaces as well, despite the slight difference in peak positions. The DOS evolution kinetics of the O–Rh(001) surface in the phase transition agrees with the trend for the O–Cu(001) phase transition as determined with VLEED [97] and PES [301]. Fig. 28 shows the exposure dependence of the unoccupied bands of the O–Cu(110) surface measured using ARIPES ðh
! ¼ 9:7 eVÞ [141] of which the resonant features are quite similar to the STS from O–Nb(110) surface (Fig. 25). It was explained in Ref. [141] that the state centred at 4 eV above EF corresponds to a Cu-3d [10] state, in keeping with the Cu-3d [9] interpretation for the ground state, because with the IPES technique an additional electron is added to the copper–oxygen system. The obviously symmetric dispersion of this state with respect to the occupied oxygen 2py states is in agreement with a two-level approximation for the O–Cu s-bond. Chen and Smith [311] suggested that the band above EF is an empty surface state, which means that the agreement of its dispersion with that of the O-2py state below EF is accidental. There appears to be no definite interpretation to this phenomenon up to now. One might, however, take a very pragmatic view of that problem. Any state in

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

609

Fig. 27. ARPES profiles for (a and b) O–Pd (110) [60] and (c and d) O–Rh(001) [166] surfaces. Shaded areas are features derived by oxygen adsorption. A notable aspect upon the reconstruction is the two new humps around
2 eV and
5 eV. It should be noted that the ARPES intensity near EF at some angles such as 0 and 44 is reduced. For O–Rh(001), additional O-derivatives can be identified at around
5 eV for the (22)-2O phase. The feature at
5.0 eV shifts to
6.0 eV with an additional feature at about
2.0 eV at some azimuth angles when the (22)p4g-2O phase is developed.

a monatomic surface layer can be viewed as a surface state. In this sense, the two different interpretations given for the unoccupied state of the O–Cu(110) surface may not be all that different as commented by Hu¨fner [59]. Nevertheless, it should be noted that the feature at +2 eV coincides with those probed using STS on the same O–Cu(001) surface. This empty feature decreases with increasing oxygen exposure.

610

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 28. IPES (inverse photoemission) spectra of Cu(110)-O ðh
! ¼ 9:7 eVÞ [141] show the similar STS resonant features to that from the O–Nb(110) surface [304]. Feature D was explained as a direct transition in the bulk band structure, while S a surface state. Feature A is an oxygen-induced orbital. It should be noted that the peak at 2.0 eV decreases with increasing oxygen exposure, indicating the occupancy of the empty surface states due to dipole formation.

PES, IPESS and XPS data are available for systems of PdO [312,313] O–Cu(110) [314] O–Nb(110) [315] AgO [316] and Bi2Sr2CaCu2O8 [317], which share considerable DOS similarities in the valence band and above. Au nanoclusters deposited on TiO2(110) substrate also exhibits a weak feature at
1.0 eV due to the interfacial oxidation [318]. XPS study [62] revealed that oxygen adsorption shifts the Pd(3d5/2) core-level by  0.6 eV towards higher binding energy, which coincides with the O-1s core-level shift detected from the O–Cu(001) surface [319]. The O-1s level (
529.5 eV) shifts 0.6 to
530.1 eV when the oxygen reacts with the Cu(001) surface. It has also been reported recently [194] that the Rh-3d binding energy increases by about 0.3 eV per bond to an oxygen adsorbate at the Rh(111) surface. Two distinct components in the Ru-3d5/2 core-level spectra have been detected from the clean Ru(0001) surface [320]. With increasing oxygen coverage to the Ru(0001) surface, the Ru-3d5/2 corelevel peaks shift to higher binding energies by up to 1.0 eV. Recently, Yang and Sacher [321] found that the Cu-2p3/2 peak shifts positively with the particle size, D, which follows the relation: DEC ðDÞ ¼ A þ B=D. The A and B are constant depend-

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

611

ing on both surface treatment and substrate type. Nitridation of the sample could raise the slope B from 2 (Cu/graphite) to 4.2 (Cu–N/polymer), demonstrating nicely the joint physical and chemical effects on the core level shift. Therefore, as physical origin (without charge transport), surface relaxation and nanosolid formation share considerable similarity in splitting and shifting the core-levels of a specimen. The core-level shift detected with an XPS shows that: (i) the O-adsorbate does capture electrons from metal atoms, and, (ii) the core-level shift of the clean Ru(0001) surface may be indicative of the bond relaxation at the surface, which modifies the crystal field upon which core-level shift depends. Therefore, chemisorption is indeed a kinetic process in which electron transport dominates. Unfortunately, an XPS is unable to reveal direct information about the process of valence charge transport. In this sense, STS and UPS are more favorable than XPS but mechanism for charge transportation must be clear. 4.1.3. Indication Comparing the DOS features detected using STS, PES and IPES, one may conclude immediately that the oxygen-derived valence DOS features are very common for all the analysed systems. Slight difference in the peak positions may result from the difference in electronegativity that determines the ease of charge transport. One might have noted that the patterns of morphology and crystallography vary indeed from situation to situation, as discussed in Section 3, but the DOS spectral features detected are quite common. The challenge for us is to find a common mechanism behind these phenomenological observations. 4.2. Specification Opinions regarding the O-induced valence DOS features are often conflicting. For instance, there are have been several arguments on the additional DOS features around
1.4 to
2.0 eV of the copper oxide: (i) O–Cu anti-bonding states [233,309,322]; (ii) O2p anti-bonding states [59,233,322]; (iii) oxygen 2s states [323,324] and (iv) the interaction of O-2p electrons with the spd hybridized electrons of Cu [301]. The additional DOS features around
5.5 eV are interpreted as O-2p states adding to the valence band of the host surface [35,59,233]. The sharp fall of the DOS features at EF > E>
3.0 eV corresponds to the disappearance of the clean Cu surface states. Therefore, a consistent definition of the O-derived DOS features in the valence band of the hosts is essential. It has been clear now that chemisorption is a kinetic process in which the valencies of the bonding constituent change and hence the sizes and positions of surface atoms change. Electrons are transported from the valence band of the host to the empty p-orbital of oxygen for the bonding; then the oxygen hybridizes with the production of lone pairs. The lone pairs polarize in turn their surrounding neighbors and the electrons of the host dipoles move from the original energy level to the higher energy levels. These sequential processes will redistribute electrons at the valence band and above of the host (Section 2.3). Table 14 summarizes the adsorbate-derived DOS features in the valence band of metals. With the BBB model developed, we are able to define the O-derived DOS features consistently (Section 2.3):

612

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

The DOS features above EF (  2.0 eV for Cu) correspond to the occupation of empty surface states by the anti-bonding dipoles, which lower the work function. Resonant peaks may come from the surface image states. The shift of the resonances for O–Nb(110) lowered energy corresponds to the work function increases due to H-like bond formation at the surface. Lone-pair production generates the DOS feature between
1.5 and
2.0 eV below EF. Features around
5.5 eV are derivatives of the O-metal bonding. The sharp fall of the DOS features near the EF results from the hole-production in the process of bond and anti-bonding dipole formation. It is to be noted that the hole-production and lone-pair production have opposite effects on the DOS distribution between
3.0 eV and EF. The former weakens the DOS intensity while the latter enhances it. What one can detect is the resultant of these two effects. The DOS change detected in this region may be insignificant if these two opposite processes are comparable in quantity. Such a DOS definition may be applied to catalytic reactions involving other electronegative elements such as N and S. DFT approaches indicated that the adsorbateinduced DOS features for both the N–Ru(0001) [257] and the O–Ru(1010) [63] surfaces are the same and the identities agree with those specified by the current BBB correlation premise. Further experimental evidence for the sp-orbital hybridization with lone pair production of nitrogen has been reported recently [92,325–327]. STS [328] revealed strong DOS features in the conduction band of CNx nanotubes close to the Fermi level (0.18 eV). The sp orbitals of other electronegative elements are also anticipated to hybridize upon reaction with a solid surface but further confirmation is necessary.

5. TDS: bond nature and bond strength 5.1. Identity similarity One of the striking features observed from the O–Pd(110) surface is that the kinetics of surface adsorption, as detected with the exposure-resolved TDS, coincides with the trend in the work function change () [62]. The TDS and the work function observations are given in Fig. 29 and their coincidence is summarized as follows: (i) The four TDS peaks assigned as b1, b2, g1 and g2 oscillate with increasing oxygen exposure. The changes of the TDS peak intensities show clearly the discrete stages of reaction: o b2 emerges first upon introduction of oxygen to the surface and the b2 peak saturates at 2.5 L exposure, and then keeps constant until 240 L; o b1 emerges at 1.5 L and increases gradually in intensity until 240 L; o after an extremely high oxygen-exposure (22,800 L), a reversal, or oscillation, in the spectral intensities of b1 and b2 occurs. b1 is substantially more intense than b2, which differs significantly from the trends at lower oxygen exposures, where b1 is always weaker than b2. At the extremely

613

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Table 14 The adsorbate-derived DOS features adding to the valence band of metals (unit in eV). Holes are produced below EF. All the data were probed with ARPES unless otherwise indicated Other methods [Ref.]

Definition

Anti-bond dipole > EF

O–Cu(001) [309] VLEED [88] 1.2 O–Cu(001) [322] O–Ni(001) [329] O–Cu(110) [35,141, 233,330] O–Cu(poly) [301] O–Rh(001) [160] O–Pd(110) [62] O–Al(poly) [331] O–Gd(0001) [80] O–Ru(0001) [332] O–Ru(0001) [333] O–Ru(0001) [334] O–Ru(0001) [335] O–Ru(1010) [63] MgO/Ag(001) [336] O–Co(Poly) [337] O-diamond (001) [338] O–C(nanotube) [339] N-Cu(001) [340]

Lone pair
Metal
O–M bond

1.50.5
2.1
1.37;
1.16


3.0 1.0


6.5 1.5


1.50.5
2.10.5
1.5
3.1
2.00.5


3.0 1.0

EF 
6.0
6.5 1.5

SXES STS [47]

2.0 1.3 0.5

DFT

1.0

STS

1.0

Ab initio DFT

1.5 1.7 2.5

UPS UPS

XAS XES [329] DFT

N–Ru(0001) [257] (O, S)–Cu(001) [340] TiCN [341] XPS a-CN [342] XPS CN [343] XPS (N, O, S)–Ag(111) [344]

0.8 3.0 3.0


3.0
1.01.0
0.8
4
3.0
2
3.0
3.01.0
2.0
3.0


1.2
1.0
3.0
1.3 0.01.0
4.5
2.3
3.4


3.0 1.0
0.5, 3.0


6.5 1.5
5.8
4.5 1.5


1.0,
8.0
6.0
5.5 1.5
4.4
5.5,
7.8
5.8
5.0
0.7


4.0


5.0


5.6
5.5
6.0
6.0
5.7
7.1
8.0

Previous Surf. states O2s [324] Anti-bond O-2p [35,59,233] explanations antibond [334] O2p [59,233] O–Cu O–M O–M anti-bond [322]

higher exposure, two troughs, g1 (600 K) and g2 (705 K), emerge at the expense of slowing the intensity-increase of b2. (ii) The exposure-derived  of the O–Pd(110) surface agrees with that of the O–Ru(0001) surface [86,263] and the O–Gd(0001) surface as well [80]. The trend for the  for the O–Pd(110) surface can be summarized as follows [62]: o the  reaches 500 meV at 1.0 L oxygen exposure and then to a maximum 520 meV at 1.5 L; o from 1.5 to 2.5 L the  decreases from 520 meV to 430 meV;

614

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 29. (a) TDS profiles from Pd(110) surface [62] exposed to oxygen at 304 and 400 K, showing the oscillation of the TDS peaks with increasing oxygen exposure, and, (b) the oxygen-exposure dependence of the work function change  derived from the UPS spectra of the O–Pd(110) surface [62]. The separated regions correspond to the O
1, O
2-hybrid and H-like bond formation.

o the  reduces from 430 meV at 2.5 L and less significantly to 400 meV at 10 L. Zheng and Altman [345] observed the similar trend of TDS from an O–Pd(001) surface. The Pd(001) surfaces exposed to oxygen at 350 K temperature results in a single TDS peak at 850 K. The peak shifts from 850 K at lower coverage to 800 K after a 10 L exposure at which the highest temperature peak saturates. After the 800 K peak saturates, a new TDS peak appears at 700 K that saturates at an exposure of

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

615

20 L. Meanwhile, another peak shows up at 650 K, which shifts towards 700 K with further increasing exposure. Above 1000 L the oxygen uptake slows and an additional peak shoulder at 600 K is observed. Fig. 30 shows the exposure-resolved TDS profiles recorded from the O–Rh(110) surface [173]. The TDS profiles exhibit five desorption maxima around 797, 835, 909, 1095 and 1150–1190 K. Besides the addition of the fifth peak, the peak temperatures for the O–Rh(110) surface are slightly higher than those for the O–Pd{(001),(110)} surface. Comelli et al. [27] have observed the peaks b2 and b3 at lower exposures to the O–Rh(110) surface. Bowker et al. [346] also resolved three TDS peaks of b2–b4 from the same O–Rh(110) surface. The emerging trend and the oscillation of these peaks are substantially the same as the set for the O–Pd{(001),(110)} surfaces though the peak positions differ slightly. The broadened peak b5 which shifts to lower temperature with increasing population of desorption states indicates the second-order desorption kinetics. All the other TDS maxima are invariant with coverage according to the first-order kinetics. By combining the TDS profiles and LEED crystallography of the O–Rh(110) surface, Schwarz et al. [173] correlated the TDS peaks to the LEED patterns of various reconstructed phases. Comelli et al. [27] re-interpreted the TDS data of the O–Rh(110) surface based on the known adsorbate’s locations. It was suggested that the most stable structure is the p2mg phase and so the adsorbate binds to the surface more strongly. Due to the repulsive interactions, the adsorbate in the unreconstructed troughs binds to the surface weakly.

Fig. 30. TDS profiles for the Rh(110) surface exposed to oxygen at 573 K [173]. Features oscillate with a similar trend to those for the O–Pd(110) surface upon increasing oxygen exposure.

616

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Similarly, TDS studies [158,347–349] revealed that the O–Rh(001) surface reaction involves three phases. The first phase gives no desorption feature up to 350 K, when it converts to a state tentatively assigned as ‘oxygen atomic state’—or physisorption. For temperatures above 500 K, a new state appears, which was assigned as ‘oxide state’—chemisorption. The latter two states yield oxygen ions of different kinetic energies. Fisher and Schmieg [165] reported three peaks (820, 920, and 1325–1200 K with the corresponding enthalpy of 210, 260 and 360 kJ/mol) in the TDS spectrum of the same O–Rh(001) surface. A kinetic Monte Carlo simulation of the O–Rh(001) TDS features [350] derived the desorption peaks of 820, 925, and 1250 K agreeing with the measurement. The state of desorption at T > 1200 K was related to the p(22)-O structure (corresponding to the O
1 state at oxygen coverage smaller than 0.5 ML, as analyzed in this report) and it is a second order. The 920 K state showed first order kinetics, which was related to the c(22)-2O phase (corresponding to the O
1 state at 0.5 ML coverage). The 820 K peak was associated with desorption from the c(22)p4g-2O structure (O
2 dominates with H-like bond involvement at 0.5 ML coverage, see Section 3.6). This implies that the adsorption enthalpy of oxygen decreases with the evolution of the O
1 to O
2. The initial precursor states are more stable than the fully developed ‘p4g’ phase, according to Ref. [350], which is conflicting with the descriptions of Schwarz et al. [173] and Comelli et al. [27] for the O–Rh(110) surface. Based on their TDS measurement and the time-of-flight profiles from the O–Rh(111) surface, Peterlinz and Sibener [192] suggested that: (i) sub-surface oxygen forms first on the Rh(111) surface at a temperature below 375 K; (ii) the subsurface oxygen starts to segregate to the surface at 375 K and, (iii) the desorption occurs at above 650 K. Desorption is associated with a fairly sharp TDS peak at about 800 K. The velocity distribution of oxygen molecules was found to be the same for the same oxide bulk to that for the surface with adsorbed oxygen. This can be indicative that oxygen desorbing from the surface is in the same state as the original state of oxygen before desorbing from the surface. Logan et al. [351] obtained a set of TDS data from the O–Rh(111) surface in the temperature range of 850–1050 K, higher than those reported by Peterlinz and Sibener who used lower oxygen doses. Using the time-of-flight spectroscopy, Gibson et al. [53,193] found that the velocity distribution of the desorbed oxygen follows approximately the Maxwell– Boltzmann description. They relate such characteristics to a ‘hot’ desorption, that is, the temperature of the desorbing gas is higher than the surface temperature. The gas temperature of 1175 K is higher than the sample temperature detected with TDS (700–1100 K), which implies that the TDS value of the heat of adsorption should be corrected slightly. Bo¨ttcher et al. [352] detected two TDS peaks at 400 and 1100 K from the Ru(0001) surface at oxygen coverage below 0.25 ML. This differs somewhat from the phase diagram reported by Piercy et al. [353] The phase diagram shows two peaks at 754 and 555 K. Despite the slight difference in the number of the TDS peaks and the binding strength that varies with the surface orientation and with materials, one may conclude that the TDS features of oxide surfaces share considerable similarities as compared in Fig. 31 showing the temperatures of the TDS peaks for the O–Rh and O–Pd surfaces.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

617

Controversies surrounding the origins of the TDS identities are apparent, as addressed above. For the O–Pd(110) surface example, one opinion [62,247,354] relates the b1 feature to sub-surface oxygen and the more strongly bound b2 peak to surface oxygen. Another opinion [355] assumes that the two peaks come from the 18 O and 16O isotopes. There is still a dispute on the reason for the oscillation of b1 and b2 for O–Pd(110). Bondzie et al. [62] attributed the oscillation to a process of oxidation–reduction while Bassett and Imbihl [354] and Ladas et al. [247] related the oscillation to the filling and depletion of sub-surface oxygen. As indicated by Comelli et al. [27] the detailed nature of the sub-surface state or states of oxygen has not been defined precisely though the sub-surface oxygen has been widely reported on the Rh surfaces. Gottfried et al. [356] investigated gold (110) surface oxidation by bombarding the surface with energetic oxygen ion. TDS revealed four TD peaks at 415, 545, 620 and 750–950 K. These peaks are associated with chemisorbed atomic oxygen, oxygen atoms chemically dissolved in the bulk as well as gold oxide. The peak positions are found to depend strongly on the ion energy (penetration depth) and substrate temperature as well. Increasing the ion energy from 1.0 to 5.0 keV, the 750 K peak shifts to 850 K with attenuation of peak intensity of the lower temperatures, indicating the bulk site occupation. With a constant ion energy, 2.0 keV, the 750 peak shifts to 800 K with announced low temperature peaks present when the sputtering time is prolonged from 1 to 20 min. The heating rate in the TDS measurement also affects the TDS features. Raising the heating rate has the similar effect to prolong the sputtering duration.

Fig. 31. Comparison of the transition temperatures and the number of TDS peaks for O–Pd(110) [62] and O–Rh(001) [165], O–Rh(110) [173] and O–Rh(111) [351] surfaces. The number of the TDS peaks is nearly the same although the binding strength varies with the surface orientations and with materials.

618

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

5.2. Specification Thermal desorption, known as bond breaking, is actually a reverse process of adsorption. As noted by Redhead [61], TDS can clearly separate the multiple binding states of the adsorbate atoms. Therefore, the peaks in the TDS profiles are associated with the individual process in which a specific bond forms. The similarity of the TDS signatures and generality of the DOS features for all the analyzed systems exhibit the inherent correlation of bond formation and charge transportation. Encouragingly, the process of electronic transportation and the nature of the chemical bonds involved are intrinsically common for all the analyzed systems, even though the patterns of observations by other means may be different. This enables us to relate the TDS signatures to the bonds of different nature, and to the DOS features in the valence band, for the O–Pd(110) surface example, as listed in Table 15. Similarly, the TDS signatures in Fig. 30 can also be specified correspondingly. The peaks correspond to the activation energies for the individual bond breaking. It should be a helpful practice for interested readers to interpret the TDS of the O–Rh(110) surface. The fifth peak may relate to the contracting ionic bond at the initial stage of reaction and the bond length relaxes with the development of the tetrahedron, which lowers the activation energy. The two peaks at 754 and 555 K appearing in the O–Ru(0001) phase diagram [353] can also be related to the ionic bond and the non-bonding lone pair interaction, which has been confirmed to exist in the DFT approaches [63,257], but the TDS features [352] at low oxygen coverage may correspond to the on-surface oxygen (400 K) and the first contracting bond (1100 K) between the O
1 and the Ru atom underneath. As will be shown in Section 7, such specification allows the TDS oscillation for O–Pd{(001), (110)} to be related to the bond forming kinetics, agreeing with observations using other means, such as PES and VLEED, for other systems with chemisorbed oxygen.

Table 15 Correlation between the bond nature and the TDS and valence DOS features for the O–Pd(110) and O–Pd(001) surfaces Bond nature

Ionic bond Non-bonding lone pair H-like bond H-like bond Anti-bonding dipole

TDS features (K)

DOS features

O–Pd(110) [62]

O–Pd(001) [345]

O–Pd(110)

b2(816) b1(774) g2(705) g1(660) 

850!800 800 700 600–650


5 eV
2 eV  recovery  recovery +2 eV

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

619

6. EELS and Raman: fingerprints of weak interaction 6.1. EELS: dipole vibration In a high-resolution EELS study, He and Jacobi [64] observed a dipole-active stretch vibration mode, ?(Ru–O), from an O–Ru(0001) surface. The energy of the p(Ru–O) mode shifts from 54 to 81 meV, when the (22)-O (O
1 dominates at 0.25 ML coverage) and the (21)-O adlayer (O
2 and lone pairs dominate at 0.5 ML) develops into the p(11)-O phase (O
2 and H-like bonds dominate at 1.0 ML coverage). The low-frequency EELS and infrared peaks have also been observed from other oxide surfaces [357,358]. The observed trend is just the same as that occurring to the O–Rh{(111), (110), (001)} surfaces upon oxygen adsorption. HREELS [359] revealed that the characteristic losses for the O–Rh(111) surface increase from 62 to 68 meV with oxygen coverage. On the O–Rh(001) surface [165], a single loss at 48 meV at lower oxygen coverage (O
1 dominates) shifts to 54 meV at higher oxygen coverage (O
2 dominates). Upon the Rh(110)-(21)p2mg-O structure formation, two losses at 46  1 and 64  1 meV are observed simultaneously under dipole scattering conditions. The EELS peak transits at low temperature from 36 to 30 meV [290] when the O–Ag(001) surface changes from the O
2 derived phase to the O
1 derived one. The transition from lower energies to slightly higher indicates a higher binding energy since the (22)p2mg structure is more stable. However, the origin and the energy shift of the detected mode still has to be defined although this observed energy difference has usually been suggested as fortuitous [27,181]. A nuclear inelastic scattering of synchrotron radiation measurement [360] of the vibrational DOS of nanocrystalline (6–13 nm) a-Fe with oxide covered surface revealed: (i) enhanced population of low-energy vibrational modes around 18 meV, (ii) a broadening of the DOS peaks at 30–35 meV and, (iii) an additional intensity at 40–50 meV. The softened feature (i) is attributed to vibrational modes of interface atoms, arising from the high fraction of interfacial sites connected with the small crystallite size. The broadening mode (ii) is caused by phonon confinement and the stiffening mode (iii) from oxidation. Experiments and computer simulations indicate that the DOS contains low- and high-energy modes not exhibited by coarse-grained counterparts. These modes seem to originate from vibrations of atoms with reduced atomic coordination and modified local environment, i.e., at surface/interface sites. It is concluded that oxidation similarly contributes to the low-energy DOS but additionally brings about stiff modes above the high-frequency cutoff of bulk a-Fe. It is surprising to note that the energy of the stretch vibration of O–M in EELS around 50 meV coincides with the typical strength of the hydrogen bond detected using infrared and Raman spectroscopy from H2O, protein and DNA [104]. The energy for an ionic bond is normally around 3.0 eV and the energy for a Van der Waals bond is about 0.1 eV. Therefore, the vibrations detected with EELS from the Ru, Rh and Ag oxide surfaces correspond to the weak non-bonding interaction between the host dipole and the oxygen adsorbate. For the O–Ru and O–Rh surfaces, it has been clear that, in the precursor states at lower oxygen coverage, the weak interaction between the dipole and oxygen is dominated by the O
1 that

620

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

induces the dipoles. A lone pair replaces the O
1 when the O
1 evolves into the O
2 with increasing oxygen coverage. The lone-pair dominates the weaker part of a hydrogen bond. Finally, a H-like bond forms, which stabilizes the bond network at the surface by reducing the dipole moment. Therefore, it is understandable that the interaction between oxygen adsorbate and the metal dipoles increases with the evolution from O
1, O
2 to O
2 with H-like bond formation. Dipole-oxygen interaction through a lone pair seems to be stronger than that through the O
1. The H-like bond formation stabilizes the surface bond network and hence raises the frequencies of vibration. Therefore, the detected EELS and vibration DOS features correspond to the non-bonding interaction and their energy-shifts agree with the transition of the non-bonding components from O
1 induction to lone pair interaction and then to the H-like bond contribution in the process of oxidation. The soften mode in the vibration DOS of nano-Fe oxide may be due to interparticle interaction, which even weaker than the nonbonding interaction. 6.2. Raman: lone-pair in oxides, nitrides and bio-molecules It is known that the lone-pair is produced intrinsically by the sp-orbital hybridization. The number of lone-pairs in a tetrahedron follows the ‘4-n’ rule (Section 2.2), where ‘n’ is the valence value of the electronegative additive. Vibration of the dipole induced by the lone pairs should be detectable by Raman spectroscopy in the frequency range below 1000 cm
1. An experimental survey with Raman (HeNe laser, normal incidence) has been conducted to examine the following specimens [327]: (i) Al2O3 and TiO2 powders; (ii) thin films of Ti nitride (TiN) and amorphous carbon nitride (CN); and, (iii) films of amorphous carbon (a-C) and Ti carbide (TiC). As anticipated and shown in Fig. 32(a), the lone-pair features of the oxides (n=2) are stronger than those of the nitrides (n=3) while no such features can be resolved from carbides (n=4). The appearance and the relative intensity of these low-frequency Raman features support the modeling prediction and the rules of ‘4 n’ for lone-pair formation as well. The detected Raman features are quite similar to that of H2O, protein and DNA [104]. The peak positions depend on the reduced atomic mass, =m1m2/(m1+m2), of the components, and the force constant, fk, of the weak interaction [! / (fk/)1/2]. The multi-peak corresponds to different orders of the Fourier coefficients in the numerical solutions. These features in bio-molecules used to be attributed to the hydrogen bond vibrations with  50 meV binding energy. It is known that the hydrogen bond is actually composed of a lone pair (‘:’) on one side and a covalent bond (‘–’) on the other side of the B+/dipole (A
n–B+/dipole : A
n). The B is less electronegative than A of which the sp orbitals hybridize upon reaction. The covalent bond vibration contributes to the spectra at much higher frequencies. Therefore, Raman low-frequency features come from the vibration of the lone-pair-induced dipoles that seem to be common to oxides, nitrides and bio-molecules. 6.3. Confirmation: ultra-elasticity of nitride surfaces The ultra-hard and self-lubricative behavior of nitride surfaces may provide further evidence for the lone-pair interaction. The quasi-tetrahedron bond geometry of

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

621

Fig. 32. (a) Low-frequency Raman shifts indicate that weak bond interaction exists in Ti and Al oxides and TiN and amorphous CN, which correspond to the non-bonding electron lone pairs generated during the sp-orbital hybridization of nitrogen and oxygen. Therefore, the peak intensities of oxides are stronger than that of nitrides while there are no such peaks at all for a-C and TiC [327] films. (b) The elastic recovery profiles for GaAlN and a-C evidence the lone-pair interaction dominating at the nitride surface and the strong intralayer ionic interaction [327].

a nitride prefers the C3v symmetry because a nitrogen atom needs three electrons from its surrounding atoms for a tetrahedron formation [92]. This explains why most nitride compounds such as group III-nitride wide-band-gap semiconductors prefer the cubic fcc(111) or the hexagonal hcp(0001) orientations. Ideally, nitrogen tends to locate at the hcp hollow site to form a tetrahedron with three identical bonds to the surface atoms and with one lone pair interacting with the atom underneath. Therefore,

622

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

the nitride surface-bond network is composed of ions (N
3 and M+) and these ions are surrounded with densely packed electron clouds. On the other hand, instead of the original strong interlayer metallic bonds, lone pairs dominate the interaction between the interatomic layers. Hence, the nitride surface-bond network of densely packed electrons ensures the hardness at the surface atomic layer while the weak interlayer interaction gives rise to the high elasticity perpendicular to the surface. Nanoindentation profiles in Fig. 32(b) confirm the predictions of high elasticity of the nitride surfaces contributed by the lone pair interaction. For GaAlN film [361], the elastic recovery is as high as 100% under lower indentation load (0.7 mN) compared with that of amorphous carbon film. The GaAlN surface is also harder than the a-C film under the lower indentation load. The absence of lone pairs in a-C film makes the carbide less elastic than a nitride under the same scale of indentation load. For CN and TiN, the elastic recovery ranges from 65 to 85% with higher load (5 mN) of indentation [362]. The plastic deformation of the nitride films at higher indentation force indicates the existence of a critical value of load that breaks the lone pair interaction. Therefore, the non-bonding interlayer interaction enhances the elasticity of nitride surfaces at very low load. Such high elasticity and high hardness by nature furnishes the nitride surfaces with self-lubricative property for nano-tribological applications.

7. Kinetics of bond forming and bond switching 7.1. Four-stage bond forming kinetics The correlation of the chemical bond, surface morphology, and valence DOS and the bond strength to the spectral signatures of LEED, STM, PES/STS, TDS and EELS enables the kinetics of oxide tetrahedron formation to be readily understood, in general. As an additional example, we may look at the particular O–Pd(110) surface based on the kinetic TDS and UPS profiles (please refer to Fig. 29 for the profiles and Section 3.3 for structural details) obtained during the increase of oxygen exposure:  Stage 1 (Y < 1.5 L): O
1 dominates giving a Pd5O cluster with one ionic bond to the Pd atom underneath and four surface metal dipoles. In TDS, the ionic bond feature b2 emerges before the lone pair feature b1 which does not appear yet until the turning point at 1.5 L. The O
1-induced anti-bonding dipoles reduce  considerably. The highest  value at 1.5 ML is evidence of the precursor in which O
1dominates. This agrees with data derived with VLEED from the O–Cu(001) surface at an oxygen-coverage lower than 25 L. No lone pair features can be detected at the O
1 dominated stage.  Stage 2 (1.54Y42.5 L): O
2-hybridization occurs and completes, giving rise to a tetrahedron with two ionic bonds and two non-bonding lone pairs. The lone pair feature b1 emerges upon the sp orbitals of the O
2 being hybridized. The lone pair feature b1 becomes apparent at 2.5 L. The bond feature b2 increases its intensity until the number of the ionic bonds saturates at 2.5 L.  drops from 520 meV at 1.5 L down to 430 meV at 2.5 L because the

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

623

number of dipoles decreases. Two of the four surface dipoles are retained and the remaining two become Pd+ or metallic Pd when the Pd5O evolves into the Pd4O tetrahedron.  Stage 3 (2.5 < Y4240 L): Interaction develops between the adsorbate and the anti-bonding dipoles. The non-bonding feature b1 increases its intensity gradually indicating that more lone pairs are produced and the lone-pair interaction develops. The constant intensity b2 implies that the oxygen coverage saturates gradually with the exposure. In fact, there exists no direct correspondence between the oxygen coverage and the exposure. The slight decrease of  from 430 meV at 2.5 L to 400 meV at 10 L implies that the Hlike bonds start to form. The H-like bonds reduce the width of the anti-bond sub-band and the work function reduction is restored, consequently.  Stage 4 (Y  22,800 L): The number of H-like bonds increases with increase in the O: M ratio at the surface. Each O
2 needs three surface neighbors of which one bonds to the oxygen and the rest two are to be polarized. Therefore, the increased number of H-like bonds replaces some of the original ionic bonds at the surface. The development of H-like bond features g1(600 K) and g2 (705 K) can be regarded as a process in which the H-like bond dominates. H-like bonds replace the ionic bonds at the surface, and the ionic bond peak b2 is thus apparently lowered relative to the lone-pair peak b1. The considerable increase of b1 intensity indicates that more lone-pairs have been produced, and, thereby, oxygen coverage is increased. Unfortunately, the database of  at higher exposures is lacking but we may anticipate that the  will recover further as the H-like bond narrows the width of the anti-bond band. This is also the case for the hcp(1010)-(21)p2mg-2O
2 phase of Co and Ru. The TDS oscillation of the O–Pd{(001),(110)} corresponds to the sequence of bond formation: (i) an ionic bond forms first, (ii) a lone pair follows upon second bond formation, and then, (iii) an H-like bond replaces the ionic bond at the surface with increase in the number of oxygen atoms adsorbed. These processes are the same as for the O–Cu(001) surface, and this has been quantified with VLEED (Section 3.2.3). The trend also agrees with the reaction kinetics of the polycrystalline O–Cu surface revealed with PES [30]. In general, the identities of kinetic UPS and VLEED correspond to the oxygen-derived DOS features. The exposure-resolved TDS signatures can be related to the processes of bond forming. Therefore, it is easy to view the oxide bond and band forming kinetics from any complete kinetic spectra database, using the original BBB model that has been proven appropriate. It has been found general to all the analyzed samples that an oxide tetrahedron forms in four discrete stages: (i) O
1 dominates initially at very low oxygen coverage; (ii) O
2-hybridization begins with lone pair and dipole formation upon second bond formation; (iii) interaction develops between lone pairs and dipoles, and finally, (iv) H-like bonds form at higher exposure and the H-like bonds replace the surface ionic bonds. These processes give rise to the corresponding DOS features in the valence band and modify the surface morphology and crystallography,

624

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

accordingly. Therefore, the events of sp-hybrid bonding, non-bonding lone pair, antibonding dipole and the H-like bonding are essential in the electronic process of oxidation, and in the catalytic reactions involving other electronegative additives as well. 7.2. Bond switching: O-floating and O-diffusing We now turn to the kinetics of O-diffusing in bulk oxidation and O-floating during epitaxial growth of metals on oxygen pre-recovered metal surfaces. It is known that, in the process of oxidation and corrosion of metals, oxygen breaks the adsorption barrier and the metal–metal bond to move into the bulk. The bulk oxidation can be observed with the naked eye as the oxide powders peeling off the metals, known as rusting. However, in epitaxial growth of metals onto oxygen precovered metal surfaces, oxygen atoms always float up to the surfaces. These two opposite processes seem to be very complicated and involve the kinetics and dynamics of oxide bond switching. Understanding the driving force and the kinetic process of oxide-bond switching is particularly important to both fundamental understanding and technological applications. AES [363], ICISS [242], work function measurements [198,364] and STM observations [241,365] have revealed that oxygen atom is always present on the surface of the grown Cu films deposited on an oxygen pre-covered Ru(0001) surface. Under certain conditions (YO=0.2–0.4 ML, T  400 K), the work function, monitored during film deposition, oscillates with a period of one monolayer of copper epitaxial growth. It was explained that oxygen serves as a surfactant for a layer-wise Cu growth on the O–Ru(0001) surface, periodically inducing a high density of islands. For lower YO, densely packed triangular islands are partially covered with an O/Cu surfactant structure. The O–Cu structure is locally ordered in a distorted hexagonal lattice, namely, with the hcp(0001) or fcc(111) features. The structure also consists of O–Cu–O strings inducing the observed corrugation. One possible mechanism proposed by Wolter et al. [241] is that the strain of the first Cu layer promotes the forp mation of the Cu2O-like (32 3) structure. Once the structure is formed, elements of it float onto the top of the growing film and act as a surfactant layer for further Cu film growth. SIMS and secondary electron emission examination [366] revealed that the Cu layers deposited on the c(22)-O/Ni(001) substrate are always covered by an adsorbed layer of oxygen. Wulfhekel et al. [367] also detected similar phenomena in the epitaxial growth of Cu on the O–Cu(111) surface. The oxygen-induced He-scattering features oscillate at  400 K. Interestingly, Yata et al. [368] observed that oxygen atoms segregate back to the surface during the epitaxial growth of Cu on the Cu(001)p p p p ( 22 2)R45 -2O surface, while the ( 22 2)R45 -2O structure is retained at the grown surface. It has been found that co-adsorption of Ni and O on W(110) and subsequent annealing to 500–1000 K leads to segregation of Ni and O, with the formation of Ni crystallites largely < 111> oriented along the surface normal. The heights of these Ni ‘towers’ can be adjusted by varying the amounts of co-adsorbed oxygen and Ni [369].

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

625

Therefore, oxygen floating in the process of Cu/O/Ru(0001), Cu/O/Cu(111) and Cu/O/Cu(001) growth and Ni cluster on O–W(110) surface is driven by a common, yet poorly known, mechanism. As pointed out by Wolter et al. [241] the detailed mechanism of oxygen segregation on the top of a metal surface is still an open question. The ideas reported in the present treatise have provided us with a possible mechanism for the kinetics and dynamics of oxide bond switching. It has been shown in Sections 3.2.3 and 7 that oxidation takes place in four discrete stages in which O
1 forms first and then O
2 follows with sp-orbital hybridisation and lone pair production. VLEED investigation [209] has revealed that annealing the O–Cu(001) surface at a ‘dull red’ temperature removes the lone-pair DOS features from the z0(E) profile, which means that the hybridised sp orbitals of the O
2 dehybridize [99]. Hence, annealing at a certain temperature provides forces that reverse the reaction by oxide bond breaking rather than enhancing. One should note that oxygen floating occurs at  400 K. At the elevated temperature, the processes of O
2 de-hybridizing and oxygen re-bonding tend to occur because annealing activates the bond breaking. At the thermally activated state, oxygen can adjust itself towards a stable tetrahedron with two bonding and two non-bonding orbitals. It is worth mentioning that the lone pair induced dipoles tend to be directed into the open end of a surface. The dipole forms periodically with the epitaxial layer growth, observed experimentally as the periodic change of the work function. Therefore, the strong repulsion between the dipoles and the repulsion between the lone pairs provide the driving force for oxygen to float up under a thermal activation in the process of homo- or hetero-epitaxial growth of metals [67]. We noted that the interaction between the dipoles and oxygen is rather weak ( 50 meV, Section 6). With an external stimulus or under certain circumstances, the dipoles may escape from the bound by lone pairs, and then oxygen reactants have to re-bond to other atoms to form the stable tetrahedron, which is the case of bulk oxidation or rusting. Although O-diffusing and O-floating are inverse processes, they share the same mechanism of bond switching. The mechanism for the oxide bond switching may be extended to bioelectronics such as the folding, signaling and regulating of DNA, proteins and NO, in which the lone pair, dipole and H-like bond may dominate.

8. Application: I. Bond contraction and charge transport 8.1. Introduction The study of nanocrystalline materials with dimensions less than 100 nm is an active area of research in physics, chemistry, and engineering [370,371]. Nanocrystals have large surface-to-volume ratios, and surface effects take on a significance that is normally inconsequential for bulk materials. The small volume can confine free carriers, allowing observation of quantum behavior. While of immense intrinsic interest, the study of nanocrystals is also propelled by technological promise. Various physical properties such as mechanical strength [372], plasticity [373], melting

626

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

temperature [374], sintering ability [375,376], diffusion [377], and electronic structures [378,379] as well as the chemical reactivity [380,381] may be dependent upon particle size. That nanomaterials may display novel or enhanced properties compared to traditional materials opens up possibilities for new technological applications. There have been numerous studies on the relationship between the size of nanocrystals and their properties. However, consistent insight into the size and shape dependency of the nanostructured solid is still lacking. So far, we have developed two essential concepts through the study of surface oxidation, which should be of immediate application. One is the surface bond contraction and the other is sp-orbital hybridization of O, N and C in reaction. Predictions of the functions and potential applications of the bonding events at a surface with chemisorbed oxygen are summarized in Table 16. The bond contraction is not limited to an oxide surface but it happens at any site, where the atomic CN reduces. It is known that the physical properties of a solid are often related to atomic interaction and to the distribution of the valence charges. The spontaneous bond contraction enhances the single bond energy of the remaining bonds of the lower-coordinated atom. Catalytic reaction modifies directly the occupied valence DOS by charge transportation or polarization. Band-gap expansion and surface bond contraction are suggested to play major roles dictating the property changes of a solid. Importantly [382], all the detectable physical quantities relating to the cohesive energy or binding energy density will change with the dimension of a nanosolid of which the portion of surface atoms and the curvature of the surface vary with its dimension. Developments so far have led to the following advances in applications. 8.2. Nano-solid: bond order–length–strength (BOLS) correlation With reduced dimensions of solids or devices, quantum and interface effects become dominant. The size-and-shape dependency of the physical properties of a nano-solid has attracted tremendous interest. The striking significance of nanometric materials is that the conventionally detectable quantities are no longer conTable 16 Summary of model predictions and potential applications of compounds with electronegative additives Events

Characterization techniques

Functions

Anti-bonding (dipole) >EF Holes
STM/S, IPES UPS

Work function-reduction (f) Cold-cathode field emission

Photoemission Raman/FTIR STS, VLEED

Band-gap expansion PL Blue-shift UV detection Polarization of metal electrons High elasticity Far-IR activity

STM/S, UPS EELS -Recovery

XRD, LEED UPS, XPS Surface bond XRD/XPS contraction

Mass transport atomic shift phase change Crystal field cohesive energy Hamiltonian

Potential applications

Surface bond network stabilization Compound formation Origin for the tunability of nano-solids

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

627

stant but they are adjustable by simply controlling the shape and size of the solid. The continuous change of the properties has been leading to a revolution in materials science and device technology. It is possible to tune the physical performance of a device by adjusting the sizes of particles that compose the medium of the device. However, from a fundamental point of view, the origins and the general trend of the property change are yet to be understood, although there are often several models describing a specific phenomenon. As will be shown in the following sections that the CN imperfection induced bond contraction at the surface and the rise in the portion of surface atoms may unify the enormous variations of nano-solid properties. 8.2.1. Principle The finding of surface bond contraction has led to a bond order–length–strength (BOLS) correlation mechanism that is given as follows [383]:  di ¼ c i d "i ¼ c
m ð8:1Þ i "  ci ¼ 2= 1 þ exp½ð12
zi Þ=ð8zi Þ where d is the bulk bond length and " is the single bond energy of atom in the bulk. The i is counted up to at most three from the outermost atomic layer to the center of the particle, as no apparent CN(zi) reduction is expected for i> 3. The m is an adjustable parameter introduced to describe the change of binding energy. Normally, for metals or semiconductors [384], m is around one, whereas, for compounds and alloys, m around four [327,385]. The contraction coefficient ci should be anisotropy and vary with solid dimension because the surface curvature of the solid changes with particle size. Fig. 33 illustrates the BOLS correlation which states that the lengths of the remaining bonds of the lower-coordinated atom reduce (with coefficient ci) spontaneously; as such, the single bond energy ("i) will rise with CN(zi) reduction. The bond energy rise and the reduction of the CN contribute to the cohesive energy of the specific atom, which relates the process of self-organization growth, phase transition and thermal stability. The bond energy rise raises the binding energy density in the relaxed surface region, as the bond number in a circumferential unit area does not change. The binding energy density rise contributes to the Hamiltonian and related properties such as the entire band structure and the surface mechanics. Most strikingly, a recent DFT calculation [386] reveals that for Au, Cu, Pt, Pd, Ni and Ag single atomic chains, the binding energy per bond is (
3 to
1 eV) 2–3 times larger in the chains than the single bond energy (
1 to
0.4 eV) of the bulk fcc structures. Meanwhile, the equilibrium atomic separation contracts by up to 10% (Cu, Ag)  15% (Pt). The binding energy rise due to reduced coordination is attributed to the multi-atom effect that enhances the interatomic interaction of the single atomic chain. This finding concurs with the current BOLS correlation mechanism. For a spherical particle with radius R there are k atoms arranged along the R. The thickness of the ith atomic shell is di. The number of atoms in the ith atomic shell is Ni=4R2i di, Ri=[k
(i
0.5)]d; the total number of atoms of the entire sphere is N=4(kd)3/3. Then the number or volume ratio of the ith atomic shell to that of the bulk is,

628

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 33. Illustration of the BOLS correlation mechanism which states that the bond length reduces with the reduction of the atomic CN(zi); the bond energy of the shortened bond will rise. Large open circles and the square are data after Goldschmidt [101] and Feibelman [72].

   i ¼ Ni =N ¼ 4½ðk
ði
0:5ÞÞd 2 di = 4ðkdÞ3 =3   ci ¼ 3  1
½ði
0:5Þ=k2 ; ðR ¼ kdÞ k Generally, the surface-to-volume ratio for solid (L=0) or hollow (L > 0) a spherical dot (p=3), a rod (p=2) and a plate (p=1) can be expressed as: ip ðk; L; ci Þ ¼

p½k
ði
0:5Þp
1 ci kp
L p

ð8:2Þ

As the k is an integer, the property change will show quantized features at small particle sizes. For a solid, the pi is proportional to pk
1, or pD
1, the dimensional size (k, and D) and dimensionality (p) of the solid. 8.2.2. Application: lattice strain and surface mechanics 8.2.2.1. Lattice contraction and strain energy For a freestanding nano-solid, the lattice constants are often measured to contract while for a nano-solid embedded in a matrix of different materials or passivated chemically, the lattice constants may expand. For example, oxygen chemisorption could expand the first metallic interlayer by up to 10–25% [65] though the oxygenmetal bond contracts [78,110]. Reddy et al. [387] detected with XRD that the lattices of a ZnMnTe ally contract by as high as 7–8% and the amount of contraction decreases with increasing particle sizes. They ascribed such contraction as the atomic density rise in the core of the nanoparticles. Using XRD and Raman, Zhang et al. [388] measured that the lattice constant of Wurtzite CdSe nanocrystals contracts 0.06–0.23% and suggested that the surface tension provides the force driving the surface optimization/reconstruction and hence to minimize the non-radiative traps originated from the dandling bonds on the surface, which cause the red shift of the LO frequency of the nanosolid. Yu et al. [389] found that the mean lattice constants

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

629

of Sn and Bi nano-particles contract with the decrease of particle size. The c-axis lattice contracts more significant than the a-axis lattice. Yu et al. have attributed such lattice variation to the super-saturation of the vacant lattice sites inside the particle. Stoneham [390] ascribed the lattice contraction of Bi and Sn as the effect of surface stress and the compressibility of these materials. Nanda et al. [391] adopted a liquid-drop model to illustrate the lattice strain and indicated that the anisotropic lattice contraction comes from the anisotropy of compressibility and thermal expansion coefficient of the corresponding bulk materials. Despite the different physical mechanisms, all the arguments could fit the experimental measurements very well. In previous work [392], we showed that the mean lattice contraction of an isolated nano-solid is averaged over the entire solid. The change of the mean lattice constant of the entire solid originates from the spontaneous contraction of the bond at the surface and the rise in the surface-to-volume ratio. Actually, surface stress and surface energy results from, not in, the bond contraction as there is no external stimulus (pressure, field, heat, etc.) applied to the surface. For instance, the compressibility,   1 @V  ¼
 ; V @P T and the thermal expansion coefficient,   1 @V  ¼  ; V @T P are intrinsic properties of a solid. They describe the response of the lattice (V / d 3),  DV Dd   DP; T ¼ const ð8:3Þ ¼3 ¼   DT; P ¼ const V d to the external stimulus such as pressure, P, or temperature change, T. The external stimulus simply provides a probe detecting the responses: compression or expansion. However, the degree of easiness of lattice change depends on the shape of the inter-atomic interaction potential. For example, if the pair-wise atom potential well is broader, the compressibility and the thermal expansion coefficient of the atomic pair would be higher, and vise versa. Interestingly, a recent X-ray diffraction study [131] revealed that the compressibility and the phase transition pressure of nanometric PbS particles are no longer constant but change with varying particle size. It was found that the compressibility increases and the transition pressure decreases considerably with reducing particle size. This indicates that reducing particle size not only changes the minimal binding energy but also the shape of the potential well. Actually, the compressibility and thermal expansion vary with the shape of potential well but the pressure (surface stress) depends on the minimal binding energy at equilibrium atomic separation, as will be shown in the next section. Therefore, one needs to identify what the cause is and what the effect would be. It is not applicable to assume a constant compressibility and a constant thermal expansion coefficient in dealing with a nanometric solid. As will be shown in the

630

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

following sections, the surface stress and interface energy are derivatives of the binding energy that is enhanced at the surface by the spontaneous bond contraction. The mean lattice constant in a nano-particle can be derived from the relation: " # " # X X p d ¼ Nd0 þ Ni ðci
1Þd0 =N ¼ d0 1 þ i ð c i
1Þ i¼2

i¼2

and, the relative change of the mean lattice-constant is Dd d
d0 X p ¼ ¼ i ðci
1Þ < 0: d0 d0 i¼2

ð8:4Þ

The relative change in the mean lattice-constant of a particle depends on both the surface-to-volume ratio ip and the bond-contraction coefficient ci. By adjusting the ci values, the trends of lattice contraction of Sn and Bi particles can be quantified, as for the Sn sample shown in Fig. 34(a). The possible errors in measurement may result from the accuracy of shapes and sizes of the particles that could lead to different ip values [382]. The fact that the c-axis lattice contracts more significantly than does the a-axis lattice may results from the anisotropy of the CN or from the bond strength in different directions. Such a bond contraction should lead to an increase in binding energy, or strain energy, at equilibrium atomic separation. Therefore, the premise of spontaneous bond contraction is able to describe the origin of the size-dependency of lattice contraction and its derivatives of a nanometric solid. Fig. 34(b) shows the distribution of nearest-neighbor (NN) atomic distances (bond length) for relaxed Ag nanocrystals [393]. The NN distance, for the 2.0, 2.5, and 3.5 nm size crystals, becomes shorter than the bulk one. For the 5.0 nm crystal, 60% of the atoms have the bulk value but 40% have a shorter NN distance. Fig. 34(c) shows the size-dependence of the average NN distance for the relaxed Ag, Cu, and Ni nanocrystals. One can see that the average NN distance for the three elements, Ag, Cu, and Ni, is shortened by as much as 1.6–2.0% for small nano-crystals and about 0.6% for relatively large ones, as compared with the bulk value. Using extended X-ray-absorption fine structure, Montano et al. [394] measured the microstructure of Cu microstructures (0.7–1.5 nm mean diameter) and found the Cu–Cu bond contracts from the bulk value of 0.255–0.221 nm for Cu2 dimer and the structure remains the fcc structure. The trends of NN distance contraction in Fig. 34(c) agree with the current modeling predictions [see sample in Fig. 34(a)]. For Ag nanocrystals, Kara and Rahman [393] found that the atomic force constants experience a stiffening of up to 120% when compared with that for bulk Ag, which should be one of the consequences of binding energy enhancement, as discussed. By examining the bond length between neighboring atoms in Ag, Cu, Ni, and Fe in different structures with different coordination numbers, Kara and Rahman [393] pointed out that these elements belong to a class of elements showing a strong bondorder–bond-length correlation [72]. Because of this correlation, the bond length between an atom and its neighbors would decrease with decreasing coordination.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

631

Fig. 34. (a) Simulation of the experimental observations (broken lines) of Yu et al. [389]. (b) Distribution of bond length (nearest neighbor distance) in relaxed Ag nanocrystals with diameters 2.0 (dashed line), 3.5 (long-dashed line), and 5.0 nm (dot-dashed line); the bold vertical line represents the NN distance in the bulk. (c) Relative average NN distance as a function of the nano-crystalline diameter for Ag (solid line), Cu (dot-dashed line), and Ni (long-dashed line) [393]. Agreement [392] in (a) is acceptable within error by introducing different bond contraction parameters, as indicated. An atomic radius of 0.162 nm was used for the Sn atom.

Thus the dimer bond lengths (2.53, 2.22, 2.15, and 2.02 A˚, for Ag, Cu, Ni, and Fe, respectively [72]), are shorter than the NN distance in their respective bulk values by 12.5% for Ag, 13.2% for Cu, 13.6% for Ni, and 18.6% for Fe. Note that Fe shows the strongest bond-order–bond-length correlation, as far as the dimer bond-length is concerned. The pattern is similar for the surface relaxations of the top layer atoms for these elements, for several crystallographic orientations. Because of the lowered CN at the surface, elements show a contraction of the interlayer separation at the surface. Experiments and first principle calculations agree that the interlayer

632

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

separation between the first and the second layers contracts for these four elements. Bond contraction has also been detected as (multi) layer relaxation from a number of clean metal surfaces, as listed in Table 3. It was predicted [72] that bonds to under-coordinated Ti or Zr should be unusually short, given the small ratio,  0.7, of dimer bond length to NN distance for these elements. The argument and findings agree with the premises of Pauling and Goldschmidt, though the extent of contraction may vary from situation to situation. According to Pauling and Goldschmidt, all the elements will suffer the global shrinkage because of the reduced CN at surfaces, and the global shrinkage has no selectivity of elements. 8.2.2.2. Surface stress and Young’s modulus Surface stress links the microscopic bonding configuration at an interfacial region with its macroscopic properties [395,396]. It plays a central role in the thermodynamics and acoustics of solid surfaces. During the last decade, increasing interest has been paid to processes that are strongly influenced by surface stress effects such as reconstruction, interfacial mixing, segregation and self-organization at solid surfaces. However, detailed knowledge about the underlying atomistic processes of surface stress is yet lacking [395,396]. The BOLS correlation premise may provide us with insight into the physical origin of the enhanced surface stress and Young’s modulus, as discussed as follows. The CN-derived bond contraction has been defined as di ¼ ci d by introducing a contraction coefficient ci < 1. Hence, a specific bond length and the correspondingly local properties such as interatomic potential u(di) at equilibrium atomic separation will change: Ddi ¼ ci
1 < 0; d Duðri Þ  D"i ¼ c
m
1>0  ¼ i uðrÞ r¼d "

ð8:5Þ

The Young’s modulus, B, and the surface stress or hardness, P, can be derived as [327]: @uðrÞ  " / 3;  @v r¼d d @P  @2 uðrÞ  " ¼v / 3 B ¼
v   @v r¼d @v2 r¼d d P¼


ð8:6Þ

Both the P and B share the same dimension and they depend uniquely on the binding energy at equilibrium atomic separation, or the bond length. In another word, the hardness is a sum of bond energies per unit volume [397]. The derivative is independent of the exact form of an interatomic potential. Therefore, the relative changes of the surface stress (corresponding to hardness) and the Young’s modulus can be expressed as:

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

DB DP D" Dd
3 ¼ c
m ¼ ¼
1
3ð c i
1Þ i B0 P0 " d

633

ð8:7Þ

This relation implies that the B and P at a surface are both higher than the bulk values due to the spontaneous bond contraction and that the tensile surface stress dominates at an elemental solid surface [395,396]. A study to compare the hardness and Young’s modulus of the surface with the bulk value was carried out using nanoindentation [327]. The force-depth profile in Figure 35 shows a maximal 200% hardness at  10 nm depth of a TiCrN thin film (2 mm thick) deposited on tungsten carbide substrate. The peak shift is due to the surface roughness (Ra=10 nm was confirmed with an atomic force microscopy). The geometrical shape of the nanoindenter tip determines the subsequent decrease in the profile with depth. Shi et al. [398] have detected 180% increase of both the hardness and Young’s modulus of the bulk value of amorphous carbon thin films deposited on silicon substrate. Caceres et al. [361] have also observed the similar enhancement of both the Young’s modulus and the hardness at an AlGaN surface. Solving Eq. (8.7) with the measured value of DP=P ¼ 1; we found the ci values of 0.770, 0.824, 0.860 and 0.883 that correspond to m=1, 2, 3, and 4. Therefore, m=3–4 is reasonable, if the surface bond contracts by 12–14%, according to Goldschmidt [101]. The m> 1 value is an intrinsic property of compounds and alloys [399]. It is not surprising that, if one can establish the functional dependence of any physical property, Q(r, "), on the atomic distance or its derivatives, it is possible to express the size-induced change of the property Q(r, ") of a nano-solid by using the quantized statistical relation: DQðDÞ X DQi ðdi ; "i Þ ¼ ; ð8:8Þ i Qð1Þ Q i43 P  43 Eqs. (8.3 and 8.8) indicate that, for a solid nano-rod i¼1 i / 1=k DQ=Q depends on the inverse radius (1/k) of the rod; while for a hollow nanotube with

Fig. 35. Nanoindentation depth profile of a TiCrN thin film on a tungsten carbide substrate with roughness of 10 nm. The hardness increases by 200% compared to the bulk value [327].

634

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

P  43 limited number of walls  1, Q/Q approaches a constant value and the i i¼1 Q(D) is much greater than that for the corresponding bulk value. These predictions agree with what has been discovered by Wong et al. [400]. By using atomic force microscopy, they found that the multiwall carbon nanotubes are about two times as stiff as the SiC nano-rods and that the strengths of the SiC nano-rods are substantially greater than those found for large SiC structures (600 GPa). The Young’s modulus is 610 Gap and 660 Gap for SiC rods of 23.0 and 21.5 nm across, respectively. For hollow carbon tubes, the modulus is 1.28  0.59 TPa with no apparent dependence on the diameter of the nanotubes. Zeng et al. [401,402] have found that the hardness and elasticity of nanometric TiN/CrN and TiN/NbN multi-layered thin films increase with reducing the structural wavelength (optimal at 7.0 nm). This may be indicative of the bond contraction at the surface and interfaces. These results provide direct evidence for the essential concept of surface-bond contraction that enhances the mechanical strength at a surface and of nanobeams. It is important to note that, in the Nanoindentation test, errors may arise due to the shape and size effect of the tip. However, such errors may affect the precision of the derived m values but never affect the origin and the general trend of the measurements. By taking the relative change of the quantity into account errors due to the measuring technique should be removed. 8.2.3. Other applications The bond contraction at the surface has indeed enormous effects on various physical properties of nano-solids. A theoretical calculation conducted by Ian and Huber [403] reveals that the Fe–W and Fe–Fe interlayer contracts by
10% compared to the corresponding bulk Fe–W and Fe–Fe interlayer spacings. Compared to the Fe bcc bulk moment of 2.2B, the magnetic moment for the surface layer of Fe is enhanced (i) by 15% to 2.54B for 1 ML Fe/5 ML W (110) and (ii) by 29% to 2.84B for 2 ML Fe/5 ML W (110). The inner Fe layer for 2 ML Fe/5 ML W(110) has a bulk-like moment of 2.3B. The significant surface relaxation of Fe(310) (
12%) [404] and Ni(210) (
12%) [405] has also been found to enhance the atomic magnetic momentum by up to 27%. The surface relaxation of the (110), (211), (311), (511), (331) and (221) surfaces of Al, Ag, Cu and Pd has been found to lead to a shift in the frequencies of the surface states and to a change in the number and localization of the states [406]. It has been found [407] that the vibrational free energy and the heat capacity of the step and the terrace atoms on a Cu(711) surface are sensitive to the local atomic environment. The vibrational contribution to the excess free energy of the step atoms near room temperature is a significant fraction of the kink formation energy. Batra [68] concluded that the Al(001) surface relaxation has an effect on the total bandwidth for the relaxed monolayer, which is about 1.5 eV larger than the value for the bulk truncated monolayer. The cohesive energy is increased by about 0.3 eV per atom upon relaxation. The concept of surface-bond contraction has been incorporated into a number of other observations [399]. For instance, the band-gap expansion of nanometric semiconductors will lead to the reduction of dielectric constant and hence the blue shift of the photo-absorption edges of a nanometric semiconductor [384]. The sur-

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

635

face stress enhancement has an influence on the Gibbs free energy that determines the transition behavior of ferroelectric [408] and pyroelectric [409] properties of nanometric PZT oxides. It has been shown that the CN-imperfection-induced relaxation influences not only the band-gap of a solid but also the entire band structure of the nano-solid including the core-level shift, bandwidth and the band tails [383]. The electronic structure of nano-composites was proposed to deviate from one of the coarse-grained materials if a large volume fraction consists of electronically modified regions at interface boundaries [410]. Typical samples of applications are introduced below. Ferroelectrics constitutes a special group of materials with high dielectric constants. The dielectric properties vary with external stimulus, such as temperature, electric field, pressure and particle size. Introducing such materials into a photonic crystal could thus help modulate its band-gap, which is not only sensitive to the excitations but also tunable by varying the particle sizes. In line with this, a silicon– dioxide colloid crystal infilled with barium titanate (BaTiO3) has been synthesized [411] with a method combining a self-assembly process and a sol-gel technique. In the vicinity of the ferroelectric phase-transition point of BaTiO3 (100–150  C), the photonic band-gap of the resulting assembly exhibited strong temperature dependence. At the Curie point, a 20-nm red shift in the band-gap has been detected. This is also where optical transmittance is at a minimum. The tuning of the band-gap could be used not only for simple on-off switching, but also in devices requiring more localized control of light propagation. The atomic cohesive energy, or the sum of the binding energy of a single bond over the CN of an atom, determines the thermal energy required to melt the atom of a solid. The critical thermal energy also causes the order-disorder transition of a ferromagnetic nano-solid. This principle has enabled the melting behavior of (Cu, Au, Ag, Al, InAl, Sn, etc.) nano-solids [412] to be consistently formulated, which agrees with experimental observations. The predicted size dependency [413] of the Curie temperatures for the spin–spin order–disorder transition of ferromagnetic (Fe, Co, Ni) solids also agrees with measurements. The approach favors closely the atomistic model of Jiang and Shi [414] who relate the amplitude of atomic vibration to the melting behavior of a surface and the surface energy. With few adjustable parameters, the model based on Lindemann and Mott premises for the relationship of vibrational melting entropy and melting temperature has applied well to the sizedependent melting of compounds [415], metals [416], nanotubes [417], polymers [418], glasses [419], inert gases [420], ice [421], and semiconductors [422]. In conjunction with Jiang’s approach, the current BOLS premise could provide deeper insight into the physical origin and the general trends of the phase transition behavior of a nanometric solid. Matching the BOLS modeling predictions to the Auger photoelectronic coincidence spectroscopies (APECS) from surfaces or to the size dependence of the APECS from nanosolids derive information about the single energy levels of an atom isolated from solid and their shifts upon solid formation as well as the bonding of a monatomic chain [423–425]. With the measured values of coalescent temperature of the tip-end (Tm=1593 K) [426] and the known product of the Young’s modulus with the wall (bond) thickness (Yt ffi 0.3685 TPa.nm) [400,427,428] for a

636

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

single-walled carbon nantube (SWCNT) as well as their functional dependence on atomic coordination, bond length and bond energy, represented by the BOLS correlation, a single C–C bond in the SWCNT has been determined to be  0.142 nm thick and  0.116 nm long, associated with  68% bond energy rise in magnitude [429]. The melting point of the tube wall is found  12 K higher than the tube terminal. Besides, the predicted diameter-dependence matches the observed trends of Tm-suppression and Y-enhancement of nanorod and multi-walled carbon nanotubes (MWCNTs). Findings provide not only consistent insight into the origin of the Y-enhancement and the Tm-suppression of the NTs but also an effective method obtaining information that is beyond the scope of direct measurement using currently available techniques. Further work towards a single and simple model aiming to generalize the shape and size dependency and reconcile all the reasonable models for nano-solids to the effect of CN imperfection is in progress. 8.3. Catalytic effect on band-gap expansion Solids made of elements in group-II and group-III are conductors without apparent gap existing between the conduction and the valence band of the bulk. Upon introducing the electronegative additives to the conductors, a band-gap EG is generated due to the mechanism of electron-hole pair production, as discussed in Section 2.3. The width of a band-gap can be detected by measuring the wavelength of the absorbed or emitted photons [430]. In the following sections, we will show that the band-gap expansion mechanism not only leads to a new finding of the blue light emission of a PZT oxide but also provides a consistent understanding of the blue shift in photoluminescence of both oxides and nitrides. We also show that both the physical-size effect and the catalytic effect enhance each other in the band-gap expansion. 8.3.1. Blue light emission of PZT The band-gap-expansion mechanism implies that it is possible to discover or invent new sources for light emission with a desired wavelength by controlling the extent of catalytic reaction. Oxide ceramics are much cheaper and easier to produce compared with the group-III nitrides. Driven by this motivation, we have searched for visible light emission during the sintering of PZT [Pb(ZrxTi1
x)O3] ceramics [431]. Intense blue-light was indeed found to emit from the sample under Ar+ ultraviolet (UV) irradiation. This was first observed in a vacuum chamber in which we were depositing the conducting Au layer by Ar+ sputtering on to the samples for scanning electron microscopic characterization. The light is stable after a period of 2-year ageing [432]. The Pb(ZrxTi1
x)O3 pellets were synthesized using the traditional sintering process [431]. The dried slurry of the ball-milled mixture of PbO, TiO2 and ZrO2 powders was pressed into pellets and then calcined in air at 900  C for 2 h. After calcination, the pellets were crushed and wet-ball milled again. The pellets were sintered in air at 500  C for 2 h, and subsequently 1200  C for another 2 h. For all the sintering profiles, the heating and cooling rate was 5  C/min.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

637

The photoluminescence (PL) and photo-excitation (PE) spectra and the temporal profiles were measured using a Spex Fluorolog-3 spectrometer in ambient atmosphere. The excitation beams were applied directly onto the surface of the pellets and the signals were collected in front of the surface. Fig. 36(a) plots the PL (I–l) spectra of the PZT samples. There is only one broad band located at 475  50 nm. The single excitation mode implies that the mechanism of direct band-gap transition dominates. The corresponding excitation spectra (monitored at 475 nm PL) in Fig. 36(b) exhibit that the excitation band is in the range of 305  45 nm. Changing

Fig. 36. (a) The photoluminescence (PL) spectra and (b) the photo-excitation (PE) spectra for the PZT ceramics obtained at room temperature. The peak corresponds to 2.65 eV (467 nm) and the FWHM is 0.5 eV. The lifetime (c) varies from 0.03 to 0.6 ms [432]. (d) Room temperature blue PL of PZT observed under Ar+ UV radiation in a glass chamber for film deposition.

638

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 36. (continued)

the x value causes a negligible shift of both the PL and PE peaks. The fluorescence decay curves of the PZT samples show that the lifetime of the photons varies markedly from 0.03 ms (at x=0.5) to 0.6 ms (at x=1.0). The 0.6 ms lifetime is much longer than that for other reported defect luminescence. However, the mechanism for the unusual temporal behavior is not clear though x=0.5 corresponds to the phase boundary between the tetragonal and the rhombohedral ferroelectric in the phase diagram [431]. This finding not only justifies the BBB modeling predictions but also adds another advantage to the PZT oxide ceramics. The BBB modeling argument provides the possible mechanisms for the band-gap production and the x-value effect on the PL and PE peak shift. The energy difference between the PE and PL gives information about electron–phonon interaction termed as Stokes shift, in conjunction with non-irradiative combination of the carriers [430]. As it is known, the band-gap depends on both the valence DOS distribution and the interatomic interaction. The mechanism of electron-hole pair production in Section 2.3 anticipates that a band-gap generates between the conduction and the valence band upon reacting with electronegative additives. On the other hand, the width of the band-gap depends on the first Fourier coefficient of the

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

639

crystal field. Further, the crystal field is a function of atomic distance and the quantity of charge transportation during the reaction. There are several means for enhancing the interatomic potential. For instance, the spontaneous lattice contraction enhances the binding energy density in the relaxed region and the corresponding properties. Moreover, solid transits from one phase to another that is more stable, which enhances the atomic interaction. Even further, defects production in materials processing may form centers of photoluminescence and influence the crystal field and charge distribution as well. Therefore, besides the direct gap expansion by hole-production, modifying the crystal field by chemical reaction and lattice revision at the surface or through phase transition will expand intrinsically the band-gap. Except for the similar electronic configuration (d4), Ti (3d4;
=1.5; r=0.1467 nm; r(4+)=0.068 nm) differs slightly from Zr (4d4;
=1.3; r=0.1597 nm; r(4+)=0.079 nm;) in the electronegativity (
), metallic and ionic radius (r). Adding an element with a smaller atomic radius could raise interatomic potential energy, which expands the band-gap intrinsically; increasing atoms of lower
value means that a higher amount of charge transportation to oxygen would take place, which would expand the band-gap extrinsically. The smaller radius of Ti and the lower
value of Zr may compensate for each other in the gap expansion when the x-value is adjusted. The atomic-size–electronegativity compensation mechanism may explain why the bandgap of the PZT is less sensitive to the x-values, though the actual mechanism should be much more thoroughly investigated. It was found that samples of ZrO2 and TiO2 or their mixtures without Pb presence give no light though the samples were prepared under the same processing conditions. This means that the Pb plays some unclear yet important catalytic roles in determining the PL features. On the other hand, the temporal behavior of the photons is a high challenge. The actual mechanism for the PL of PZT should be much more complicated and an understanding of the role of Pb and the temporal behavior would be necessary. Visible light emission of other oxides has been reported [433] such as SrCu2O2 [434], Sr2CeO4 [435], SiON [436], and ITO/SiO [437]. The current band-gap expansion mechanism may provide interpretation for the photoluminescence of these oxides. 8.3.2. O-induced blue-shift in PL Fig. 37 shows the room temperature PL spectra of (a) the Al0.3Ga0.7As/GaAs ‘quantum-well’ lasers [438], (b) the SiO2 films containing Si nanoparticles (2–5 nm in diameter) [440] and (c) ZnO/Ga2O3 nanofibers [439]. Rapid thermal annealing at 920  C for 120 s causes a blue shift and intensified PL spectra of the devices coated with an anodic SiO2. The ZnO/Ga3O2 fibres were synthesized at 850  C in a tube furnace and atmosphere using graphite powder as catalysis. Adding Ga2O3 improves the quality of the PL. Thermal oxidation of the Si/SiO2 system also shifts the PL to high energy but the intensity of the PL peak is reduced significantly. The blue shift of the PL peak is often attributed to a ‘quantum-size’ effect in which a reduction of the average nanocrystal size leads to emission at shorter wavelengths [440]. The width of the emission band is attributed to a wide distribution in nanocrystal sizes.

640

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 37. Room temperature PL spectra of (a) the GaAs–Al0.3Ga0.7As ‘quantum-well’ laser with and without anodic SiO2 annealed at 920  C for 120 s [438]. (b) 100 nm thick SiO2 films containing Si nanocrystals oxidized for 0, 3, 10, 15, 20, 25, and 29 min at 1000  C. (c) ZnO–Ga2O3 nanofibres (150 nm in diameter) synthesized at 850  C in tube furnace using graphite powder as catalysis [439], by heating the mixture of ZnO+Ga2O3+C (Curve I), and ZnO+C (Curve II). For each spectrum, the corresponding oxidation time is indicated. An arrow indicates the band-gap wavelength of bulk Si. The inset shows schematically the change in the nanocrystal size distribution upon oxidation [440]. Thermal annealing of the oxidized laser causes a blue shift and enhancement of the PL spectrum.

According to the band-gap expansion mechanism developed in this work, it is easy to understand that the oxidation-induced blue-shift of the PL is purely due to the effect of electron-hole pair production during annealing when oxygen switches its bond and moves into the bulk. On the other hand, annealing also enhances the socalled ‘quantum-size’ effect in the device with oxygen diffusion because the nonoxide core reduces its size upon being oxidized of the outer atomic layers [438]. Oxide emission cannot be detectable within the infrared wavelength range. There are two possible effective ways that are hence applicable to tune the wavelength of the PL of a compound solid. One is to reduce the physical size and the other is to enhance the catalytic reaction of the system. The rise in the portion of surface atoms in conjunction with surface bond contraction enhances the crystal field in the surface region and hence the Hamiltonian that determines the entire band structure including the band-gap. Catalytic reaction removes electrons from the top of the valence band, which widens the gap directly. It is known that the quantum efficiency (YPL) of photoluminescence follows the relation [430]: YPL=Pr/(Pr+Pnr),

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

641

where Pr and Pnr are the probability of radiative and non-radiative combination of electron-hole pairs. Electron transiting from the conduction band tail to the valence band tail is responsible for the radiative combination, while electron transiting from the conduction band tail to the defect states that are located within the band-gap is responsible for the non-radiative irradiation [441]. The decrease of the PL intensity of Si in SiO2 indicates that oxidation creates defects in this particular system due to the nonbonding lone pair production. 8.3.3. PL of III- and IV-nitride Similarity in the band-gap-enlargement of oxides and nitrides has been widely noted. Fig. 38 shows the band-gap enlargement of (a) III-nitrides and (b) amorphous Ge- and Si-nitrides. It is noted that nitrogen incorporation into the group-III metallic solids generates a considerably large band-gap when the compound forms. The width of the band-gap depends on the bond length [442] or the electronegativity of the corresponding element (
Al=1.5,
Ga=1.6, and
In=1.7). Nitrogen expands the band-gap of the semiconductive a-Ge and a-Si from  1.1 to  4.0 eV [443], depending on nitrogen content. Nitrogen also widens the band-gap of amorphous carbon (aCNx : H) [444]. Corresponding band-gap changes of a-CNx : H films have been observed in the He-II valence band spectra showing a recession of the leading edge of more than 0.9 eV while the optical band-gap widens from 0 to more than 1 eV. Reynolds et al. [445] suggested that the green-band of ZnO and the yellow-band of GaN share some common yet unclear mechanisms. Chambouleyron et al. [443] related the band-gap expansion of a-Ge:N and a-Si:N compounds to the substitution of Si–Si or Ge–Ge bonds by stronger Si–N or Ge–N bonds. It was suggested that, as the N content increases, the nitrogen lone-pair band develops and that the lone-pair band dominates the valence-band maximum as the stoichiometry is reached. The largest optical band-gap is obtained for the stoichiometric compound. On the contrary, for smaller N content, Si–Si or Ge–Ge bonds dominate the valence band maximum. For amorphous semiconductors, it is generally accepted that the transition of carriers is between the conduction-band tail and valence-band tail states. Luminescence spectra [430] of the a-Si:H showed that the n-type (phosphorous) doping shifts the luminescence peak of the a-Si:H from 1.1 to 0.81 eV, and the p-type (boron) doping shifts the peak to 0.91 eV. This can be easily understood in terms of impurity levels. The shallow n-donor levels and the deeper p-acceptor levels are located within the initial band-gap (1.1 eV width) near to the band tails, which should narrow the gap, as observed. However, the luminescence peak of the a-Si:N:H compound moves to higher energy with increasing nitrogen concentration [446]. The broadened band-gap through nitridation could not be explained in terms of the traditional donor effect, though the nitrogen addition is always believed as n-type doping. Clearly, the band-gaps of metal oxides, III-nitride, and IV-nitride are enlarged by the same hole-production mechanism proposed in the current exercise (Section 2.3). The change of bond nature and bond length has an effect on the crystal field, and consequently, the width of the band-gap; charge transport in the reaction re-populates with valence electrons of the host materials.

642

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Fig. 38. (a) Band-gap expansion of the group III-nitride [442] and (b) the N-concentration dependence of the optical band-gap (ETauc) of a-Ge:N:H [447,448] and a-Si:N:H [449] thin films [443]. Nitrogen generates a band-gap to the metallic group III materials and the width of the band-gap depends on the bond length or the electronegativity.

8.4. Joint size and catalytic effects: PL of nanometric SiO2 Owing to the new phenomena of blue shift in photoluminescence, the investigation of nano-crystalline or porous Si oxide has attracted tremendous interest. Nanostructured SiO2 compounds also have potential applications as low-dielectric layer, in the deep submicron integrated circuit (DSIC) technology. The emitted wavelengths of the nanometric Si oxide varies in a wide spectral range from infrared to the ultraviolet, depending on the shape and size of the particles and the process of

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

643

chemical treatment [450]. There are several conflicting models concerning the mechanisms responsible for the band-gap expansion. Existing models include quantum confinement [451,452], phonon-assisted free-exciton collision [453], impurity centers [454], surface alloying [455], cluster interaction and oxidation effects [456]. According to the quantum confinement theory [451], electrons in the conduction band and holes in the valence band are confined spatially by the potential barrier of the surface. Because of the confinement of both electrons and holes, the lowest energy optical transition from the valence to the conduction band increases in energy, effectively increasing the band-gap. Within a simple effective-mass approximation, the confined gap is given as [457,458]: EG ðRÞ ¼ EG ð1Þ þ where

1 

h
2 2 1:786e2 þ 0:284ER
2 2R "r R

ð8:9Þ

¼ m1 þ m1 , being the reduced mass of an electron-hole (e-h) pair, is an h

e

adjustable parameter, where ER is the Rydberg (spatial correlation) energy for the bulk semiconductor:   e4  ¼ 13:56 2 ER ¼ ðeVÞ ð8:10Þ "r m e 2"2r "20 h
2 The effective dielectric constant "r and the effective mass, , describe the effect of the homogeneous medium in the quantum box. For CdS [459], the "r=5.5, me=0.19, and mh=0.8. The energy of the freely moving carriers is responsible for the band-gap expansion according to this expression. The width of the confined band-gap grows as the characteristic dimensions D of the crystallite decrease. One question may arise whether the carriers move indeed freely inside the solid that is far greater than the localized length (several Angstrom) and the nano-solid contains numerous atoms each of which acts as a trapping center? The free-exciton collision [453] model suggests that the excitation laser heats the free excitons that collide with the boundaries of nanometer-sized fragments. According to this model, laser heating the free-excitons up to a temperature in excess of the activation energy required for the self-trapping give rise to the extremely hot self-trapping excitons (STEs). Because the resulting temperature of the STEs is much higher than the lattice temperature, the cooling of STEs is dominated by the emission of lattice phonons. However, if the STE temperature comes into equilibrium with the lattice temperature, the absorption of lattice phonons becomes possible. As a result, the blue shift of the STE-PL band is suggested to originate from the activation of hot-phonon-assisted electronic transitions. The blue shift of the STE-PL band depends on the temperature of laser-heated free-excitons, which, in turn, is determined by the size of nanometer-sized (silica example was considered only) fragments. This happens because the temperature (kinetic energy) of the laserheated free-exciton increases with the number of collisions within the boundary of confined regions, which tends to be higher with decreasing the size of the silica fragments in nanoscale materials. The energy gained by laser heating increases with

644

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

decreasing nano-particle diameters in an exp(1/D) fashion. It is indicated that the blue shift of the PL bands in nanoscale materials in general does not need to be related always to a quantum confinement effect. A recent first-principle study by Luo et al. [460] of the Si/O(001) multi- (n Si+1O) layers suggested that the quantum-confinement effect does not provide the observed optical functionality. They referred the oxygen enlarged band gap of Si to the effects that are of atomic origin. It is suggested that by adding oxygen atoms, one introduces new non-tetrahedral bonds and extra electronic states. These new states interact with the Si host states, resulting in changes in the energy levels and in the characteristics of the host wave functions. The added oxygen-layers are suggested to have two effects. First, the strongly electronegative oxygen atoms (assumed as rigid spheres without valence alternation) form a potential barrier that causes quantum confinement on the Si host states. This will lower the energies of the Si-valence band states and increase the energies of the Si-conduction band states, thus increasing the band gap. The second effect of the added oxygen-layer is the lowering of the crystal symmetry that can lead to significant coupling between the Si states that otherwise do not couple, as well as between the Si and oxygen states. The coupling between the Si states will be large if they have the same folded symmetry, and their overlaps with the oxygen state are large. Coupling results in level repulsion that pushes up the valence band maximum and lowers the conduction band minimum, thus lowering the band gap, opposite of the quantum confinement. The combined effect of quantum-confinement, level repulsion and splitting makes the band structure of the (8Si +1O) layers a direct gap of 1.4 eV in the center of Brillouin zone of Si. Here we show that the PL frequency shift of Si oxide nanoparticles is determined by the joint effect of oxidation and surface-bond contraction—BOLS and BBB mechanism [382]. It is known that the frequency-shift of PL reflects the band-gap change of the system. The band-gap depends on the crystal field or a sum of the binding energies over the entire solid. The interatomic potential, or binding energy, depends on atomic distance and charge quantity of the atoms [382]. Shortened bond length and alternatively charged ions enhance the crystal field and hence the bandgap. Chemical reaction also enhances the crystal field by weakening the ‘screen effect’ due to the charge repopulation upon reaction. The process of charge transfer further widens the band-gap directly by emptying the occupied DOS below EF, as specified in the BBB model (Section 2.3). Considering the effect of surface relaxation, the Hamiltonian of a solid is defined as: 2 2 ^ ¼H ^0þH ^ 0 ¼
h
r þ Vatom ðrÞ þ Vcrystal ðr þ RC Þ½1 þ Dsurf  H 2m

ð8:11Þ

where the Vatom(r) is the intra-atomic trapping potential of an isolated atom and the ^ 0 ¼ Vcrystal ðr þ RC Þ½1 þ Dsurf  is the periodic potential of the surrounding atoms, H i.e., the inter-atomic binding potential or crystal field. RC is the lattice constant. surf is the contribution from surface relaxation:

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Dsurf ¼ and, ¼

645

X X Dvðdi Þ

þ ¼ i i c
m
1 þ ; i v ð d Þ i43 i43

nVðDÞ N 2 v ð dÞ

ð8:12Þ

where v(d) is the binding energy density and v(di) / "i is proportional to the single bond energy as no bond number decrease occurs between two circumferential atomic layers. The "i is the bond energy at equilibrium atomic separation. describe the contribution from cluster interaction V(D), which becomes insignificant with increasing particle dimension D. The particle size determined band-gap and the core-level shift of a nano-solid are expressed as follows [382]: X

DEG ðDÞ DEcore ðDÞ ¼ ¼ Dsurf ¼ i c
m
1 þ

ð8:13Þ i EG ð1Þ Ecore ð1Þ i43 Eq. (8.12) indicates that the band-gap expansion and the core-level shift of a nano-solid result from the surface-bond contraction (ci) and the rise in surface-tovolume ratio ( i) that depends on the shape (p, L) and size (k) of the particles as well as the type of atomic interaction (m). Fig. 39 shows that the modeling predictions agree with observations of the sizedependent PL energy of SiOx nano-particles. Bond contraction of 12% and 4% was

Fig. 39. Agreement between prediction [385] and observations [474] of the PL blue shift of nano-sized Si oxide. Data 1–13 are after Refs. [468–480].

646

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

applied, respectively, to the first two atomic layers of a spherical dot. It can be seen that the curve of m=4 follows better the trend of observation than the curve of m=2 [385]. The resultant m value describes well both the size and the oxidation effects of smaller particle sizes. The cluster interaction seemed to play an insignificant role in the band-gap expansion. Based on the same premise, consistency between predictions and observations of the blue shift in the PL of CdS and CdSe nanosolid has been achieved [461]. This agreement evidences that a surface bond contracts more than it does on a flat surface when the surface curvature increases and that the Cd–Se bond contracts faster than the Cd–S bond when the solid approaches to the lower end of size limit. On the other hand, XPS investigations of O–Cu [462], O–Sn, O–Ta [463], ZnS [464] and CdS [465] nanoclusters have revealed additional features at higher binding energy besides the well-known chemical shift (see Fig. 40). A core-level shift towards higher binding energy of the clean Ru(0001) surface has also been detected with high-resolution XPS [320]. Both the components of core-level shifts result from the increase of crystal field experienced by the core electrons, which are dominated by different mechanisms (the charge transportation and the bond contraction). Zacchigna et al. [159] have found that a clean Rh(001) surface has a core-level shift of
0.65 eV relative to the bulk value, while with O addition, the shift extends
0.40 eV further towards higher binding energy. Photoemission from highly oriented pyrolytic graphite by Balasubramanian et al. [466] reveals two C 1s components

Fig. 40. XPS Cu 2p3/2 core-level shifts of O–Cu nano-particles of (a) 4.0 (b) 6.0 and (c)  25 nm in diameter which indicate the increase of binding energy with reduced particle size, with, as yet, unclear mechanism. The slight decrease (10
1 eV level) of the separation between the main peak and the satellite peak with reducing particle size and the slight increase of the peak intensity Is/Im were ascribed as the increase of ionicity of the O–Cu bond [462].

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

647

separated in binding energy by 0.12 eV. The high binding energy component is ascribed as atoms in the outermost atomic layer and the other to atoms in deeper layers. This interpretation is based on the relative intensity change with incident beam energy and the incident angles. The intensity of the higher binding energy line (surface feature) increases with lower photo energy and smaller incident angles. These size-induced core-level offsets may provide further evidence for the spontaneous bond contraction at the surface that contributes directly to the crystal field [467]. The band-gap expansion and core-level shifts are consequences of the binding energy that can be enhanced by either surface-bond contraction and charge transportation. Strikingly, these effects enhance each other on the electronic structures and related properties of a nano-solid. Hence, the overall PL features and the corelevel shifts of nano-compounds can be consistently understood in terms of bond contraction and chemical reaction effects. The former defines the band-gap; the latter widens it by charge transportation. The above analysis may lead to a conclusion about the band-gap generation of nanometric compounds. Although it cannot be accounted for using the traditional wise of impurities in semiconductor physics, the phenomenon of blue shift in the light emission from the compound semiconductors can be consistently explained with the current BBB correlation for both oxides and nitrides [92], and the size effect for nanometer-sized particles. The occupied DOS at the top of the valence or conduction band are emptied by electron transportation for bond and anti-bonding dipole formation. The width of the generated band-gap depends on: (i) the DOS distribution of the host, (ii) the electronegativity difference and, (iii) the stoichiometry of the compound. 8.5. Work function reduction: cold cathode field emission 8.5.1. Current understanding Field emission from semiconductors in general can originate from the valence band, gap states, or the conduction band, from an accumulation layer at the surface. Previous models have attributed field emission from diamond and DLC to the low affinity of diamond [481,482], the so-called antenna effect of conducting channels [483], emission from defect gap states [484], and band bending at depletion layers [485]. Emitters doped with N or O could reduce the threshold of the field emission. Experimental evidence [486–489] has shown that an addition of N to CVD polycrystalline diamond thin films significantly reduces the threshold of cold-cathode emission of the diamond. The threshold of the N-doped diamond is even lower than those doped with boron and phosphorous. It has recently been found that boron nitride coated graphite nanofibers also emit electrons at much reduced (from 1.5 to 0.8 V/mm) the threshold with high (102 level) intensity compared with the uncoated carbon fibers [490]. It is explained that introduction of the BN nanofilm leads to a significant reduction in the effective potential barrier height [491]. A tendency of N-buckling outward the BN nanotubes has been observed theoretically [492]. This is

648

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

explained as arising from the different hybridizations of B and N in the curved hexagonal layer and the N-bucking is expected to form a surface dipole. Similarly, ceramic oxides, such as PZT, can also be used as cold-field electron emitters [493,494]. Pickett [495] found that Cs–O–C surface bonds reduce the work function of diamond surface to F=1.25 eV and he indicated that the low-F surface promotes non-charging electron emission of the surface. Lin [339] proved experimentally that lowering the work function of a cathode by adsorbing both oxygen and electropositive metallic elements on its surface is more effective than by adsorbing simply the metallic element. He noted that an appropriate quantity of oxygen adsorbed on the cathode surface would be essential to result in a better dispenser cathode and also high quantum yield negative electron affinity photo cathodes, while fluorine would play an important part in the electron emission stability. First-principles calculations by Park et al. [496] reveal that the calculated emission currents of carbon nanotubes are significantly enhanced when oxygen is adsorbed at the tip. This finding is attributed to the change in the electronic structure by oxidation and the local field increase at the adsorption site. Treated by O2 and O3, the emission current of the carbon-nanotube array was found to increase by  800% along with a decrease of the onset field emission voltage from 0.8 to 0.6 V/mm [497]. This is in conflict with the experimental results [498] at higher oxygen exposures ( > 65 L) that cause adverse effect on the field emission. Several models have been developed to explain the effect of N and O lowered threshold of cold emission. Nitrogen is soluble in diamond, but it forms a distorted substitutional site with one long bond. This site forms a deep singly occupied donor level, localized mainly on the carbon orbital of the weak C–N bond, 1.7 eV below the conduction band edge Ec, and also a doubly occupied state just above Ev due to the quasi-N lone pair of this C–N bond [499]. Nitrogen can also form aggregates in diamond. The simplest is the nearest-neighbor substitutional pair, known as the A center [500]. The two nitrogen atoms relax away from each other, so that the N lone pair states interact to form two filled states, just below Ev and about 1.5 eV above Ev [499]. The carbon dangling bonds within grains have an energy  1.4 eV above Ev. C–H and C–O bonds possess dipoles which cause a surface dipole layer and a voltage step at the surface [501]. The C–H bond has a positive H site that lowers the affinity; while the C–O bond has a negative O that raises the affinity. It is also explained [500] that nitrogen impurities in the diamond can aid emission if they form deep donor levels. They would create a depletion layer, which causes band bending at the back contact. At sufficiently high donor concentrations, this band bending narrows the tunneling distance there, and allows emission into the diamond conduction band. Some N electrons are donated to intrinsic defect levels, and current is injected into the N levels to hop across the diamond to the front surface, where it is emitted. A possible mechanism for films like SiO2 is proposed as due to ohmic loss or a space charge layer [485], so that the Fermi level of the back contact now lies close to the vacuum layer, and that electrons could tunnel from the back contact into the diamond or DLC conduction band, become hot electrons, and be emitted into the vacuum. The N-lowered threshold has also been attributed to a certain sub-band formed by the N above EF, while such a band is lacking with P or B doping because boron is a

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

649

shallow substitutional acceptor in diamond with a level at 0.38 eV above the valence band edge Ev, and phosphorus can act as a shallow donor with a level 0.46 eV below Ec [499]. As expected, a Cs- and Li-dopant might add their higher energy states to the conduction band of diamond to reduce the work function of the diamond. However, this measure was found less effective than N-doping [502]. 8.5.2. Explanation As an additional consequence of chemisorption of oxygen and nitridation discussed in this work, the local work function changes with altering the surface atomic valencies. The dipole towards the open end of a surface will easily emit electrons under an external electrical stimulus. This may provide an additional possible mechanism for lowering the threshold voltage of emitters in cold-cathode field emission. It is known that the emitting-current density (j) follows the Fowler–Nordheim (FN) relation [503]: j¼

  AE 2  B3=2 exp  E

where A and B are constant  is the dimensionless field enhancement factor (for a smooth surface =1), and E and  the external electric field and the work function, respectively. Obviously, reducing the work function is the only means to enhance the capability of a cold cathode. The possible mechanisms of reducing the work function may include the charge tunneling [493], surface roughening or nano-structuring that enlarge the local curvature of the emitters, as well as chemical adsorption. The low-threshold effect of the cold-cathode field emission due to oxygen and nitrogen addition could be a clear evidence for the BBB modeling predictions. The presence of the anti-bonding states above EF lowers the work function that eases the cathode field emission. Actually, the phosphorus-3p electrons are more mobile than are the N-2p and O-2p electrons, but P-doping gives little reduction of the threshold with respect to the O and N doping. The work function of Cs and Li (  3.5 eV) is much lower than that of other metals ( 5.0 eV). However, adding the Cs and Li to the diamond surface improves little in the emission properties. Therefore, lowering the threshold of field emission is the special ability of electronegative additives such as nitrogen and oxygen. The lone pair induced dipoles lowers the work function and hence the threshold of field emission. However, it is anticipated that the production of the H-like bond at the surface due to over-dosing may have detrimental effect on the work function reduction, such as the case [498] of carbon nanotubes with overdoped oxygen. 8.6. Magnetic enhancement Inclusion of oxygen and nitrogen could enhance the magnetic momentum by as high as 25% of the ferromagnetic materials [504,505]. Coey and Sun [506], and Pan et al. [507] found that the addition of N to the rare earth–Fe(Co) (R–Fe) system

650

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

improves the Ms and the Tc considerably. The Ms of the R–Fe(Co) alloys were increased by about 30–40% relative to their parent alloys. The N-enhanced Ms may enable these kinds of materials to be used in high-density data storage and as new kind of strong permanent magnets. However, mechanism for the magnetic enhancement is yet to be clear. The BBB correlation indicates that the chemical reaction could transport electrons from the less electronegative element to the higher ones. The charge transport modulates the valence state and hence the angular momentum that is the sum of spin and angular contributions. It can be seen from Table 17 that the total angular momentum increases when the Fe alters its atomic states to Fe+ or Fe dipole. In the former, the Fe donates one 3d electron to the sp3 hybrid orbital of the N acceptor. In the latter, a Fe 3d electron is excited to its own outer shells, 4p or 4d, corresponding to the higher antibonding energy. The N
3 and its electrons contribute nothing to the magnetization. The total momentum of the Fe+ is 2.5 and the Fe dipole is 4.0 (3d54s24p1) or even 5.0 (3d54s24d1). The average momentum is then 2.875 or 3.125, 25–40% greater than is a pure Fe (2.22 measurement). Apparently, this interpretation, being well account for the experimental observations, should be reasonable. For the N
3+3Fe+1+Fe (dipole) unit, the estimated average angular-momentum increases to 2.875 or 3.125 relative to the pure Fe (2.22 measurement), which agrees with the reported findings.

9. Application: II. Synthetic diamond The original ideas of sp-orbital hybridization of C, N and O upon interacting with solid surfaces have been applied to solving critical issues in the synthesis of diamond films, which has led to progress as introduced below:  Preferential oxidation of diamond {111};  Adhesion enhancement between diamond and metal substrate;  Grain size and annealing effects on the dielectrics of diamond.

Table 17 Variations of angular momentum of Fe with its atomic states [325] P P L= Li Valence state Configuration S= Si Fe Fe+ Fe+2 Fe (dipole 1) Fe (dipole 2)

3d64s2 3d54s2 3d54s1 3d54s24p1 3d54s24d1

2 2.5 2.5+0.5 2.5+0.5 2.5+0.5

0 (L-frozen) 0 0 0+1 0+2

Dipole 2 corresponds to the antibonding states being well above the EF. a The total angular momentum is governed by the Hund’s rule.

P Ja= (LS)i 2 (2.22) 2.5 3.0 4.0 5.0

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

651

9.1. Thermal oxidation Oxidation and graphitization of synthetic diamond is an important issue for practical applications. Oxidation limits the performance of diamond used in cutting tools, optical windows and electronic devices when the diamond piece is experiencing working conditions under which its temperature rises in air or in other gaseous ambient. Johnson et al. [508] heated diamond optical windows in the temperature range of 970–1170 K in air and examined the infrared (l=10.6 mm) transmittance of the windows. They found that the forward scattering of the infrared from the diamond windows changes from 0.8% before heating to 6.2% after heating at 1070 K for 255 seconds while the transmittance drops by 6–12% after heating. Nimmagadda et al. [509] suggested that oxidation occurs preferentially at regions of grain boundaries, local defects and diamond-like carbon phase during the earlier stages of oxidation. An investigation by Zhu et al. [510] revealed that diamond oxidation occurs at  1070 K and suggested that diamond oxidation depends on the crystal orientation and defect density. Thermal-programmed desorption (TPD), EELS and AES investigating the oxidation of single crystal diamond (111) and (110) surfaces [511] revealed that molecular oxygen is easily chemisorbed on the clean diamond surfaces at room temperature. Carbon monoxide was the only product of thermal desorption from both surfaces. Apart from a low temperature desorption peak present in all TPD spectra, two CO desorption peaks at 790 and 1030  C were observed for the C(111)-(21) surface, whereas, only one desorption peak in the 760–890  C range was observed in case of the C(110) surface. In a mimic of low-earth-orbital atomic oxygen dominated environment, Li et al. [512] examined the reaction of polycrystalline diamond films with energetic atomic oxygen of  5.0 eV at ambient temperature. They found a low reaction rate of R=8.2810
26 cm3/atom (volume eroded by per oxygen atom) at room temperature, suggesting the erosive-resistant of CVD diamond in the space ambient. Hence, it is widely believed that the {001} faces have fewer defects than the {111} faces and thus the {001} face is more resistant to oxidation. Understanding of diamond oxidation so far is constrained to the ‘defect density’ mechanism. Recently, Theije et al. [513] examined the oxidative etching of the diamond {111} surface using ‘dry’ oxygen, ‘wet’ oxygen/water vapor mixture and in molten potassium nitrate. They found upon dry-oxygen etching the {111} surface roughens very fast and becomes unstable. It is proposed that the rugged surface is due to the chemical roughening in which oxygen destabilizes the surface by different bond configurations. In a thermal oxidation of the hydrogenated diamond {100} surface, Pehrsson et al. [514] found that the {100} surface is largely inert to oxygen below 1220 K. Thermal oxidation with un-activated oxygen occurred only after the surface hydrogen had been removed, usually at about 1220 K. Obviously, the {100} surface is far harder to be oxidized than the {111} surface. The preferential oxidation of the diamond {111} surface is in contrast to the ‘periodic bond chain’ theory [515] that predicted the {111} to be the only stable face of diamond and, therefore, the {111} should be the face with slowest etching rate or more oxidative resistant than other faces [516].

652

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

Challenges for the underlying mechanism for the oxidation of a diamond yet remain. We need to answer why are the densely packed diamond {111} planes oxidized easier than diamond {001} or {220} planes, and how does the oxygen penetrate into diamond {111} face in the course of reaction? Here we present our findings on the geometrical selectivity of oxidation with a mechanism based on the premise of oxide bond switching (Section 7.2). Fig. 41 shows the SEM image of the oxidized diamond single crystal obtained after heating at 1070 K in air [267]. It is seen that diamond erosion occurs preferentially in the planes possessing C3v symmetry. However, the rectangular planes experience no change apart from the defect sites. Generally, the {111} planes grow in CVD synthesis at a relatively higher temperature than the {220} planes [517]. At 1–2% CH4/H2 gas ratio, the synthetic diamond changes at 1100  50 K from {220} plane dominated to a {111} dominated phase. The {111} planes can form hexagonal platelets, truncated hexagonal platelets, decahedrons (pseudo five-fold symmetry), icosahedral (20 faces) and triangular shapes, depending on stacking errors [518]. Nevertheless, the {111} faces cannot form platelets with 90 corner angles. Therefore, the square or rectangle platelets in Fig. 41(a) can be identified as the {220}, or equivalent planes. It is important to note that the lone pairs of oxygen possess the special ability of inducing dipoles. The interaction between the dipole and the oxygen, and hence the dipole to the bulk, is very weak ( 50 meV, Section 6) as there is no electron sharing between the dipole and the oxygen. Due to the strong repulsion of the non-bonding lone pairs, the dipoles tend to locate at the open end of a surface. The loosely bounded dipoles therefore tend to be eroded away of the surface in the process of corrosion. The atomic density ratio of the {111} plane to that of the {220} plane is given as [refer to Fig. 41(b and d) being the lattice parameter]:

Fig. 41. (a) SEM image showing that oxygen atoms penetrate into the {111} plane throughout the course of reaction. (b) Schematic illustration of the geometrical environment of the {111} and {220} planes of diamond. The {111} plane accommodates more easily an oxide tetrahedron than the {220} plane, as it is harder for oxygen to find a nearest neighbor C in the second atomic layer [524].

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

nf111g ¼ nf220g

1 pffiffiffi 2 2d sinð60 Þ=2 1:5 pffiffiffi 2d 2

653

rffiffiffiffiffi 32 ¼ >1 27

p The {111} inter-plane distance is ( 3-1)d/2 (from p the plane comprising atoms labeled 123 to atom 0). The {220}-plane separation is 3d/2 (from atom 0 to atom 4). It is seen from the diamond unit cell in Fig. 41(b) that each of the two planes cuts through three C atoms but the geometrical arrangement of these atoms in the planes is entirely different. The three atoms (labeled 1, 2 and p p 3) in the {111} plane form a C3v hollow with identical edges of a= 2d, and of ( 3-1)d/2 in depth. The diamond p bond length is 3d/2 or 1.54 A˚. However, the three atoms (labeled 0, 1 and 2) in the p {220} plane are arranged in such a way that the triangle edges are at d, d and 2d. Therefore, it is easier for an oxygen adsorbate to find the fourth C atom (labeled 0) underneath the C3v-hollow site in the {111} plane to form a quasi-tetrahedron. However, the atom (labeled 3 or 4) underneath the {220}-plane composed of atoms labeled 0, 1, and 2 could hardly form a frame in which an oxide tetrahedron could fit. Therefore, it is harder to facilitate the oxide tetrahedron in the {220} plane than in the {111} plane. This explains why the densely packed {111} plane oxidizes easier than the {220} plane. Although the atomic sizes and the geometrical arrangement of the considered metals (Sections 3–4) are different, the oxide tetrahedron formation is essentially the same, as we have already noted in the context. During the process of oxidation of the diamond, oxygen atoms penetrate easily into the {111} planes and loosen the C atoms at the surface by forming the weakly bound dipoles. The dipoles are readily eroded away at elevated temperatures and then the oxygen seeks new partners for the tetrahedron. Oxygen can always survive by fitting itself into a suitable bonding environment to build up the tetrahedron through bond switching. It may be questioned that diamond is not essentially the same as metals in the process of reaction. In fact, we have shown that the factors dominating the oxidation of a solid surface are: (i) the difference in electronegativity, (ii) the scale and geometry of the lattice, and, (iii) the temperature and oxygen pressure. Electronegativity determines the nature of the bond or the easiness and amount of charge transportation between the bonding constituents. The atomic geometry and lattice constant determines the way of facilitating the oxide tetrahedron. The temperature and the ambient oxygen pressure determine the rate of reaction. Therefore, the mechanism of oxidation should be valid for a solid surface whether it is a metal or not. No specification of element or its phase is necessary for oxide tetrahedron formation. It is clear now that diamond erosion starts at 750 K in air showing strong geometric selectivity, an indication of oxide tetrahedron formation as in the case of metal surface oxidation. It is also clear now that the {111} planes provide a more suitable bonding environment than the loosely packed {220} planes. Oxygen penetrates into the bulk by bond switching and leaves behind the weakly bound dipoles that are eroded away during the process of corrosion. Loh and his co-workers [519] have recently detected, using the UPS He-II emission, that the lone-pair DOS

654

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

features (around
3.0 eV below EF) are present at the diamond {111} planes with chemisorbed oxygen. This provides further evidence that the oxide tetrahedron forms on the non-metal diamond surface with lone pair production. Therefore, the preferential oxidation of diamond {111} shows that our model premise is valid not only for metals but also for the non-metallic diamond. 9.2. Adhesion improvement The poor adhesion of diamond films to metal substrates has been a long-standing issue that prevents practical applications of the excellent properties of synthetic diamond. Earlier analysis (Section 3.6) [100] for Ni surface reaction indicated that the C–Ni(001) bond experiences strong compression and the surface is covered with Ni+ and Ni2+ while the N–Ni(001) bond stress is slightly tensile and the surface comprises alternating Ni+ and Ni+/diple. It is possible to extend the ideas (summarized in Table 13) to the adhesion of diamond to metals by adopting the conclusions derived in Section 3.6. XRD measurement [520] has confirmed the prediction that carbon turns the tensile stress of Ti surface to be compressive. Comparison of the surface morphology of TiC with that of TiN indicates that the surface stresses are different in nature [521]. Based on these predictions and justifications, we have designed and inserted a graded TiCN interlayer between metal substrate and the synthetic diamond to neutralize the interface stress, which has improved the adhesion between diamond and the Ti substrate substantially [521]. Fig. 42 compares the cross-section of the diamond/Ti system with and without the graded TiCN interlayer. It can be seen that the graded TiCN interlayer is able to remove the cracks and has enabled the diamond to adhere to the Ti substrate strongly. The critical scratch load for the adherent diamond is as high as 135 N compared to the adhesion strength of 55 N for nanostructured diamond on the tungsten carbide substrate. These exercises may provide a prototype for joining metals with non-metals in composite materials. The success of this approach evidences that the extension of the sp-hybrid bonding of oxygen to carbon and nitrogen is on an essentially correct track. The extension of the sp-hybrid bonding to nitrogen has also led to the new understanding that the

Fig. 42. SEM cross-section observation of diamond with or without the designed graded TiCN buffer layer. The porous carbide with strong bond repulsion prohibits the adhesion while the buffer layer allows the diamond and Ti substrate to bond strongly.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

655

crystalline carbon-nitride formation is harder than the hexagonal SiCN crystallite [326], and that overdosed ( > 75% partial pressure) nitrogen in diamond deposition could turn the diamond to be SiCN [522]. Further results show that the surface bond contraction and the lone pair interaction play dominant roles in the extraordinary mechanical performance of the hard and self-lubricant nitride surfaces [327]. 9.3. Dielectric relaxation and transition Lowering the dielectric constant of a medium is increasingly demanded by the ‘copper-low k’ DSIC technology and for photonic devices. Dielectric reduction will lead to the blue shift in the photo absorption, in particular for semiconductors having an indirect band-gap transition [384]. Diamond films with reduced domain size may have high potential to serve as the dielectric layer in DSIC technology, provided suitable thermal performance and adequate adhesion. The mechanism of surface bond contraction implies that adjusting the particle size or introducing pores into the solid can reduce the dielectric constant. This is because the spontaneous bond contraction of the surface atoms enhances the crystal field of the system. The dielectric constant is inversely proportional to the square width of the band-gap (E
2 g ) [523], which depends functionally on the crystal field [384]. An impedance (resistance and capacitance) spectroscopic investigation reveals that [524,525]: (i) reduction of the particle size from the micrometer scale to the nanometer scale can reduce the static dielectric constant of the synthetic diamond films; (ii) raising the measuring temperature has the opposite effect to that caused by reducing the particle size; and, (iii) for the nanostructured diamond, the impedance transits at  520 K. The transition corresponds to the mechanism of electron-polarization at the grain interior and electron-polarization at grain boundaries, respectively. It is suggested that the grain interior is activated at lower temperature and the impurities at grain boundaries are activated at higher temperature. The activation energies for the two mechanisms were measured as 0.13 and 0.67 eV, respectively. The opposite effects of heating and particle-size reduction on the dielectric behavior of the nanometric diamond are due to the variation of the mean atomic separation that expands upon heating while contracts with reduced particle size. Lattice contraction enhances the crystal field and subsequently decreases the dielectric constant. On the other hand, the crystal field can be weakened by lattice expansion under thermal excitation. Hence, heating the sample has the opposite effect to that caused by reducing the grain size on the dielectric constant of the diamond.

10. Summary 10.1. General understanding 10.1.1. Essential events at a surface Two events are found essential at a surface with and without adsorbed oxygen. The underlying mechanism and the consequences of these events should be useful in practice:

656

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

 First, a chemical bond contracts at a surface or at sites surrounding defects where the CN of the atom reduces. An extension of the concept of ‘atomic size shrinks with its CN reduction’, initiated by Goldschmidt and Pauling long ago, to solid surfaces has led to the bond order-length-strength correlation mechanism, which dictates the tunable properties of a nano-solid of which the portion of surface atoms increases with reducing particle size. The BOLS correlation modifies both the cohesive energy of an atom at a surface and the binding energy density in the relaxed surface region.  The BOLS correlation is expected to provide impact on the fields of nanometric materials and surface science as well, as we have demonstrated.  Second, it is essential for O, N, and C atoms to hybridize their sp orbitals upon reacting with atoms whether in the gaseous phase or at a solid surface. It is valid to adopt the orbital configuration of a single molecule, such as CH4, NH3 and H2O, to the interaction of O, N and C with atoms at the solid surfaces of transition metals, noble metals and a non-metal diamond. Such exercises have led to the correlation of chemical-bond–valence-band–potential-barrier with the often overlooked events of lone pair non-bonding, H-like bonding, and dipole anti-bonding, which have indeed brought us enormous new insight into the electronic process of oxidation and practical applications.  The combination of BOLS and BBB mechanisms may cover the major sources that determine the unusual behavior of a surface. Bond contraction at a surface provides perturbation to the Hamiltonian that defines the entire band structure and related properties of a solid; chemical reaction causes a repopulation of valence electrons in the valence band and modifies the binding energy as well. Catalytic reaction changes the bond nature while the physically sizing of the solid changes the average bond length. All the physical properties should be derivatives of the Hamiltonian of the system or the DOS distribution in the valence band of the solid. 10.1.2. Bond nature and bond forming kinetics It has been clear that surface oxidation is a kinetic process in which O
1 forms first and then the O
2 follows with sp-orbital hybridization and, H-like bond formation, if necessary. An oxide tetrahedron forms intrinsically, which is independent of the bonding environment or the nature of the host element. A host atom may donate more than one-electron to different oxygen atoms while one oxygen atom can never catch more than one-electron from a specific host atom because of the directionally hybridized orbitals. The sp orbitals of an oxygen atom cannot hybridize until its two bonding orbitals are fully occupied. The two bonding orbitals can be occupied by sharing electrons with atoms of a metal or a non-metal, or even by dragging the electron-cloud of dipoles, thereby, being able to stabilize the primary M2O tetrahedron. The production of the non-bonding lone pairs and the induced dipoles is also intrinsic and is independent of the environment or the nature of the bonding constituents. One oxide surface may contain atoms with different valencies: O
1; O
2; M+; M+2; Mdipole; M+/dipole and Mvacancy. An oxide system is composed of various

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

657

chemical bonds: the ionic or polar-covalent bond between the oxygen and the host atom, the non-bonding lone pair of oxygen, anti-bonding host dipoles, and the H-like bond. Covalent bond and anti-bond can never form between the adsorbate and the host substrate atom due to the huge difference in electronegativity between them [107]. Although most of these atomic valencies and bonding events are often overlooked, these bonding events and the corresponding atomic valencies play crucial roles in the process of catalytic oxidation and in determining the properties of an oxide system, as demonstrated in this report. An oxide tetrahedron forms in the following discrete stages: (i) First, an ionic or polar-covalent bond forms. The surface bond contracts because of CN imperfection. Ions polarize their surrounding atoms. (ii) The second contraction bond follows. (iii) The sp orbitals of oxygen hybridize with the production of two non-bonding lone pairs. (iv) Interaction develops between the adsorbate and the lone pair induced antibonding dipoles. (v) At higher oxygen coverage, H-like bonds may form by dragging the electron cloud of the dipoles to the bonding orbital of the oxygen adsorbate. Measurements have revealed that the bond process is reversible by heating or by bombardment of energetic particles. Bond switching is responsible for the bulk oxidation and for oxygen floating in the process of epitaxial growth of metal on oxygen pre-covered metals. Oxide tetrahedron formation could be the cause of observed atomic dislocation, phase formation and transition, charge and mass transportation. The theme of sp-orbital hybridization has been extended to the reaction of carbon and nitrogen with a Ni(001) surface. Although the observed patterns of reconstruction and morphology for the O–Rh(001) and the (C, N)–Ni(001) surfaces are the same, the valencies of surface atoms, the bond stress and the driving forces for these surfaces are quite different. This is simply due to the variation of the valencies of the p p adsorbates. A ‘rhombi-chain’ model or a c(4 24 2)R45-16O
2 phase structure could describe these ‘p4g’ reconstructions better. 10.1.3. Orientation specificity of the tetrahedron Formation of the oxide tetrahedron and its kinetics are common while the adsorbate-site specification, the order of the two bonds formation and the orientation of the oxide tetrahedron vary with the bonding environment. Except for the initial stage of oxidation, the oxygen adsorbate prefers a position inside a tetrahedron. However, the scale and geometry of the lattice, and the electronegativity of the host determine the bond-formation order and the site-and-orientation specificity of the tetrahedron.  Oxygen prefers the next-nearest-neighboring C4v hollow site of the fcc(001) surface. However, the orders of the ionic bond formation at the Cu(001) and (Rh, V, Ag)(001) surfaces are opposite owing to their different atomic sizes and the values of electronegativity. Different patterns of reconstruction

658

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

of the O–Rh(001) and O–Cu(001) surfaces are suggested to arise from nothing more than the minor difference in their atomic sizes [radius, 1.342(Rh)–1.277(Cu)=0.065pA˚] and electronegativity [2.2(Rh)–1.9(Cu)=0.3]. The smaller hollow [d=2R( 2–1)=1.058 A˚] and the lower electronegativity of the Cu(001) surface allow the oxygen [d=1.32 A˚] to bond to one Cu surface atom first, and then to another Cu underneath. The wider hollow [1.112 A˚] and the higher electronegativity of the Rh(001) surface permit oxygen to sink into the hollow and form the first bond to the Rh atom underneath. The inverse orders of ionic bond formation generate entirely different patterns of reconstruction and morphology, as well as the different surface atomic valencies and phase structures on the fcc(001) surface of Cu, Rh, V and Ag. Adsorption of oxygen to the Rh(001) surface creates the Rh5O radial and then the Rh4O tetrahedron that gives the ‘p4g’ clockwise reconstruction and the ‘rhombi-chain’ fashion. Oxygen reaction with the Cu(001) surface gives rise to the off-centered pyramid and then the missing-row structure in which the pairing ‘Cudipole : O
2 : Cudipole’ strings form. The pairing Cudipole–Cudipole crosses over the missing row vacancy. Alternation of the valency of oxygen from O
1 to O
2 gives rise to the off-centered CuO2 pairing pyramid into the Cu3O2 pairing tetrahedron on the Cu(001) surface. The analysis also applies to the phase transition of O–V(001) and O–Ag(001) surfaces.  The fcc(110) surface of Cu, Ni, Ag and Pt could fit the oxide tetrahedron ideally by locating the adsorbate at the long-bridge hollow site with ‘missingrow’ production. However, the slightly opened fcc(110) surface of Rh and Pd allows the oxygen to locate at the fcc(111) facet site in a zigzag fashion along the close packed direction without row missing. The former produces the alternative rows of metal vacancies and the single ‘Mdipole : O
2 : Mdipole’ string. The latter leads to the sequential phases that correspond to the M5O, M4O and M3O structures with H-like bonds dominant at the surface. The observable differences on these fcc(110) surfaces result from nothing more than the difference in the atomic size and electronegativity.  The slight geometrical difference between the (Rh, Pd)-fcc(110) surface and the (Co, Ru)-hcp(1010) surface allows the oxygen adsorbates to prefer the troughs in different ways. Oxygen adsorbates prefer the fcc(111) troughs located beside a certain close-packed metal row at the fcc(110) surface. In contrast, the adsorbates locate in the same fcc(111) facet troughs but locate between two neighboring metal rows at the hcp(1010) surface. Such a difference in the adsorbate’s site-specificity generates entirely different patterns in STM observations. The zigzagged dipole-row is seen at the (Rh, Pd)(110) surface but the grouped octopoles appear at the (Ru, Co)(1010) surfaces at 0.5 ML oxygen coverage.  Oxygen prefers the hcp(0001) hollow site in both the (Rh, Al)(111) surfaces and the Ru(0001) surface. Although the basic tetrahedron remains during the course of reaction the electronic configurations of the constituent atoms alternate in the process of oxidation. Electron transportation leads to the ‘C3v-radial’ and then the ‘pairing-row’ type reconstruction and, finally, H-like

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

659

bonds dominate at the surface. In comparison, oxidation of the Cu(111) and the Pd(111) [526] surfaces gives rather complicated patterns of reconstruction.  At the O
1-induced precursor phases, O
1 forms one bond to a host atom nearby and the O
1 polarizes its surroundings. The O
1-induced dipoles interact with the O
1 adsorbate through the non-bonding O
1Mdipole interaction that is rather weak as revealed with EELS from the O–Ru and O– Rh surfaces at the initial stages of oxidation (Section 6.1). It seems that oxygen tends to form the first bond with an atom at the transition metal surface such as the Cu(001) and Co(1010) surfaces, because of the lower electronegativity (< 2.0) and shorter atomic radius (< 1.3 A˚). The O
1–O
1 dimer rests above the Co atoms forming the ‘8’-shaped Co2O2 bond with two Co atoms at the Co(1010) surface because of the dimension of the C2v hollow of Co(1010) (dimension of 2.504.06 A˚2). The O
1 locates 40.4 A˚ eccentrically (  0.18 A˚) above the Cu(001) hollow (2.553.61 A˚2) and forms the CuO2 off-centered pyramid at the surface. In contrast, oxygen sinks into the hollow site and bonds firstly to the noble metal (Ag, Rh, Pd and Ru) atom underneath. Therefore, it is safe to say that O
1 prefers the on-surface position of a transition metal and then the O
2 buckles in, and that O
1 and O
2 locate at a sub-surface position inside the surfaces of noble metals throughout the course of reaction.  In the O
2 valence situations, the sp-orbital hybridization allows the O
2 to seek four suitable neighbors for a stable quasi-tetrahedron. Oxygen prefers the nearly central position of the tetrahedron rather than any other alternatives. Therefore, difference in the bonding environment determines the orientation of the tetrahedron of which the lone pairs tend to point into the open end of the surface. The Cu(001) surface supports the pairing Cu3O2 tetrahedron, while the (Cu, Ni, Ag, Pt)(110) surface supports the primary Cu2O tetrahedron. The Rh(001) surface supports the single Rh2O tetrahedron with rotation, which forms the rhombi-chain along the < 11 > direction. The O–Cu(001) surface differs only from the O–Cu(110) surface in that the O–Cu–O chain rotates itself by  45 to fit the crystal orientation. p p The V(001) surface allows the ‘radial’ and then the short ordered ( 23 2)R45 -4O
2 reconstruction. The missing row type Ag3O2 structure is more stable at lower temperature while the ‘radial’ Ag5O (O
1) is table at room temperature and above.  Oxygen can penetrate into the bulk or float back to the surface subjecting to external conditions, such as raised temperature or higher exposure, through bond switching. Oxide tetrahedron formation shows strong geometric selectivity at the hardest diamond surface and the diamond {111} plane is preferable throughout the course of reaction. The weakly bound dipoles are readily eroded away under thermal excitation. 10.1.4. Consequences of bond forming The external conditions determine the site-and-orientation specificity of the basic oxide tetrahedron, which generate versatile changes in observations. The main identities observed so far from the oxide surfaces are summarized below:

660

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

 O–M–O protruding chains form at most of the surfaces. It is recognized that all the O–M–O chains are zigzagged by the lone-pair non-bonding states rather than by any other kinds of chemical bonds. The observed metal dipoles in the O–M–O chains are not those belonging to the M2O system, as these dipoles exchange no electrons with the O
2. Those belonging to the M2O molecule are perpendicularly connected to the O–M–O chain and they cannot be detected by STM imaging because of the reduced size and lowered energy levels of occupancy.  The missing-rows could form on a very limited number of surfaces due to oxygen chemisorption. The missing atoms in the O–(Cu, Ag)(001) surface and the O–(Cu, Ni, Ag, Pt)(110) surfaces are isolated during bond formation as the metallic neighbors of the missing-row atoms have combined with oxygen adsorbates. These ‘extra’ atoms for tetrahedron formation are readily squeezed away from their regular positions by a small disturbance such as being dragged out by the two neighboring O
1 at the initial stage of Cu(001) surface reaction.  H-like bond formation is, however, more general than the missing row in the surfaces analyzed. The process of H-like bond formation compensates for the lack of an atom for the tetrahedron formation, in which the adsorbate drags the electron-cloud of the dipoles to its bonding orbitals. This process sharpens the ‘tip’ of the ‘honey-comb’ protrusions such as in the (Co, Ru)(1010)c(24)-4O
2 phase (Fig. 16). There are no atoms to be missing. Importantly, the H-like bond lowers the STM protrusions and stabilizes the system by reducing the dipole moment at the surface, as do the phases of (Co, Ru)(1010)-(21)p2mg-2O
2, (Rh, Pd)(110))-(21)p2mg-2O
2 and Ru(0001)(11)-O
2. Except for those phases induced by O
1 and indicated with ‘missing row’ in Table 4, all the phases contain the H-like bond. At higher oxygen coverage, the H-like bonds interlock all the surface atoms. Such bond interlocking may allow the top layer to prevent atomic diffusion into the bulk, which may act as a barrier that should be of use in many fields.  Because of oxygen penetrating into the top layer for tetrahedron formation, the first interlayer spacing expands, which depends on the bond geometry. The production of metal ions in the second layer strengthens the interaction between the ionic second layer and the third metallic layer. Therefore, the spacing between the second and third atomic layer often contracts. 10.1.5. Driving forces behind reconstruction It is known that the binding energy for a metallic bond is about
1.0 eV and for an ionic bond it is about
3.0 eV. The energies for the H-like bond and the nonbonding lone pair are around
0.05 to
0.1 eV. Disregarding the minor energies for the weak bound states, the transition of metallic bond to the contracted ionic bond is associated with a gain in energy: DE ¼
3:0=ð1
QÞ
ð
1:0Þ ¼
ð2 þ QÞ=ð1
QÞ <
2:0 ðeV=bondÞ:

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

661

The net gain in energy provides forces that drive the reconstruction, though the orbital hybridization and the anti-bonding formation would consume some amount of energy. Analysis of the Rh(001)-O
2 clock reconstruction indicates that the driving force for the atomic dislocation also comes from other sources such as the alternative electrostatic repulsion and attraction along the < 11 > rhombi chain, and a response of bond tension that stabilizes the clock rotation. 10.1.6. Factors controlling bond formation It has been clear that the sp orbitals of an oxygen atom hybridize intrinsically upon the full occupancy of the two bonding orbitals. Formation of the oxide tetrahedron is independent of the nature of the host element and the geometrical environment. However, external factors determine the site-and-orientation of the oxide tetrahedron and the order of the two ionic bonds of the tetrahedron. The extrinsic dominating factors are: (i) (ii) (iii) (iv)

difference in electronegativity between the bond constituents lattice geometrical orientation and the scale of lattice constant of the host valencies of the oxygen adsorbate substrate temperature and oxygen exposure as well as aging conditions.

Obviously, only the last one is controllable. The valencies of oxygen can be changed by varying oxygen exposure or by annealing the sample at a certain temperature. The multiphase ordering on the metal surfaces results from the variation of the oxygen valency under various bonding circumstances. However, it is still puzzling how the oxygen exposure changes the valencies of the oxygen. 10.2. Capability-enhancement of probing techniques The developed knowledge allowed us to enhance the capabilities of the following techniques in terms of atomic valencies, bond geometry, valence DOS, bond strength and bond forming kinetics as summarized below. 10.2.1. STM and STS  The striking significance of the current modeling exercise is that, based on the STM and STS observations, the individual atomic valencies at the surface have been identified and the kinetics of oxidation has been formulated. For instance, electron clouds of metal dipoles, induced by either the O
1 or the non-bonding lone pairs of the O
2, dominate the STM protrusions; ions of metals or oxygen (M+, O
1 or O
2) or vacancies of missing atoms cause the STM depressions. Even though it locates well above the surface atomic plane, the oxygen adsorbate is still not detectable by an STM because the occupied energy states of the oxygen are still lower than the EF of metals, even though the size of the adsorbate increases. The M+ is undetectable because of both the reduced atomic size and the lowered energy states of occupancy.

662

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

 The various shapes of STM protrusions represent the configuration of dipoles. For example, (i) a single dipole forms on the Cu(110)-O
2 surface; (ii) an engaged-cogwheel and paired dipoles (quadruple) grow on the Cu(001)O
1 surface; and, (iii) the grouped dipoles (octupole) form a congested array p p on the (Co, Ru)(1010)-c(24)-4O
2 and the V(001)-( 23 2)R45 -4O
2 surfaces.  The unique STS profiles covering energies of EF  2.5 eV from the O–Cu(110) surface give the on-site DOS information about the lone pairs ( < EF) and anti-bonding dipoles (> EF). The STS features from O–Nb(110) show the resonant features due to the surface image potential, which is in agreement with inverse PES observations.

10.2.2. PES, TDS, EELS and VLEED As analyzed in Sections 4–6, the spectral identities of STS, PES, TDS, EELS and VLEED correspond to individual processes of bond formation or their consequences on the energetic behavior of valence electrons. Oxygen-induced phase ordering and patterns of observations vary indeed from situation to situation. However, formation of the primary tetrahedron, the oxygen-derived DOS features and the bond forming kinetics are naturally common for all the analyzed representatives.  The four valence DOS features in a PES correspond to the oxygen-host bonding (
5 eV), non-bonding lone-pairs of oxygen (
2 eV), electron holes of the host (4EF) and the anti-bonding host dipoles (5EF), rather than simply the addition of the 2s or 2p states of the isolated oxygen to the valence DOS of the host.  The change of work function corresponds to the bandwidth of the antibonding dipole states, which provides information about the dipole formation and H-like bond formation. Dipole formation widens the bandwidth of the anti-bonding states and reduces the work function; H-like bond formation narrows the anti-bonding sub-band and causes the reduced work function to restore, which stabilizes the entire system. The reduced work function could be of use in the cold-cathode electron emission but the H-like bond formation may have an adverse effect.  TDS provides information about the activation energy for individual bond breaking. The oscillation of the TDS signatures of the O–Rh and O–Pd surfaces has been related to the sequence of bonding, non-bonding and the Hlike bonding contribution.  The energy increase of the dipole-stretch mode in EELS relates directly to the strength of non-bonding interaction that varies from the O
1Mdipole to the O
2:Mdipole and the H-bind-like contribution. Raman spectroscopy gives information at low frequencies about the non-bonding lone pair interaction, being similar to that revealed by EELS. As the weaker part of a hydrogen bond, the lone pair interaction is found to exist in the oxides, nitrides and biomolecules such as protein and DNA chains.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

663

 It has been verified that LEED at very low energy ( < 30 eV) provides a unique technique that can collect nondestructive information from the top atomic layer and the VLEED covers the valence band in energy. VLEED gives simultaneous information about the surface morphology, lattice geometry and valence DOS features, though sophisticated decoding and a proper model is required.  The BOLS correlation enables the APECS (a combination of AES and XPS) to reveal information about the single energy levels of an atom isolated from solid and their shifts upon solid formation as well as the bonding of a monatomic chain, which is beyond the scope of direct measurement in convention. 10.3. Findings in applications The knowledge of BOLS and BBB correlation has led to a systematic understanding of the catalytic effect and the shape-and-size dependency of a nano-solid, which has led to some new findings in nano-tribology, nano-dielectrics and nanophotonics, etc. The BBB correlation knowledge has also led to the process design for strengthening diamond-metal adhesion. Designing new functional materials for photo and electron emitters and photonic crystal with tunable optical band-gap has also been possible. Progress made in these practical applications could be indicative that our approach is in a correct track, which should bring along new findings. Further efforts, I am sure, towards materials design based on the BOLS and BBB correlations would be more effective, interesting and rewarding.

11. Recommendations With the bond-forming premise, we have been able to generalize various observations of oxidation occurring at different surfaces. Developed knowledge should pave the path towards bond engineering. Further work would be even more encouraging and rewarding. Recommendations on further study are given below.  In crystallographic studies of chemisorbed surfaces, it might not be essential to locate precisely the static positions of the adsorbate atoms at a surface, as the reaction is a kinetic process in which charge transportation dominates and all the involved atoms move collectively and continually. Identifying the nature, the dynamics and kinetics of bond formation, and their consequences on observations and practical applications should be a foremost concern in a study. The bond angle and bond length are retractable depending on surface coordinates and the bonding environment. During the reaction, all atoms change their sizes and valencies, and hence hard-sphere models have limitations. As demonstrated, a small variation of oxygen positions in the O– Cu(110) surface gives an entirely different physical picture. Therefore, in simulating diffraction data (such as the LEED I–V profiles), it would be

664

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685









necessary to replace the traditional trial-error wisdom of individual displacement of the hard spheres with variation of bond geometry and atomic valency. The value of the R-factor in LEED optimizations may be improved; on the other hand, if one allows the oxygen to reside inside a tetrahedron and allows the bond to contract at surface. Due to the correlation among the parameters, the number of numerical solutions may not be unique and physical constraints should be applied. It would be more realistic in theoretical approaches to consider the actual occupancy of the orbitals that may hybridize, by the lone pairs and antibonding dipoles, in particular. The interatomic potential for the O
2 : Mdipole (lone pair) differs from that for the O
2–M+ (ionic bond). Such modifications may improve the outcome of calculations such as minimizing the system total energy. Bond contraction is also an intrinsic process at the surface and a bond switching mechanism would be necessary. Exercises in terms of rigid adsorbate being trapped in the potential well at the surface without breaking the barrier seem to be less adequate in chemisorption studies. Recent photoelectron diffraction measurement [91] indicated that the bond lengths of CO, NO and NH3 molecules are much shorter (by up to 0.79 A˚) on a Ni(001) and a NiO(001) surface than those determined by theoretical calculations and, therefore, the currently widely-used theoretical methods are questioned as in very serious errors in the particular cases [91]. It is pointed out that the shortened bond length indicates true bond formation, and not simply electrostatic interaction as has been concluded based on currently popular theoretical treatments [91]. Importantly, chemisorption is a process in which charge transportation dominates. Atomic dislocation is only one of the numerous consequences of bond formation. Therefore, it is necessary to examine the system by all the effective means, and their correlation in terms of crystallography, morphology, desorption and electronic spectroscopy to avoid misinterpreting the results. The sp-orbital hybridization may extend to catalytic reactions involving other electronegative elements. Interestingly, based on DFT pseudo-potential and tight-binding theoretical calculations, Lefebvre et al. [527] pointed out that the electronic structures of the tin monochalcogenide (SnX, X=O, S, Se, or Te) family possesses similar DOS features that are consistent with the presence of a lone pair. This might be indicative the essentiality of sp orbital hybridizations for other electronegative additives. The weak non-bonding interaction, such as the lone pair and H-like bond, may play an important role in bioelectronics such as the folding, signaling, and regulating of protein and DNA chains as well as cell binding. The non-bonding weak interaction and the dipole formation may also contribute to the mechanism for medicinecell interaction, such as messaging and regulation of NO. It is expected that the lone pair and the dipoles may play dominant roles in determining the high-Tc superconductivity of nitride and oxide compounds. Furnished with the BOLS and BBB correlations and knowledge about the mentioned bonding events and the factors controlling bond switching, we

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

665

would be able to develop methods towards controllable bond making and breaking for designer materials with anticipated functions in the not too distant future.

Acknowledgements I am pleased to thank Professors P. J. Jennings, A. Stelbovics, S. M. Thurgate, G. Hitchen, S. Y. Tong, C. L. Bai, F. M. Ashby and E. Y. Jiang for valuable communications, which enlightened me very much. The VLEED data and calculation code provided by the Murdoch group has enabled the reported advances. I would also like to express my sincere thanks to Professor J. S. Colligon for his critical readings of the manuscript and his valuable input. Kind endorsement of the works linked to the current report by M. Donath, S. Tear, G. Russell, L. A. Bursill, L. Holland, O. S. Heavens, R. E. Hurley, and A. Mookerjee and helpful discussions with K. P. Loh, A. T. S. Wee, A. C. H. Huan are all gratefully acknowledged. I am indebt to S. Li, B. K. Tay, F. R. Zhu, K. Liao, Z. L. Dong, H. Huang, Y. Q. Fu, H. L. Bai, J. Zhou, W. T. Zheng and Q. Jiang for their support and input in one way or another. Permission reprinting diagrams from Elsevier Science, IOP, World Scientific, and AIP is also acknowledged.

References [1] Young FW, Cathcart JV, Gwathmey AT. The rates of oxidation of several faces of a single crystal of copper as determined with elliptically polarized light. Acta Metall 1956;4:45–152. [2] Ertl G. Untersuchung von oberflachenreaktionen mittels beugung langsamer elektronen (LEED). Surf Sci 1967;6:208–32. [3] Jacobsen KW, Nørskov JK. Theory of the oxygen-induced restructuring of Cu(110) and Cu(100) surfaces. Phys Rev Lett 1990;65:1788–891. [4] Liu P, Wang Y. A tight-binding molecular dynamic simulation of the Cu(110) surface covered with oxygen. Surf Sci 1999;440:81–6. [5] Liu P, Wang Y. Theoretical study on the structure of Cu(110)-p21-O reconstruction. J Phys: Condens Matter 2000;12(17):3955–66. [6] Tjeng LH, Meinders MBJ, Sawatzky GA. Relationship between atomic and electronic structure of clean and oxygen covered copper (110) surface. Surf Sci 1990;233(1–2):163–83. [7] Ricart JM, Torras J, Illas F, Rubio J. Bonding of atomic oxygen to Cu(110) and Ag(110) surfaces: a study of the nature of the interaction. Surf Sci 1994;307/309:107–12. [8] Ricart JM, Torras J, Clotet A, Sueiras JE. Atomic oxygen chemisorption on Cu(110) and Ag(110): an ab initio study. Surf Sci 1994;301(1–3):89–96. [9] Schimizu T, Tsukada M. Origin of the different formation modes of the oxygen added row overlayer on Ag(110) and Cu(110) surfaces. Surf Sci 1993;295(1-2):L1017–22. [10] Lynch M, Hu P. A density functional theory study of CO and atomic oxygen chemisorption on Pt(111). Surf Sci 2000;458(1–3):1–14. [11] McClenaghan ND, Hu P, Hardacre C. A density functional theory study of the surface relaxation and reactivity of Cu2O,100). Surf Sci 464(2–3 2000;223-232. [12] Zhang CJ, Hu P. Why must oxygen atoms be activated from hollow sites to bridge sites in catalytic CO oxidation? J Am Chem Soc 2000;122:2134–5. [13] Noguera C. Insulating oxides in low dimensionality: a theoretical review. Surf Rev Lett 2001;8:121–67.

666

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[14] Besenbacher F. STM studies of metal surfaces. Rep Prog Phys 1996;59:1737–802. [15] Sinnott MJ. The solid state for engineers. New York: John Wiley and Sons Inc; 1963. [16] Wintterlin J, Behm RJ. Adsorbate on metals. In: Gu¨nthert HJ, Wiesendanger R, editors. The scanning tunnelling microscopy I; 1992. Springer-Verlag, Berlin [Chapter 4]. Bai CL. Scanning tunnelling microscopy and its applications. Springer Series in Surface Science vol. 32. Springer: Berlin. [17] Bonnell DA. STM and STS of oxide surfaces. Prog Surf Sci 1998;57(3):187–252. [18] Van Hove MA, Cerda J, Sautet P, Bocquet ML, Salmeron M. Surface structure determination by STM vs LEED. Prog Surf Sci 1997;54:315–29. [19] Kittel M, Polcik M, Terborg R, Hoeft J-T, Baumgartel P, Bradshaw AM, et al. The structure of oxygen on Cu(100) at low and high coverages. Surf Sci 2001;470(3):311–25. [20] Henrich VE. Metal-oxide surfaces. Prog Surf Sci 1995;50:77–90. [21] Besenbacher F, Nørskov JK. Oxygen chemisorption on metal surfaces: general trends for Cu, Ni and Ag. Prog Surf Sci 1993;44:5–66. [22] King DA. Chemisorption on metals: a personal review. Surf Sci 1994;299/230:678–89. [23] Ertl G. Reactions at well-defined surfaces. Surf Sci 1994;299/230:742–54. [24] Van Hove MA, Somorjai GA. Adsorption and adsorbate-induced restructuring: a LEED perspective. Surf Sci 1994;299:487–501. [25] Liu W, Wong KC, Zeng HC, Mitchell KAR. What determines the structures formed by oxygen at low-index surfaces of copper. Prog Surf Sci 1995;50:247–57. [26] Leibsle FM, Murray PW, Condon NG, Thornton G. STM studies of reaction on metal surfaces and model oxide supports. J Phys D: Appl Phys 1997;30:741–56. [27] Comelli G, Dhanak VR, Kiskinova M, Prince K C, Rosei R. Oxygen and nitrogen interaction with rhodium single crystal surfaces. Surf Sci Rep 1998;32:165–231. [28] Over H. Crystallographic study of interaction between adspecies on metal surfaces. Prog Surf Sci 1998;58(4):249–376. [29] Freund H-J, Kuhenbeck H, Staemmler V. Oxide surfaces. Rep Prog Phys 1996;59:283–347. [30] Noguera C. Dipolar oxide surface. J Phys Condens Matter 2000;12:R367–410. [31] Gilles Renaud. Oxide surfaces and metal-oxide interfaces studied by grazing incidence X-ray scattering. Surf Sci Rep 1998;32:1–90. [32] Mattsson A, Panas I, Siegbahn P, Wahlgren U, Akeby H. Model studies of the chemisorption of hydrogen and oxygen on Cu(100). Phys Rev 1987;B36:7389–401. [33] Bagus PS, Illas F. Theoretical analysis of the bonding of oxygen to Cu(100). Phys Rev B 1990;42: 10852–7. [34] Fritsche L, Weimert B, Kranefeld H, Noffke J. Bonding in the Cu(110)-p(21)-O system analyzed in terms of Cu–O–Cu chain properties. Surf Sci 1993;291(3):309–16. [35] DiDio RA, Zehner DM, Plummer EW. An angle-resolved UPS study of the oxygen-induced reconstruction of Cu(110). J Vac Sci Technol 1984;A2(2):852–5. [36] Jensen F, Besenbacher F, Lagsgaard E, Stensgaard I. Kinetics of oxygen-induced reconstruction of Cu(100) studied by scanning tunneling microscopy. Phys Rev B 1990;42:9206–9. [37] Arvanitis D, Comelli G, Lederer T, Rabus H, Baberschke K. Characterization of two different adsorption states for O on Cu(100): ionic versus covalent bonding. Chem Phys Lett 1993;211:53–9. [38] Lederer T, Arvanitis D, Comelli G, Troger L, Baberschke K. Adsorption of oxygen on Cu(100). 1. Local-structure and kinetics for 2 atomic chemisorption states. Phys Rev B 1993;48(20):15390–404. [39] Stolbov S, Kara A, Rahman TS. Electronic structure of the c(22)O/Cu(001) system. Phys Rev B 2002;66:245–405. [40] Stolbov S, Rahman TS. Role of Long range interaction in oxygen superstructures formation on Cu(001) and Ni(001). Phys Rev Lett 2002;89:116101. [41] McCullen EF, Hsu CL, Tobin RG. Electron density changes and the surface resistivity of thin metal films: oxygen on Cu(100). Surf Sci 2001;481(1–3):198–204. [42] Koch R, Schwarz E, Schmidt K, Burg B, Christmann K, Rieder KH. Oxygen adsorption on Co(101
0): different reconstruction behavior of hcp(101
0) and fcc(110). Phys Rev Lett 1993;71:1047– 50.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

667

[43] Koch R, Burg B, Schmidt K, Rieder KH, Schwarz E, Christmann K. Oxygen adsorption on Co(101
0). The structure of p(21)2O. Chem Phys Lett 1994;220:172–6. [44] Taniguchi M, Tanaka K, Hashizume T, Sakurai T. Ordering of Ag–O chains on the Ag(110) surface. Surf Sci 1992;262(3):L123–8. [45] Eierdal L, Besenbacher F, Lagsgaard E, Stensgaard I. Interaction of oxygen with Ni(110) studied by STM. Surf Sci 1994;312:31–53. [46] Pascal M, Lamont CLA, Baumga¨rtel P, Terborg R, Hoeft JT, Schaff O, et al. Photoelectron diffraction study of the Ag(110)-(21)-O reconstruction. Surf Sci 2000;464(2–3):83–90. [47] Chua FM, Kuk Y, Silverman PJ. Oxygen chemisorption on Cu(110): an atomic view by scanning tunneling microscopy. Phys Rev Lett 1989;63:386–9. [48] Feidenhans’l R, Grey F, Nielsen M, Besenbacher F, Jensen F, Lægsgaard E, et al. Oxygen chemisorption on Cu(110): a model for the c(6  2) structure. Phys Rev Lett 1990;65:2027–30. [49] Jensen F, Besenbacher F, Lagsgaard E, Stensgaard I. Oxidation of Cu(111): two new oxygen induced reconstructions. Surf Sci 1991;259:L774–80. [50] Batra IP, Barker JA. Helium diffraction from adsorbate-covered surfaces: a study of the O–Ni(001) system. Phys Rev B 1984;29:5286–91. [51] Vu DT, Mitchell KAR, Warren OL, Thiel PA. Tensor LEED analysis of the Pd(100)p p ( 5 5)R27 -O surface structure. Surf Sci 1994;318(1–2):129–38. [52] Mercer JR, Finetti P, Leibsle FM, McGrath R, Dhanak VR, Baraldi A, et al. STM and SPA-LEED studies of O-induced structures on Rh(100) surfaces. Surf Sci 1996;352–354:173–8. [53] Gibson KD, Viste Mark, Sanchez Errol C, Sibener SJ. High density adsorbed oxygen on Rh(111) and enhanced routes to metallic oxidation using atomic oxygen. J Chem Phys 1999;110(6):2757–60. [54] Schwegmann S, Over H, De Renzi V, Ertl G. The atomic geometry of the O and CO+O phases on Rh(111). Surf Sci 1997;375:91–106. [55] Stampfl C, Schwegmann S, Over H, Scheffler M, Ertl G. Structure and stability of a high-coverage (11) oxygen phase on Ru(0001) Phys. Rev Lett 1996;77:3371–4. [56] Meinel K, Wolter H, Ammer C, Beckmann A, Neddermeyer H. Adsorption stages of O on Ru(0001) studied by means of STM. J Phys: Cond Matter 1997;9:4611–9. [57] Briggs D, Rivie`re JC. Practical surface analysis, vol. 1. USA: John Wiley and Sons Press; 1990. [58] Ertl G, Rhodin TN. The nature of surface chemical bond. USA: North-Holland Press; 1979. [59] Hu¨fner S. Photoelectron spectroscopy: principles and applications. USA: Springer-Verlag; 1995. [60] Yagi K, Higashiyama K, Fukutani H. Angle-resolved photoemission study of oxygen-induced c(24) structure on Pd(110). Surf Sci 1993;295:230–40. [61] Redhead PA. The first 50 years of electron stimulated desorption (1918–1968). Vacuum 1997;48:585– 96. [62] Bondzie VA, Kleban P, Dwyer DJ. XPS identification of the chemical state of subsurface oxygen in the O/Pd(110) system. Surf Sci 1996;347:319–28. [63] Schwegmann S, Seitsonen AP, De Renzi V, Dietrich D, Bludau H, Gierer M, et al. Oxygen adsorption on the Ru(101
0) surface: anomalous coverage dependence. Phys Rev B 1998;57(24): 15487–95. [64] He P, Jacobi K. Vibrational analysis of the 11-O overlayer on Ru(0001). Phys Rev 1997;B55: 4751–4. [65] Jennings PJ, Sun CQ. In: O’Connor DJ, Sexton BA, Smart RC, editors. In the surface analysis methods in materials science. 4th ed. New York: Springer-Verlag; 1992. [66] Adams DL, Nielsen HB, Andersen JN, Stensgaard I, Feidenhans’l R, Sørensen JE. Oscillatory relaxation of the Cu(110) surface. Phys Rev Lett 1982;49(9):669–72. [67] Frenken JWM, Van der Veen Allan JFG. Relation between surface relaxation and surface force constants in clean and oxygen-covered Ni(001). Phys Rev Lett 1983;51(20):1876–9. [68] Batra IP. Lattice relaxation in aluminum monolayers. J Vac Sci Techn 1985;A3(3):1603–6. [69] Gupta RP. Lattice relaxation at a metal surface. Phys Rev B 1981;23(12):6265–70. [70] Hwang Y-N, Park S-H, Kim D. Size-dependent surface phonon mode of CdSe quantum dots. Phys Rev B 1999;59(11):7285–8.

668

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[71] Li YS, Jona F, Marcus PM. Low-energy-electron-diffraction study of multilayer relaxation on a Pb{311} surface. Phys Rev B 1991;44(15):8267–73. [72] Feibelman PJ. Relaxation of hcp(0001) surfaces: a chemical view. Phys Rev B 1996;53(20):13740–6. [73] Stolbov S, Rahman TS. Relationship between electronic and geometric structures of the O/Cu(001) system. J Chem Phys 2002;117(18):8523–30. [74] Feidenhans’l R, Grey F, Johnson RL, Mochrie SGJ, Bohr J, Nielsen M. Oxygen chemisorption on Cu(110): a structural determination by X-ray diffraction. Phys Rev 1990;B41:5420–3. [75] Kiejna A, Lundqvist BI. First-principles study of surface and subsurface O structures at Al(111). Phys Rev 2001;B63:085405. [76] Kiejna A, Lundqvist BI. Stability of oxygen adsorption sites and ultrathin aluminum oxide films on Al(111). Surf Sci 2002;504(C):1–10. [77] Hansen LB, Stoltze P, Nørskov JK, Clausen BS, Niemann W. Is there a contraction of the interatomic distance in small metal particles? Phys Rev Lett 1990;64(26):3155–8. [78] Sun CQ. Oxygen-reduced inner potential and work function in VLEED. Vacuum 1997;48:865–9. [79] Rotermund HH. Investigation of dynamic processes in adsorbed layers by photoemission electron microscopy (PEEM). Surf Sci 1993;283:87–100. [80] Zhang J, Dowben PA, Li D, Onellion M. Angle-resolved photoemission study of oxygen chemisorption on Gd(0001) Surf. Sci 1995;329(3):177–83. [81] Thurgate SM, Sun CQ. Very low-energy electron diffraction analysis of oxygen on Cu(001). Phys Rev 1995;B51:2410–7. [82] Lindroos M, Pfnu¨r H, Held G, Menzel D. Adsorbate induced reconstruction by strong chemisorption: Ru(001)p(22)-O. Surf Sci 1989;222:451–63. [83] Pfnu¨r H, Lindroos M, Menzel D. Investigation of adsorbates with LEED at very low energies (VLEED). Surf Sci 1991;248:1–10. [84] Thurgate SM, Hitchen G, Sun CQ. Surface structural determination by VLEED analysis. In: MacDonald RJ, Taglauer EC, Wandelt KR, editors. The surface science: principles and applications. New York: Springer; 1996. p. 29. [85] Stampfl C, Scheffler M. Theoretical study of O adlayers on Ru(0001) Phys. Rev B 1996;54:2868–72. [86] Materer N, Starke U, Barbieri A, Do¨ll R, Heinz K, Van Hove MA. Reliability of detailed LEED structural analysis: Pt(111) and Pt(111)-p(22)-O. Surf Sci 1995;325:207–22. [87] Wang YM, Li YS, Mitchell KAR. LEED crystallographic analysis for the half monolayer structure formed by O(0001) at the Zr surface. Surf Sci 1995;342(1-3):272–80. [88] Sun CQ, Tay BK, Sun XW, Lau SP. Solution certainty of the Cu(110)-(2  1)-O2
crystallography. Int J Mod Phys B 2002;16(1-2):71–8. [89] Sun CQ, Bai CL. Modeling of non-uniform electrical potential barriers for metal surfaces with chemisorbed oxygen. J Phys: Condens. Matter 1997;9(27):5823–36. [90] Sun CQ, Bai CL. Oxygen-induced non-uniformity in surface electrical potential barrier. Mod Phys Lett 1997;B11(5):201–8. [91] Hoeft JT, Kittle M, Polcik M, Bao S, Toomes RL, Kang JH, et al. Molecular adsorption bond length at metal oxide surfaces: failure of current theoretical methods. Phys Rev Lett 2001;87(8): 086101. [92] Sun CQ. A model of bonding and band-forming for oxides and nitrides. Appl Phys Lett 1998; 72(14):1706–8. [93] Sun CQ, Bai CL. A model of bonding between oxygen and metal surfaces. J Phys Chem Solids 1997;58(6):903–12. [94] Sun CQ. What effects in nature the two-phase on the O–Cu(001)? Mod Phys Lett 1997;B11:81–6. [95] Sun CQ. Origin and processes of the O–Cu(001) and O–Cu(110) biphase ordering. Int J Mod Phys B12(10) 1998:951–64. [96] Sun CQ. Spectral sensitivity of the VLEED to the bonding geometry and the potential barrier of the O–Cu(001) surface. Vacuum 1997;48(5):491–8. [97] Sun CQ. Exposure-resolved VLEED from the O–Cu(001): bonding kinetics. Vacuum 1997;48:535–41. [98] Sun CQ. O–Cu(001): I. Binding the signatures of LEED, STM PES in a bond-forming way (review). Surf Rev Lett 2001;8(3):367–402.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

669

[99] Sun CQ. O–Cu(001): II. VLEED quantification of the four-stage Cu3O2 bond- forming kinetics (review). Surf Rev Lett 2001;8(6):703–34. [100] Sun CQ. The sp orbital hybrid bonding of C, N and O to the fcc(001) surface of nickel and rhodium (review). Surf Rev Lett 2000;7(3):347–63. [101] Goldschmidt VM. Ber Deut Chem Ges 1927;60:1270. [102] Pauling L. Atomic radii and interatomic distances in metals. J Am Chem Soc 1947;69(1):542–53. [103] Keenan CW, Kleinfelter DC, Wood JH. General college chemistry. 6th ed. USA: Harper and Row Publishers; 1979. [104] Han WG, Zhang CT. A theory of non-linear stretch vibration of hydrogen bonds. J Phys: Cond Mat 1991;3:27–35. [105] Crabtree RH. A new type of hydrogen bond. Science 1998;282:2000. [106] Hobza P, Havlas Z. The fluoroform. . .ethylene oxide complex exhibits a C–H. . .O anti-hydrogen bond. Chem Phys Lett 1999;303(3-4):447–52. [107] Atkins PW. Physical chemistry. 4th ed. Oxford: Oxford University Press; 1990. [108] Morrison SR. The chemical physics of surfaces. NY and London: Plenum Press; 1977. [109] Baraldi G, Silvano L, Comelli G. Oxygen adsorption and ordering on Ru(101
0). Phys Rev B63 2001:115410. [110] Over H, Kleinle G, Ertl G, Moritz W, Ernst KH, Wohlgemuth H, Christmann K, Schwarz E. LEED structural analysis of the Co(101
0) surface. Surf Sci 1991;254(1–3):L469–74. [111] Davis HL, Zehner DM. Structure of the clean Re(101
0) surface. J Vac Sci Technol 1980;17:190–3. [112] Halicioglu T. Calculation of surface energies for low index planes of diamond. Surf Sci 1991;259: L714–8. [113] Begley AM, Kim SK, Jona F, Marcus PM. Surface relaxation of Rh{001}. Phys Rev B 1993;48(16): 12326–9. [114] Teeter G, Erskine JL. Studies of clean metal surface relaxation experiment–theory discrepancies. Surf Rev Lett 1999;6:813–7. [115] Sokolov J, Jona F, Marcus PM. Parallel and perpendicular multilayer relaxation of Fe{310}. Phys Rev B 1985;31:1929–35. [116] Fu CL, Ohnishi S, Wimmer E, Freeman AJ. Energetics of surface multilayer relaxation on W(001): evidence for short-range screening. Phys Rev Lett 1984;53(7):675–8. [117] Teeter G, Erskine JL, Shi F, Van Hove MA. Surface roughness and LEED crystallography: analysis of flat and vicinal W(110). Phys Rev 1999;B60(3):1975–81. [118] Cho JH, Oh DH, Kleinman. Core-level shifts of low coordination atoms at the W(320) stepped surface. Phys Rev 2001;B64:115404. [119] Zhang XG, Van Hove MA, Somorjai GA, Rous PJ, Tobin D, Gonis D, et al. Efficient determination of multilayer relaxation in the Pt(210) and densely stepped kink surface. Phys Rev Lett 1991;67(10): 1298–303. [120] Tian Y, Lin K-W, Jona F. Anomalous multilayer relaxation on a Cu(331) surface. Phys Rev B 2000;62(19):12844–8. [121] Durukanoglu S, Kara A, Rahman TS. Local structural and vibrational properties of stepped surfaces: Cu(211), Cu(511), and Cu(331). Phys Rev 1997;B55(20):13894–993. [122] Adams DL, Jensen V, Sun XF, Vollesen JH. Multilayer relaxation of the Al(210) surface. Phys Rev B 1988;38(12):7913–31. [123] Cho JH, Terakura K. Plane-wave-basis pseudopotential calculations of the surface relaxations of Ti(0001) and Zr(0001). Phys Rev B 1997;56(15):9282–5. [124] Wei CY, Lewis SP, Mele EJ, Rappe AM. Structure and vibrations of the vicinal copper (211) surface. Phys Rev B 1998;57(16):10062–8. [125] Geng WT, Freeman AJ. Multilayer relaxation of Cu(331). Phys Rev B 2001;64:115401. [126] Seyller T, Diehl RD, Jona F. Low-energy electron diffraction study of the multilayer relaxation of Cu(211). J Vac Sci Technol A 1999;17(4):1635–8. [127] Bohnen KP, Ho KM. Structure and dynamics at metal surfaces. Surf Sci Rep 1993;19:99–120. [128] Rodriguez AM, Bozzolo G, Ferrante J. Structure and energetics of high index Fe, Al, Cu and Ni surfaces using equivalent crystal theory. Surf Sci 1994;307–309:625.

670

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[129] Walter S, Baier H, Weinelt M, Heinz K, Fauster Th. Quantitative determination of Cu(117) multilayer surface relaxations by LEED. Phys Rev B 2001;63:155407. [130] Tian ZJ, Rahman TS. Energetics of stepped Cu surfaces. Phys Rev B 1993;47(15):9751–9. [131] Qadri SB, Yang J, Ratna BR, Skelton EF, Hu JZ. Pressure induced structural transitions in nanometer size particles of PbS. Appl Phys Lett 1996;69(15):2205–7. [132] Gravil PA, Holloway S. Roughening and melting of stepped aluminum surfaces. Phys Rev B 1996; 53(16):11258–75. [133] Meyerheim HL, Sander D, Popescu R, Kirschner J, Steadman P, Ferrer S. Surface structure and stress in Fe monolayers on W(110). Phys Rev B 2001;64:045414. [134] Davis HL, Hannon JB, Ray KB, Plummer EW. Anomalous interplanar expansion at the surface of Be(0001). Phys Rev Lett 1992;68:2632–5. [135] Adams DL, Sorensen CS. Multilayer relaxation of the Al(331) surface. Surf Sci 1986;166:495–511. [136] Lang ND. Apparent barrier height in scanning tunneling microscopy. Phys Rev B 1988;37(17): 10395–8. [137] Lang ND. Small adsorption dipole moment need not imply small charge transfers. Surf Sci 1983; 127:L118–22. [138] Lang ND. Scanning tunneling microscopy III. Springer Series in Surf. Sci., vol. 29 Wiesendanger R, Gu¨ntherodt HJ, editors (Berlin), 1993, p. 7–12. [139] Ghijsen J, Tjeng LH, Van Elp J, Eskes H, Westerink J, Sawatzky GA, Czyzyk MT. Electronic structure of Cu2O and CuO. Phys Rev 1988;B38:11322–30. [140] Marabelli F, Parravicini GB, Salghetti-Drioli F. Optical gap of CuO. Phys Rev 1995;B52:1433–6. [141] Dose V. Momentum-resolved inverse photoemission. Surf Sci Rep 1986;5:337–78. [142] Rotermund HH, Lauterbach J, Haas G. The formation of subsurface oxygen on Pt(100). Appl Phys A: Solids and Surf 1993;57(6):507–11. [143] Lauterbach J, Rotermund HH. Spatio-temporal pattern formation during the catalytic CO-oxidation on Pt(100). Surf Sci 1994;311:231–46. [144] Lauterbach J, Asakura K, Rotermund HH. Subsurface oxygen on Pt(100): kinetics of the transition from chemisorbed to subsurface state and its reaction with CO, H2 and O2. Surf Sci 1994;313:52– 63. [145] Pendry JB. Low energy electron diffraction: the theory and its application to the determination of surface structure. London: Academic Press; 1974. [146] Jones RO, Jennings PJ, Jepsen O. Surface barrier in metals: a new model with application to W(001). Phys Rev B 1984;29(12):6474–80. [147] Jennings P. Private communication, Murdoch University, Perth; 1994. [148] Sun CQ. Modeling of bond structure and potential barrier of the O–Cu(100) surface characterized by STM/S and VLEED. PhD Thesis, Murdoch University Press; 1995. [149] Zeng HC, McFarlane RA, Sodhi RNS, Mitchell KAR. LEED crystallographic studies for the chemisorption of oxygen on the (100) surface of copper. Can J Chem 1988;66:2054–62. [150] Zeng HC, Mitchell KAR. Further LEED investigations of missing row models for the Cu(100)p p ( 22 2)R45 -O surface structure. Surf Sci 1990;239:L571–8. [151] Fujita T, Okawa Y, Matsumoto Y, Tanaka K. Phase boundaries of nanometer scale c(22)-O domains on the Cu(100) surface. Phys Rev 1996;B54:2167–74. [152] Atrei A, Bardi U, Rovida G, Zanazzi E, Casalone G. Test of structural models for Cu(001)p p ( 22 2)R45 -O by LEED intensity analysis. Vacuum 1990;41(1–3):333–6. [153] Robinson IK, Vlieg E. Oxygen-induced missing-row reconstruction of Cu(001) and Cu(001)-vicinal surfaces. Phys Rev B 1990;42:6954–62. [154] Asensio MC, Ashwin MJ, Kilcoyne ALD, Woodruff DP, Robinson AW, Lindner T, et al. The structure of oxygen adsorption phases on Cu(100). Surf Sci 1990;236:1–14. [155] Wuttig M, Franchy R, Ibach H. Oxygen on Cu(100): a new type of an adsorbate induced reconstruction. J Electr Spectro 1987;44:317–23. [156] Shen YG, Qayyum A, O’Connor DJ, King BV. Oxygen-induced surface (2  2)p4g reconstruction of Rh(001). Phys Rev B 1998;58(15):10025–30. [157] Tucker CW. Oxygen faceting on Rh(210) and Rh(100) surfaces. Acta Meta 1967;15:1465–74.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

671

[158] Ueda K, Takano A, Tanaka K. Oxygen adsorption study on Rh(100) surface by electron-stimulated desorption spectroscopy. Jpn J Appl Phys B 1995;34:3662–5. [159] Zacchigna M, Astaldi C, Prince KC, Sastry M, Comicioli C, Rosei R, et al. Photoemission from atomic and molecular adsorbates on Rh(100). Surf Sci 1996;347:53–62. [160] Alfe D, de Gironcoli S, Baroni S. The reconstruction of Rh(100) upon oxygen adsorption. Surf Sci 1998;410(2–3):151–7. [161] Baraldi A, Cerda´ J, Martı´n-Gago JA, Comelli G, Lizzit S, Paolucci G, et al. Oxygen induced reconstruction of the Rh(100) surface: general tendency towards threefold oxygen adsorption site on Rh surfaces. Phys Rev Lett 1999;82(24):4874–7. [162] Baraldi A, Dhanak VR, Comelli G, Prince KC, Rosei R. Order–disorder phase transition of oxygen on Rh(100). Phys Rev B56 1997:10511–7. [163] Baraldi A, Dhanak VR, Comelli G, Prince KC, Rosei R. O/Rh(100) p(22) ! c(22) order–disorder phase transition. Phys Rev 1996;B53:4073–7. [164] Dubois LH. Vibrational spectra of atomic adsorbates: carbon, oxygen, and sulfur on Rh(100). J Chem Phys 1992;77:5228–33. [165] Fisher GB, Schmieg SJ. The molecular and atomic states of oxygen adsorbed on Rh(100): adsorption. J Vac Sci Technol A 1983;1:1064–9. [166] Mercer JR, Finetti P, Scantlebury MJ, Beierlein U, Dhanak VR, McGrath R. Angle-resolved photoemission study of half-monolayer O and S structures on the Rh(100) surface. Phys Rev B 1997;55: 10014–21. [167] Sun CQ. Driving force behind the O–Rh(100) radial and clock reconstruction Mod. Phys Lett B 1998;12(20):849–57. [168] Wintterlin J, Schuster R, Coulman DJ, Ertl G, Behm RJ. Atomic motion and mass transport in the oxygen induced reconstructions of Cu(110). J Vac Sci Technol B 1991;9(2):902–8. [169] Bader M, Puschmann A, Ocal C, Haase J. Surface extended x-ray absorption fine-structure study of the O(2 1)/Cu(110) system: missing-row reconstruction and anisotropy in the surface mean free path and in the surface Debye–Waller factor. Phys Rev Lett 1986;57:3273–6. [170] Yarmoff JA, Donna MC, Huang JH, Kim S, Stanley WR. Impact-collision ion-scattering spectroscopy of Cu(110) and Cu(110)-(2  1)-O using 5-keV 6Li+. Phys Rev B 1986;33:3856–8. [171] Sun CQ. On the nature of the O–Rh(110) multiphase ordering. Surf Sci 1998;398(3):L320–0L326. [172] Stokbro K, Baroni S. Surface chemistry of metal–oxygen interactions: a first-principles study of O:Rh(110). Surf Sci 1997;370:166–78. [173] Schwarz E, Lenz J, Wohlgemuth H, Christmann K. Interaction of oxygen with a Rh(110) surface. Vacuum 1990;41:167–70. [174] Murray PW, Leibsle FM, Li Y, Guo Q, Bowker M, Thornton G, et al. STM Study of the oxygeninduced reconstruction of Rh(110). Phys Rev 1993;B47(19):12976–9. [175] Kundrotas PJ, Lapinskas S, Rosengren A. Theoretical study of the oxygen-induced missing-row reconstruction on the Rh(110) surface. Surf Sci 1997;377–379:7–10. [176] Gierer M, Over H, Ertl G, Wohlgemuth H, Schwarz E, Christmann K. LEED analysis of the Rh(110)-(2  2)-O phase. Surf Sci 1993;297:L73–8. [177] Dhanak VR, Prince KC, Rosei R, Murray PW, Leibsle FM, Bowker M, et al. STM study of oxygen on Rh(110). Phys Rev 1994;B49(8):5585–90. [178] Dhanak VR, Comelli G, Cautero G, Paolucci G, Prince KC, Kiskinova M, et al. (1  n) Reconstruction of the Rh(110) surface with N=2, 3, 4, 5. Chem Phys Lett 1992;188(3–4):237–40. [179] Comicioli C, Dhanak VR, Comelli G, Astaldi C, Prince KC, Rosei R, et al. Structure of Rh(110)(12) and Rh(110)-(22)p2mg-O surfaces. Chem Phys Lett 1993;214(5):438–42. [180] Comelli G, Dhanak VR, Kiskinova M, Paolucci G, Prince K C, Rosei R. Adsorption of oxygen on Rh(110) and reactivity of different overlayer structures. Surf Sci 260- 1992;270:360–4. [181] Comelli G, Dhanak VR, Kiskinova M, Pangher N, Paolucci G, Prince KC, et al. Adsorption of oxygen on Rh(110)—A LEED, Auger-electron spectroscopy and thermal-desorption study. Surf Sci 1992;260(1–3):7–13. [182] Batteas JD, Barbieri A, Starkey EK, Van Hove MA, Somorjai GA. Rh(110)-p2mg(21)-2O surface

672

[183] [184] [185] [186] [187] [188] [189] [190] [191] [192] [193] [194] [195] [196] [197] [198] [199] [200] [201] [202] [203] [204] [205] [206] [207] [208]

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685 structure determined by automated tensor LEED: structure changes with oxygen coverage. Surf Sci 1995;339:142–50. Sun CQ. On the nature of the O-Co(101
0) triphase ordering. Surf Rev Lett 1998;5(5):1023–8. Schwarz E, Ernst KH, Gonser-Buntrock C, Neuber M, Christmann K. Ordered oxygen phases on a Co(101
0) surface. Vacuum 1990;41(1–3):180–4. Gierer M, Over H, Pech P, Schwarz E, Christmann K. The adsorption geometry of the (21)-2O oxygen phase formed on the Co(101
0) surface. Surf Sci 1997;370:L201–6. Poluston S, Tikhov M, Lambert. The adsorption geometry of the (21)-2O oxygen phase formed on the Co(101
0) surface. Surf Rev Lett 1994;1:655. Orent TW, Hansen RS. Interactions of nitric-oxide and oxygen with Ru(1010) Surf. Sci 1977;67: 325–50. Jensen F, Besenbacher F, Lagsgaard E, Stensgaard I. Two new oxygen induced reconstructions on Cu(111). Surf Sci 269/ 1992;270:400–4. Sun CQ. Oxygen interaction with the Ru(0001) and Rh(111) surfaces: bond forming kinetics. Mod Phys Lett B 2000;14:219–28. Wong KC, Liu W, Mitchell KAR. LEED crystallographic analysis for the Rh(111)-(21)-O surface structure. Surf Sci 1996;360:137–43. Thiel PA, Yates Jr JT, Weinberg WH. The interaction of oxygen with the Rh(111) surface. Surf Sci 1979;82:22–44. Peterlinz KA, Sibener SJ. Absorption, adsorption, and desorption studies of the oxygen/Rh(111) system using O
2, NO and NO2. J Phys Chem 1995;99:2817–25. Gibson KD, Colonell JI, Sibener S. Velocity distribution of recombinatively desorbed O2 originating from surface and sub-surface oxygen/Rh(111). J Surf Sci 1995;343:L1151–5. Ganduglia-Pirovano MV, Scheffler M. Structural and electronic properties of chemisorbed oxygen on Rh(111). Phys Rev B 1999;59(23):15533–43. Narloch B, Held G, Menzel D. Structural arrangement by co-adsorption: a LEED IV determination of the Ru(001)-p(22) (2O+CO) structure. Surf Sci 1994;317:131–42. Fuggle JC, Madey TE, Steinkilberg M, Menzel D. Photoelectron spectroscopic studies of adsorption of CO and oxygen on Ru(001). Surf Sci 1975;52:521–41. Sun CQ. O-Ru(0001) surface bond-and-band formation. Surf Rev Lett 1998;5(2):465–71. Schmidt M, Wolter H, Wandelt K. Work-function oscillations during the surfactant induced layerby-layer growth of copper on oxygen pre-covered Ru(0001) Surf. Sci 1994;307:507–13. Schmid AK, Hwang RQ, Bartelt NC. Predicting the future of a complicated surface structure: grain coarsening of p(22)O/Ru(0001) Phys. Rev Lett 1998;80:2153–6. Pfnu¨r H, Held G, Lindroos M, Menzel D. Oxygen induced reconstruction of a close-packed surface: a LEED IV study on Ru(001)-p(21)O. Surf Sci 1989;220:43–58. Woodruff DP, Delchar TA. Modern techniques of surface science. Cambridge: Cambridge University Press; 1986. Nørskov JK. Theory of adsorption and adsorbate-induced reconstruction. Surf Sci 1994;299/300: 690–705. Kostov KL, Reauscher H, Menzel D. The role of coadsorbed oxygen in NCO formation on Ru(001). Surf Sci 287/ 1993;288:283–7. Mills DL. Theory of STM-induced enhancement of dynamic dipole moments on crystal surfaces. Phys Rev B 2002;65:125419. Kuk Y, Silverman PJ. Role of tip structure in scanning tunneling microscopy. Appl Phys Lett 1986; 48:1597. Wo¨ll C, Wilson RJ, Chiang S, Zeng HC, Mitchell KAR. Oxygen on Cu(100) surface structure studied by STM and by LEED multiple-scattering calculations. Phys Rev B 1990;42:11926–9. Tanaka K-I, Fujita T, Okawa Y. Oxygen induced order-disorder restructuring of a Cu(100) surface. Surf Sci 1998;401:L407–12. Hitchen G, Thurgate SM, Jennings P. A LEED fine structure study of oxygen adsorption on Cu(001) and Cu(111). Aust J Phys 1990;43:519–34.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

673

[209] Sun CQ. Time-resolved VLEED from the O–Cu(001): atomic processes of oxidation. Vacuum 1997; 48:525–30. [210] Van de Riet E, Smeets JBJ, Fluit JM, Niehaus A. The structure of a clean and oxygen covered copper surface studied by low energy ion scattering. Surf Sci 1989;214:111–40. [211] Bronckers RPN, De Wit AGJ. Reconstruction of the oxygen-covered Cu(110) surface identified with low energy Ne+ and H2O+ ion scattering. Surf Sci 1981;112(1–2):133–52. [212] Do¨bler U, Baberschke K, Haase J, Puschmann A. Azimuthal- and polar-angle-dependent surface extended x-ray-absorption fine-structure study: (2  1)O on Cu(110). Phys Rev Lett 1984;52:1437–40. [213] Do¨bler U, Baberschke UK, Vvedensky DD, Pendry JB. X-ray absorption near-edge structure of adsorbate-induced reconstruction (21)O on Cu(110). Surf Sci 1986;178:679–85. [214] Feidenhans’l R, Stensgaard I. Oxygen adsorption induced reconstruction of Cu(110) studied by high energy ion scattering. Surf Sci 1983;133:453–68. [215] Kuk Y, Chua FM, Silverman PJ, Meyer JA. O chemisorption on Cu(110) by STM. Phys Rev B 1990;41(18):12393–402. [216] Parkin SR, Zeng HC, Zhou MY, Mitchell KAR. LEED crystallographic determination for the Cu(110)2  1-O surface structure. Phys Rev B 1990;41:5432–5. [217] Du¨rr H, Schneider R, Fauster T. Added-row growth of the (21)O–Cu(110) reconstruction. Phys Rev B 1991;43(2):1802–4. [218] Du¨rr H, Fauster Th, Schneider R. Surface structure determination of the (2  1)O–Cu(110) reconstruction by low-energy ion scattering. Surf Sci 1991;244:237–46. [219] Jensen F, Besenbacher F, Lagsgaard E, Stensgaard I. Surface reconstruction of Cu(110) induced by oxygen chemisorption. Phys Rev B 1990;41:10233–6. [220] Coulman DJ, Wintterlin J, Behm RJ, Ertl G. Novel mechanism for the formation of chemisorption phases: the (21)O–Cu(110) ‘added-row’ reconstruction. Phys Rev Lett 1990;64:1761–4. [221] Kern K, Niehus H, Schatz A, Zeppenfeld P, Goerge J, Comsa G. Long-range spatial self-organization in the adsorbate-induced restructuring of surfaces: Cu{100}-(2  1)O. Phys Rev Lett 1991; 67:855–8. [222] Haase J, Hillert B, Bradshaw AM. Comment on ‘Oxygen chemisorption on Cu(110): an atomic view by scanning tunneling microscopy’. Phys Rev Lett 1990;64(25):3098–101. [223] Pouthier V, Ramseyer C, Girardet C, Zeppenfeld P, Diercks V, Halmer R. Characterization of the Cu(110)-(2  1)O reconstruction by means of molecular adsorption. Phys Rev B 1998;58(15):9998– 10002. [224] Vlieg E, Driver SM, Goedtkindt P, Knight PJ, Liu W, et al. Structure determination of Cu(410)-O using X-ray diffraction and DFT calculations. Surf Sci 2002;516(1–2):16–32. [225] Coulman DJ, Wintterlin J, Barth JV, Ertl G, Behm RJ. An STM investigation of the Cu(110)-C(6  2)O system. Surf Sci 1990;240:151–62. [226] Feidenhans’l R, Grey F, Johnson RL, Nielsen M. Determination of the Cu(110)-c(62)-O structure by X-ray diffraction. Phys Rev B 1991;44:1875–9. [227] Liu W, Wong K C, Mitchell K A R. Structural details for the Cu(110)-c(6 2)-O surface determined by tensor LEED. Surf Sci 1995;339(1–2):151–8. [228] Dorenbos G, Breeman M, Boerma DO. Low-energy ion-scattering study of the oxygen-induced reconstructed p(2  1) and c(6  2) surfaces of Cu(110). Phys Rev B 1993;47(3):1580–8. [229] Thurgate SM, Jennings PJ. Effects of oxygen adsorption on the LEED fine structure features of Cu(001), Cu(110), Cu(111) and Ni(001). Surf Sci 1983;131:309–20. [230] Dubois LH. Oxygen chemisorption and porous oxide formation on Cu(111):a high resolution EELS study. Surf Sci 1982;119:399–410. [231] Niehus H. Surface reconstruction of Cu(111) upon oxygen adsorption. Surf Sci 1983;130:41–9. [232] Haase J, Kuhr H-J. Reconstruction and relaxation of the oxygen covered Cu(111) surface: a SEXAFS study. Surf Sci 1988;203:L695–699. [233] Jacob W, Dose V, Goldmann A. Atomic adsorption of oxygen on Cu(111) and Cu(110). Appl Phys A 1986;41:145–50. [234] Matsumoto T, Bennett RA, Stone P, Yamada T, Domen K, Bowker M. Scanning tunneling microscopy studies of oxygen adsorption on Cu(111). Surf Sci 2001;471(1–3):225–45.

674

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[235] Pauling L. The nature of the chemical bond. 3rd ed. Ithaca, NY: Cornell University Press; 1960. [236] Janin E, Schenck H, Von Go¨thelid M, Karlsson UO, Svensson M. Bridge-bonded atomic oxygen on Pt(110). Phys Rev B 2000;61(19):13144–9. [237] Schoeder T, Giorgi JB, Hammoudeh A, Magg N, Ba¨umer M, Freund H-J. Oxygen-induced p(23) reconstruction on Mo(112) studied by LEED and STM. Phys Rev B 2002;65:115411. [238] Santra AK, Min BK, Goodman DW. Oxygen-induced p(1x3)-O reconstruction on Mo(112): a precursor to the epitaxial formation of MoO2(100). Surf Sci 2002;513(3):L441–4. [239] Tan KC, Guo YP, Wee ATS, Huan CHA. (21) oxygen superstructure on Cu(210) surface studied by quantitative LEED analysis and STM. Surf Rev Lett 1999;6:859–63. Guo YP, Tan KC, Wang HQ, Huan CHA, Wee ATS. Low-energy electron diffraction study of oxygen-induced reconstructions on Cu(210). Phys Revb 2002;B66:165410. [240] Sokolov J, Jona F, Marcus PM. Europhys Lett 1986;1:401. [241] Wolter H, Meinel K, Ammer Ch, Wandelt K, Neddermeyer H. O-mediated layer growth of Cu on Ru(0001). J Phys: Condens Matter 1999;11:19–37. [242] Shen YG, O’Connor DJ, Vanzee H, Wandelt K, Macdonald RJ. The growth of thin Cu films on an O-precovered Ru(0001) surface studied by low-energy ion-beams. Thin Solid Films 1995; 263(1):72–8. [243] Hitchen G, Thurgate S. Determination of azimuth angle, incidence angle, and contact-potential difference for LEED fine-structure measurements. Phys Rev B 1988;38(13):8668–72. [244] Sun CQ. Coincidence in angular-resolved VLEED: Brilluion zones, atomic shift, and energy bands. Vacuum 1997;48(6):543–6. [245] Sun CQ. Angular-resolved VLEED from O–Cu(001): valence bands, chemical bonds, potential barrier, and energy states. Int J Mod Phys 1997;B11(25):3073–91. [246] Tanaka H, Yoshinobu J, Kawai M. Oxygen-induced reconstruction of the Pd(110) surface—an STM study. Surf Sci 327(1–2 1995;L505–L509. [247] Ladas S. Kinetic oscillation during the catalytic CO oxidation on Pd(110): the role of subsurface oxygen. Surf Sci 1989;219:88–106. [248] Nishijima M. Electron energy loss spectra of a Pd(110) clean surfaces. Solid State Commun 1986; 58:75–82. [249] Sun CQ. Nature and kinetics of the O-Pd(110) surface bonding. Vacuum 1998;49(3):227–32. [250] He J-W, Ulrich M, Norton PR. Interaction of oxygen with a Pd(110) surface. II. Kinetics and energetics. J Chem Phys 1989;90:5088–93. [251] He J-W, Ulrich M, Griffiths K, Norton PR. Interaction of oxygen with a Pd(110) surface. I. Structures and coverages. J Chem Phys 1989;90:5082–7. [252] Jo M. Oxygen adsorption on Pd(110) at 300 K low energy electron diffraction and electron energy loss spectroscopy study. Chem Phys Lett 1986;131:106–11. [253] Tucker CW. Chemisorbed oxygen structures on the rhodium (110) surface. J Appl Phys 1966; 37(11):4147–55. [254] Alfe D, Rudolf P, Kiskinova M, Rosei R. HREED spectra of various oxygen structures on Rh(110). Chem Phys Lett 1993;211(2–3):220–6. [255] Brena B, Comelli G, Ursella L, Paolucci G. Oxygen on Pd(110)-substrate reconstruction and adsorbate geometry by tensor LEED. Surf Sci 1997;375:150–60. [256] Bondino F, Comelli G, Baraldi A, Rosei R. Photoelectron diffraction study of the low-temperature low-coverage oxygen layer on Rh(110). Phys Rev B 2002;66:075402. [257] Schwegmann S, Seitsonen AP, Dietrich H, Bludau H, Over H, Jacobi K, et al. The adsorption of atomic nitrogen on Ru(0001): geometry and energetics. Chem Phys Lett 1997;264(6):680–6. [258] Castner DG, Somorjai GA. LEED, AES and thermal desorption studies of the oxidation of the rhodium (111) surface. Appl Surf Sci 1980;6:29–38. [259] Starke U, Van Hove MA, Somorjai GA. Adsorbate-induced relaxation of close-packed fcc and hcp metal surfaces. Prog Surf Sci 1994;46:304–19. [260] Todorova M, Li WX, Ganduglia-Pirovano MV, Stampfl C, Reuter K, Scheffler M. Role of subsurface oxygen in oxide formation at transition metal surfaces. Phys Rev Lett 2002;89:096103. [261] Quinn P, Brown D, Woodruff DP, Noakes TCQ, Bailey P. Structural analysis of the

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[262] [263] [264] [265] [266] [267] [268]

[269] [270] [271] [272] [273] [274]

[275] [276] [277] [278] [279] [280]

[281] [282]

[283]

[284]

675

Ru(0001)(11)-O and Ru(0001)(21)-O structures by medium energy in scattering. Surf Sci 2001; 491(1-2):208–18. Over H, Kim YD, Seitsonen AP, Wendt S, Lundgren E, Schmid M, et al. Atomic-scale structure and catalytic reactivity of the RuO2(110) surface. Science 2000;287:1474–6. Madey TE, Engelhardt HA, Menzel D. Absorption of oxygen and oxidation of CO on the ruthenium (001) surface. Surf Sci 1975;48:304–28. Renisch S, Schuster R, Wintterlin J, Ertl G. Kinetics of adatom motion under the influence of mutual interactions: O/Ru(0001) Phys. Rev Lett 1999;82(19):3839–42. Huang L, Chevrier J, Zeppenfeld P, Cosma G. Observation by STM of hexagonal Au(111) surface reconstruction induced by oxygen. Appl Phys Lett 1995;66(8):935–7. Carlisle CI, Fujimoto T, Sim WS, King DA. Atomic imaging of the transition between oxygen chemisorption and oxide film growth on Ag{111}. Surf Sci 2000;470(1–2):15–31. Sun CQ, Xie H, Zhang W, Ye H, Hing P. Preferential oxidation of diamond {111}. J Phys D 2000; 33(17):2196–9. Egelhoff Jr WF, Chen PJ, McMichael RD, Powell CJ, Deslattes RD, Serpa FG, et al. Surface oxidation as a diffusion barrier for Al deposited on ferromagnetic metals. Journal of Applied Physics 2001;89(9):5209–14. Ganduglia-Pirovano MV, Reuter K, Scheffler M. Stability of subsurface oxygen at Rh(111). Phys Rev B 2002;65:245426. Reuter K, Ganduglia-Pirovano MV, Stampfl C, Scheffler M. Metastable precursors during the oxidation of the Ru(0001) surface. Phys Rev B 2002;65:165403. Daniel WM, Kim Y, Peebles HC, White JM. Adsorption of Ag, O2 and N2O on Rh(100). Surf Sci 1981;111:189–204. Heinz K, Oed W, Pendry JB. Direct low-energy electron-diffraction analysis of c(2  2)O/Ni(100) including substrate. Phys Rev B 1990;41:10179. Chubb SR, Marcus PM, Heinz K, Mu¨ller K. Adsorption-induced relaxation of Ni(001)-c(2  2)-O. Phys Rev B 1990;41:5417. Norman D, Sto¨hr J, Jaeger R, Durham PJ, Pendry JB. Determination of local atomic arrangements at surfaces from near-edge X-ray absorption fine-structure studies: O on Ni(100). Phys Rev Lett 1983;51:2052. Ho¨rmandinger G, Pendry JB, Leibsle FM, Murray PW, Joyner RW, Thornton G. Scanningtunneling-microscopy investigation of the Ni(100)-p(2  2)C surface. Phys Rev B 1993;48:8356. Onuferko JH, Woodruff DP, Holland BW. LEED structure analysis of the Ni{100}-(22)p4g-C structure: a case of adsorbate-induced substrate distortion. Surf Sci 1979;87:357–74. Alfe D, de Gironcoli S, Baroni S. The reconstruction of nickel and rhodium (001) surfaces upon carbon, nitrogen and oxygen adsorptions. Surf Sci 1999;437:18–28. Norris AG, Schedin F, Thornton G, Dhanak VR, Turner TS, McGrath R. Surface X-ray diffraction study of the Rh(100)(22)-O reconstruction. Phys Rev 2000;B62(3):2113–7. Kirsch JE, Harris S. Electronic structure studies of Ni(001) surface reconstructions resulting from carbon, nitrogen, or oxygen atom adsorption. Surf Sci 2003;522:125–42. Wenzel L, Arvanitis D, Daum W, Rotermund HHJ, Stohr J, Babershke K, et al. Structural determination of an adsorbate-induced surface reconstruction: p4g(2  2)N versus c(2  2)O on Ni(100). Phys Rev B 1987;36:7689. Arvanitis D, Babershke K, Wenzel L. Multiple-scattering effects in surface extended X-ray absorption fine structure. Phys Rev B 1988;37:7143. Kilcoyne ALD, Woodruff DP, Robinson AW, Lindner T, Somers JS, Bradshaw AM. Photoelectron diffraction study of the Ni(100) (2 2)-C(p4g) and Ni(100) (2 2)-N(p4g) structures. Surf Sci 1991;253:107. Dudzik E, Norris AG, McGrath R, Charlton G, Thornton G, Murphy B, et al. The Ni(100)(22)p4g-N reconstruction determined by surface X-ray diffraction. Surf Sci 1999;433/435: 317. Bader M, Ocal C, Hillert B, Hass J, Bradshaw AM. Surface extended-x-ray-absorption finestructure study at the carbon K edge: the p4g(22)-C/Ni(100) system. Phys Rev B 1987;35:5900.

676

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[285] Gauthier Y, Baudoing-Savois R, Heinz K, Landskron H. Structure determination of p4g Ni(100)(2 multiplied by 2)C by LEED. Surf Sci 251/ 1991;252:493–7. [286] Leibsle FM. STM study of the (22) p4g nitrogen-induced surface reconstruction on Ni(100). Surf Sci 1993;297:98–105. [287] Klink C, Olesen L, Besenbacher F, Stensgaad I, Lagsgaard E, Lang ND. Interaction of C with Ni(100): atomic-resolved studies of the ‘clock’ reconstruction. Phys Rev Lett 1993;71:4350–3. [288] Sun CQ. Mechanism for the N–Ni(100) clock reconstruction. Vacuum 1999;52(3):347–51. [289] Sun CQ, Hing P. Driving force and bond strain for the C-Ni(100) surface reaction. Surf Rev Lett 1999;6(1):109–14. [290] Fang CSA. Surface structural transition of absorption of oxygen on Ag(100). Surf Sci 1990;235: L291–4. [291] Rocca M, Savio L, Vattuone L, Burghaus U, Palomba V, Novelli N, et al. Phase transition of dissociatively adsorbed oxygen on Ag(001). Phys Rev B 2000;61(1):213–27. [292] Cipriani G, Loffreda D, Dal Corso A, de Dironcoli S, Baroni S. Adsorption of atomic oxygen on Ag(001): a study based on density-functional theory. Surf Sci 2002;501(3):182–90. [293] Savio L, Vattuone L, Rocca M, De Renzi V, Gardonio S, Mariani C, et al. Substrate reconstruction and electronic surface states: Ag(001). Surf Sci 2001;486(1–2):65–72. [294] Tjeng LH, Meinders MB, Sawatzky GA. Electronic structure of clean and oxygen covered silver (110) surface. Surf Sci 1990;223:341–68. [295] Felter TE, Weinberg WH, Ya Lastushkina G, Boronin AI, Zhdan PA, Borskov GK, et al. A LEED study of the (100) and (111) surfaces of the intermetallic. Surf Sci 1982;118:369–79. [296] Koller R, Bergermayer W, Kresse G, Hebenstreit ELD, Konvicka C, Schmid M, et al. The structure of the oxygen induced (15) reconstruction of V(100). Surf Sci 2001;480:11–24. [297] Kuk Y, Silverman PJ. Scanning tunneling microscope instrumentation. Rev Sci Instrum 1989;60: 165–80. [298] Pendry JB, Pretre AB, Krutzen BCH. Theory of the scanning tunnelling microscope. J Phys: Condens Matter 1991;3:4313–21. [299] Sun CQ, Li S. Oxygen derived DOS features in the valence band of metals. Surf Rev Lett 2000;7(3): L213–7. [300] Sesselmann W, Conrad H, Ertl G, Ku¨ppers J, Woratschek B, Haberland H. Probing the local density of states of metal surfaces by de-excitation of metastable noble-gas atoms. Phys Rev Lett 1983;50:446–50. [301] Belash VP, Klimova IN, Kormilets VI, Trubjtsjn V Y, Finkelstein L D. Transformation of the electronic structure of Cu into Cu2O in the adsorption of oxygen. Surf Rev Lett 1999;6(3/4):383–8. [302] Courths R, Cord B, Wern H, Saalfeld H, Hu¨fner S. Dispersion of the oxygen-induced bands on Cu(110)-an angle-resolved UPS study of the system p(21)O/Cu(110). Solid State Commun 1987; 63(7):619–23. [303] Uehara Y, Matsumoto T, Ushioda SIdentification of O atoms on a Cu(001) surface by scanning tunneling microscope light emission spectra. Phys Rev 2002;B66:075413. [304] Su¨rgers C, Scho¨ck M, Lo¨hneysen HV. Oxygen-induced surface structure of Nb(110). Surf Sci 2001; 471:209–18. [305] Binnig G, Frank KH, Fuchs H, Garcia N, Reihl B, Rohrer H, et al. Tunneling spectroscopy and inverse photoemission: image and field states. Phys Rev Lett 1985;55:991. [306] Jung T, Mo YW, Himpsel FJ. Identification of metals in scanning tunneling microscopy via image states. Phys Rev Lett 1995;74:1641. [307] Portlupi M, Duo L, Isella G, Bertacco R, Marcon M, Ciccacci F. Electronic structure of epitaxial thin NiO(001) films grow on Ag (0001): towards a firm experimental basis. Phys Rev B 2001;64: 165402. [308] Pantel R, Bujor M, Bardolle J. Continuous measurement of surface potential variations during oxygen adsorption on the (100), (110) and (111) faces of niobium using mirror electron microscope. Surf Sci 1977;62:589–609. [309] Warren S, Flavell W R, Thomas AG, Hollingworth J, Wincott PL, Prime AF, et al. Photoemission studies of single crystal CuO(100). J Phys: Condens Matter 1999;11(26):5021–43.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

677

[310] Emberly EG, Kirczenow G. Anti-resonances in molecular wires. J Phys: Condens Matter 1999;11: 6911–26. [311] Chen CT, Smith NV. Unoccupied surface states on clean and oxygen-covered Cu(110) and Cu(111). Phys Rev B 1989;40(11):7487–90. [312] Pillo T, Zimmermann R, Steiner P, Hufner S. The electronic structure of PdO found by photoemission (UPS and XPS) and inverse photoemission (BIS). J Phys Condens Matter 1997;9(19): 3987–9912. [313] Yagi K, Fukutani H. Oxygen adsorption site of Pd(110)c(2*4)-O: analysis of ARUPS compared with STM image. Surf Sci 1998;412-413:489–94. [314] Ozawa R, Yamane A, Morikawa K, Ohwada M, Suzuki K, Fukutani H. Angle-resolved UPS study of the oxygen-induced 21 surface of Cu(110). Surf Sci 1996;346(1–3):237–42. [315] Wang Y, Wei X, Tian Z, Cao Y, Zhai R, Ushikubo T, Sato K, Zhuang S. An AES, UPS and HREELS study of the oxidation and reaction of Nb(110). Surf Sci 1997;372(1–3):L285–90. [316] Boronin AI, Koscheev SV, Zhidomirov GM. XPS and UPS study of oxygen states on silver. J Elec Spec Rel Phen 1998;96(1–3):43–51. [317] Aprelev AM, Grazhulis VA, Ionov AM, Lisachenko AA. UPS (8.43 and 21.2 eV) data on the evolution of DOS spectra near EE of Bi2Sr2CaCu2O8 under thermal and light treatments. Physica C 1994;235–240(pt.2):1015–6. [318] Howard A, Clark DNS, Mitchell CEJ, Egdell RG, Dhanak VR. Initial and final state effects in photoemission from Au nanocluaters on TiO2(110). Surf Sci 2002;518:210–24. [319] Tillborg H, Nilsson A, Hernnas B, Martensson N. O/Cu(100) studied by core level spectroscopy. Surf Sci 1992;260–70:300–4. [320] Lizzit S, Baraldi A, Groso A, Reuter K, Ganduglia-Pirovano MV, Stampfl C, et al. Surface corelevel shifts of clean and oxygen-covered Ru(0001). Phys Rev B 2001;63:205419. [321] Yang DQ, Sacher E. Initial- and final-state effects on metal cluster/substrate interactions, as determined by XPS: copper clusters on Dow Cyclotene and highly oriented pyrolytic graphite. Appl Surf Sci 2002;195:187–95. [322] Pforte F, Gerlach A, Goldmann A, Matzdorf R, Braun J, Postnikov A. Wave-vector-dependent symmetry analysis of a photoemission matrix element: the quasi-one-dimensional model system Cu(110)(21)O. Phys Rev B 2001;63(16):165405. [323] Benndorf C, Egert B, Keller B, Seidel H, Thieme F. Oxygen interaction with Cu(100) studied by AES, ELS, LEED and work function changes. J Phys Chem Solids 1979;40:877–86. [324] Benndorf C, Egert B, Keller G, Thieme F. The initial oxidation of Cu(001) single crystal surfaces: an electron spectroscopic investigation. Surf Sci 1978;74:216–28. [325] Sun CQ. A bond-and-band model for the behavior of nitrides. Mod Phys Lett 1997;11:1021–9. [326] Fu YQ, Sun CQ, Du HJ, Yan BB. Crystalline CN forms harder than the hexagonal SiCN crystallite. J Phys D Appl Phys 2001;34:1430–5. [327] Sun C Q, Tay BK, Lau SP, Sun XW, Zeng XT, Bai HL, et al. Bond contraction and lone pair interaction at nitride surfaces. J Appl Phys 2001;90(5):L2615–7. [328] Terrones M, Ajayan PM, Banhart F, Blase X, Carroll DL, Charlier JC, et al. N-doping and coalescence of carbon nanotubes: synthesis and electronic properties. Applied Physics A 2002;74(3): 355–61. [329] Tillborg H, Nilsson A, Wiell T, Wassdahl N, Ma˚rtensson N, Nordgren J. Electronic structure of atomic oxygen adsorbed on Ni(100) and Cu(100) studied by soft-X-ray emission and photoelectron spectroscopies. Phys Rev B 1993;47(24):16464–70. [330] Spitzer A, Luth H, The adsorption of oxygen on copper surface, I. Cu(100) and Cu(110). Surf Sci 1982;118:121–35. The adsorption of oxygen on copper surface, II. Cu(111). Surf Sci 1982;118:136–44. [331] Perrella AC, Rippard WH, Mather PG, Plisch MJ, Buhrman RA. Scanning tunneling spectroscopy and ballistic electron emission microscopy studies of aluminum-oxide surfaces. Phys Rev 2002;B65: 201403(R). [332] Bo¨ttcher A, Niehus H. Oxygen adsorption on oxidized Ru(0001) Phys. Rev B 1999;60(20):14396– 404.

678

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[333] Bo¨ttcher A, Conrad H, Niehus H. Reactivity of oxygen phases created by the high temperature oxidation of Ru(0001) Surf. Sci 2000;452:125–32. [334] Bester G, Fa¨hnle M. On the elelctronic structure of the pure and oxygen covered Ru(0001) surface. Surf Sci 2002;497(1–3):305–10. [335] Stampfl C, Ganduglia-Pirovano MV, Reuter K, Scheffler M. Catalysis and corrosion: the theoretical surface-science context. Surf Sci 2002;500:368–94. [336] Altieri S, Tjeng LH, Sawatzky GA. Electronic structure and chemical reactivity of oxide-metal interfaces: MgO(001)/Ag(001). Phys Rev B 2000;61(24):16948–55. [337] Mamy R. Spectroscopic study of the surface oxidation of a thin epitaxial Co layer. Appl Surf Sci 2000;158:353–6. [338] Zheng JC, Xie XN, Wee ATS, Loh KP. Oxygen induces surface states on diamond (100). Diamond and Rel Mater 2001;10:500–5. [339] Lin LW. The role of oxygen and fluorine in the electron emission of some kinds of cathodes. J Vac Sci Technol 1988;A6(3):1053–7. [340] Tibbetts GG, Burkstrand JM, Tracy JC. Electronic properties of adsorbed layers of nitrogen, oxygen, and sulfur on copper (100). Phys Rev B 1977;15(8):3652–60. [341] Fuentes GG, Elizalde E, Sanz JM. Optical and electronic properties of TiCxNy films. J Appl Phys 2001;90(6):2737–43. [342] Souto S, M Pickholz MC, dos Santos, Alvarez F. Electronic structure of nitrogen-carbon alloys (a-CNx) determined by photoelectron spectroscopy. Phys Rev 1998;B57(4):2536–40. [343] Chen ZY, Zhao JP, Yano T, Ooie T. Valence band electronic structure of carbon nitride from X-ray photoelectron spectroscopy. J Appl Phys 2002;92:281–7. [344] Tibbetts GG, Burkstrand JM. Electronic properties of adsorbed layers of nitrogen, oxygen, and sulfur on silver (111). Phys Rev B 1977;16(4):1536–41. [345] Zhang G, Altman EI. The oxidation mechanism of Pd(001). Surf Sci 2002;504(c):253–70. [346] Bowker M, Guo QM, Joyner RW. Oxidation of carbon monoxide catalysed by rhodium (110). Surf Sci 1993;280:50–62. [347] Craig Jr JH. Characteristic energy of electronically desorbed ions from Rh(001). Appl Surf Sci 1989;35:520–6. [348] Gorodetskii VV, Niuwenhuys BE, Aachtler WHH, Boreskov WMH. Adsorption of oxygen and its reactions with carbon monoxide and hydrogen on rhodium surfaces: comparison with platinum and iridium. Appl Surf Sci 1981;7:355–71. [349] Salanov AN, Savchenko VI. Oxygen interaction with rhodium at low pressure. Surf Sci 1993;296: 393–9. [350] Hansen EW, Neurock M. First-principles-based Monte Carlo methodology applied to O/Rh(100). Surf Sci 2000;464(2-3):91–107. [351] Logan AD, Datye AK, Houston JE. The nature of the oxygen phase formated during progressive oxidation of the rhodium (111) surface. Surf Sci 1991;245:280–8. [352] Bo¨ttcher A, Conrad H, Niehus H. Thermal arrangement of oxygen adsorbed on oxygen-rich Ru(0001) Surf. Sci 2001;478:229–39. [353] Piercy P, Maier M, Pfnu¨r H. The structure of surface II, Vol. 11, p480. [Van der Veen JF, Van Hove MA, editors.]. Berlin: Springer; 1986. [354] Bassett MR, Imbihl R. Mathematical modeling of kinetic oscillations in the catalytic CO oxidation on Pd(110): The subsurface oxygen model. J Chem Phys 1990;93(1):811–21. [355] Rar A. Desorption and dissociation of oxygen admolecules on a stepped platinum (533) surface. Surf Sci 1994;318:89–96. [356] Gottfried JM, Elghobashi N, Schroeder SLM, Christmann K. Oxidation of gold by oxygen-ion sputtering. Surf Sci 2002;523:89–102. [357] Moritz T, Menzel D, Widdra W. Collective vibrational modes of the high-density (11)-O phase on Ru(001). Surf Sci 1999;427/428:64–8. [358] Frank M, Wolter K, Magg N, Heemeier M, Ku¨hnemuth R, Ba¨umer M, Freund H-J. Phonons of clean and metal-modified oxide films: an infrared and HREELS study. Surf Sci 2001;492(3): 270–84.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

679

[359] Root TW, Fisher GB, Schmidt LD. Electron energy loss characterization of NO on Rh(111). I. NO coordination and dissociation. J Chem Phys 1986;85:4679–87. [360] Pasquini L, Barla A, Chumakov AI, Leupold O, Ru¨ffer R, Deriu A, et al. Size and oxidation effects on the vibrational properties of anocrystalline a-Fe. Phys Rev B 2002;66:073410. [361] Caceres D, Vergara I, Gonzalez R, Monroy E, Calle F, Munoz E, Omnes F. Nanoindentation of AlGaN thin films. J Appl Phys 1999;86(12):6773–8. [362] Sjo¨stro¨m H, Stafstro¨m S, Boman M, Sundgren J-E. Superhard and elastic carbon nitride thin films having fullerenelike microstructure. Phys Rev Lett 1995;75:1336. [363] Kalki K, Schick M, Ceballos G, Wandelt K. Thin-film growth on an O-precovered Ru(0001) surface. Thin Solid Films 1993;228(1–2):36–9. [364] Schmidt M, Wolter H, Schick M, Kalki K, Wandelt K. Compression phases in copper/oxygen coadsorption layers on a Ru(0001) surface. Surf Sci 1993;287-B:983–7. [365] Meinel K, Ammer Ch, Mitte M, Wolter H, Neddermeyer H. Effects and structures of the O/Cu surfacant layer in O-mediated film growth of Cu on Ru(0001). Prog Surf Sci 2001;67:183–203. [366] Karolewski MA. Determination of growth modes of Cu on O/Ni(100) and NiO(100). Surf Sci 2002; 517(1–3):138–50. [367] Wulfhekel W, Lipkin NN, Kliewer J, Rosenfeld G, Jorritsma LC, Poelsema B, et al. Conventional and manipulated growth of Cu/Cu(111). Surf Sci 1996;348(3):227–42. [368] Yata M, Rouch H, Nakamura K. Kinetics of oxygen surfactant in Cu(001) homoepitaxial growth. Phys Rev B 1997;56(16):10579–84. [369] Whitten JE, Gomer R. Reactivity of Ni on oxygen-covered W(110) surfaces. J Vac Sci Technol A 1995;13:2540. [370] Edelstein AS, Cammarata RC. Nanomaterials: synthesis, properties and applications. Bristol: Institute of Physics; 1996. [371] Gleiter H. Prog Mater Sci 1989;33:223. [372] Goldstein AN, Echer CM, Alivisatos AP. Science 1992;256:1425. [373] Lu L, Sui ML, Lu K. Superplastic extensibility of nanocrystalline copper at room temperature. Science 2000;287(5457):1463–6. [374] Christenson HK. Confinement effects on freezing and melting. J Phys Condens Matt 2001;13:R95. [375] Mintova S, Olson NH, Valtchev V, Bein T. Mechanism of zeolite a nanocrystal growth from colloids at room temperature. Science 1999;283:958–60. [376] Campbell CT, Parker SC, Starr DE. The effect of size-dependent nanoparticle energetics on catalyst sintering. Science 2002;298:811. [377] Horch S, Lorensen HT, Helveg S, Laegsgaard E, Stensgaard I, Jacobsen KW, et al. Enhancement of surface self-diffusion of platinum atoms by adsorbed hydrogen. Nature 1999;398(6723):134–6. [378] van Buuren T, Dinh LN, Chase LL, Siekhaus WJ, Terminello LJ. Changes in the electronic properties of Si nanocrystals as a function of particle size. Phys Rev Lett 1998;80:3803. [379] Yoffe AD. Semiconductor quantum dots and related systems: electronic, optical, luminescence and related properties of low dimensional systems. Adv Phys 2001;50(1):1–208. Yoffe AD. Low-dimensional systems: quantum size effects and electronic properties of semiconductor microcrystallites (zero-dimensional systems) and some quasi-two-dimensional systems. Adv Phys 2002;51(2):799–890. [380] Tan OK, Zhu W, Yan Q, Kong LB. Size effect and gas sensing characteristics of nanocrystalline xSnO2(1
x)a-Fe2O3 ethanol sensors. Sensors and Actuators B 2000;65:361–5. [381] Rao CNR, Kulkarni GU, Thomas PJ, Edwards PP. Size-dependent chemistry: properties of nanocrystals. Chem-European J 2002;8(1):29–35. [382] Sun CQ, Gong HQ, Hing P, Ye H T. Behind the quantum-confinement and surface passivation of nanoclusters. Surf Rev Lett 1999;6(2):L171–6. [383] Sun CQ, Chen TP, Tay BK, Li S, Huang H, Zhang YB, et al. An extended. quantum confinement theory: surface coordination imperfection modifies the entire band structure of a nano-solid. J Phys D 2001;34(24):3470–9. [384] Sun CQ, Tay BK, Lau SP, Sun XW. Dielectric suppression and its effect on photon absorbance of nanometric semiconductors. J Phys D 2001;34:2359–62.

680

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[385] Sun CQ, Sun XW, Gong HQ, Huang H, Ye H, Jin D, et al. Frequency shift in the photoluminescence of nanostructured SiOx: surface-bond contraction and oxidation. J Phys Condens Matter 1999;11:L547–50. [386] Bahn SR, Jaconsen KW. Chain formation of metal atoms. Phys Rev Lett 2001;87:266202. [387] Reddy DR, Reddy BK. Laser-like mechanoluminescence in ZnMnTe-diluted magnetic semiconductor. Appl Phys Lett 2002;81(3):460–2. [388] Zhang JY, Wang XY, Xiao M. Lattice contraction in free-standing CaSer nanocrystals. Appl Phys Lett 2002;81(11):2076. [389] Yu XF, Liu X, Zhang K, Hu ZQ. The lattice contraction of nanometer-sized Sn and Bi particles produced by an electrohydrodynamic technique. J Phys Condens Matter 1999;11:937–44. [390] Stoneham AM. Comment on: the lattice contraction of nanometersized Sn and Bi particles. J Phys Condens Matter 1999;11:8351–2. [391] Nanda KK, Behera SN, Sahu SN. Comment on: the lattice contraction of nanometer sized Sn and Bi particles. J Phys Condens Matter 2001;13:2861–4. [392] Sun CQ. Comment on: the lattice contraction of nanometersized Sn and Bi particles. J Phys Cond Matt 1999;11(24):4801–3. [393] Kara A, Rahman TS. Vibrational properties of metallic nanocrystals. Phys Rev Lett 1998;81(7):1453. [394] Montano PA, Shenoy GK, Alp EE, Schulze W, Urban J. Structure of copper clusters isolated in solid argon. Phys Rev Lett 1996;56(19):207602079. [395] Ibach H. The role of surface stressing reconstruction, epitaxial growth and stabilization of mesoscopic structures. Surf Sci Rep 1997;29:193–263. [396] Haiss W. Surface stress of clean and adsorbate-covered solids. Rep Prog Phys 2001;64:591–648. [397] J. Robertson, 1991?? [398] Shi X, Tay BK, Flynn DL, Sun Z. Tribological properties of tetrahedral carbon films deposited by filtered cathodic vacuum arc technique. Mat Res Symp Proc 1997;436:293. [399] Sun CQ, Tay BK, Zeng XT, Li S, Chen TP, Zhou J, et al. Bond-order–bond-length–bond-strength correlation mechanism for the shape and size dependency of a nanosolid. J Phys Condens Matter 2002;14(34):7781–95. [400] Wong EW, Sheehan PE, Lieber CM. Nanobeam mechanics: elasticity, strength, and toughness of nanorods and nanotubes. Science 1997;277:1971–4. [401] Zeng XT. TiN/NbN superlattice hard coatings deposited by unbalanced magnetron sputtering. Surf Coat Technol 1999;113:75–9. [402] Zeng XT, Zhang S, Sun CQ, Liu YC. Nanometric-layered CrN/TiN thin films: mechanical strength and thermal stability. Thin Solid Films 2003;424(1):99–102. [403] Qian X, Hu¨bner W. First-principles calculation of structural and magnetic properties for Fe monolayers and bilayers on W(110). Phys Rev B 1999;60(23):16192–7. [404] Geng WT, Freeman AJ, Wu RQ. Magnetism at high-index transition-metal surfaces and the effect of metalloid impurities: Ni(210). Phys Rev B 2001;63:064427. [405] Geng WT, Kim M, Freeman AJ. Multilayer relaxation and magnetism of a high-index transition metal surface: Fe(310). Phys Rev B 2001;63:245401. [406] Sklyadneva Yu I, Rusina GG, Chulkov EV. Vibrational states on vicinal surfaces of Al, Ag, Cu and Pd. Surf Sci 1998;416:17–36. [407] Kara A, Durukanoglu S, Rahman TS. Local thermodynamic properties of a stepped metal surface: Cu(711). Phys Rev B 1996;53(23):15489–92. [408] Huang H, Sun CQ, Zhang TS, Hing P. Grain-size effect on ferroelectric Pb(Zr1-xTix)O3 solid solutions induced by surface-bond contraction. Phys Rev B 2001;63:184112. [409] Huang H, Sun CQ, Hing P. Surface-bond contraction and its effect on the behavior of nanometric PZT. J Phys Condens Matter 2000;12:L127–32. [410] Gleiter H, Weissmu¨ller J, Wollersheim O, Wu¨rschum R. Nanocrystalline materials: a way to solids with tunable electronic structures and properties? Acta Mater 2001;498:737–45. [411] Zhou J, Sun CQ, Pita K, Lam YL, Zhou Y, Ng SL, et al. Thermally tuning of the photonic bandgap of SiO2 colloid-crystal infilled with ferroelectric BaTiO3. Appl Phys Lett 2001;78(5):661–3.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

681

[412] Sun CQ, Wang Y, Tay BK, Li S, Huang H, Zhang Y. Correlation between the melting of a nanosolid and the cohesive energy of a surface atom. J Phys Chem B 2002;106:10701. [413] Zhong WH, Sun CQ, Tay BK, Li S, Bai HL, Jiang EY. Curie temperature suppression of ferromagnetic nanosolids. J Phys Condens Matter 2002;14:L399–405. [414] Shi FG. Size dependent thermal vibrations and melting in nanocrystals. J Mater Res 1994;9:1307. [415] Jiang Q, Zhang Z, Li JC. Superheating of nanocrystals embedded in matrix. Chem Phys Lett 2000; 322(6):549–52. [416] Jiang Q, Liang LH, Li JC. Thermodynamic superheating and relevant interface stability of lowdimensional metallic crystals. J Phys Condens Matter 2001;13(4):565–71. [417] Jiang Q, Aya N, Shi FG. Nanotube size-dependent melting of single crystals in carbon nano-tubes. Appl Phys A 1997;64:627–9. [418] Jiang Q, Shi HX, Zhao M. Melting thermodynamics of organic nanocrystals. J Chem Phys 1999; 111:2176–80. [419] Jiang Q, Shi HX, Li JC. Finite size effect on glass transition temperatures. Thin Solid Films 1999; 354:283–6. [420] Wen Z, Zhao M, Jiang Q. The melting temperature of molecular nanocrystals at the lower bound of the mesoscopic size range. J Phys Condens Matter 2000;12:8819–24. [421] Jiang Q, Liang LH, Zhao M. Modeling of the melting temperature of nano-ice in MCM-41 pores. J Phys Cond Mat 2001;13:L397–401. [422] Zhang Z, Zhao M, Jiang Q. Melting temperature of semiconductor nanocrystals in the mesoscopic size range. Semicond Sci Technol 2001;16:L33–5. [423] Sun CQ, Pan LK, Fu YQ, Sun XW, Li S., Tay BK. Communicated. The 3d5/2 and 2p3/2 levels of a Cu atom isolated from solid and their shifts upon nanosolid formation andf surface passivation. [424] Sun CQ, Tay BK, Fu YQ, Li S, Chen TP, Bai HL, Jiang EY. Discriminating crystal binding from atomic trapping of a core electron at energy levels shifted by surface relaxation or nanosolid formation. J Phys Chem B Lett 2003;107(2):411–4. [425] Sun CQ, Pan LK, Fu YQ, Tay BK, Li S. Size dependence of the 2p-level shift of nanosolid silicon. J Phys Chem B 2003;107(22):5113–5. [426] An B, Fukuyama S, Yokogawa K, Yoshimura M. Surface superstructure of carbon nanotubes on highly oriented pyrolytic graphite annealed at elevated temperatures. Jap J Appl Phys (part 1) 1998; 37:3809–11. [427] Yakobson BI, Brabec CJ, Bernholic J. Nanomechanics of carbon tubes: instabilities beyond linear response. Phys Rev Lett 1996;76:2511–4. [428] Falvo MR, Clary GJ, Taylor RM, Chi V, Brooks FP, Washburn S, Superfine R. Bending and buckling of carbon nanotubes under large strain. Nature 1997;389:582–4. [429] Sun CQ, Bai HL, Tay BK, Jiang EY. Communicated. The dimension and strength of a single C–C bond in carbon nanotubes. [Submitted to J Phys Chem B]. [430] Street RA. Hydrogenated amorphous silicon. Cambridge: Cambridge University Press; 1991 [p 276, 315, 295]. [431] Jin D, Hing P, Sun CQ. Growth kinetics and electrical properties of Pb(Zr0.9Ti0.1)O3 doped with Cerium oxide. J Phys D: Appl Phys 2000;33:744–52. [432] Sun CQ, Jin D, Zhou J, Li S, Tay BK, Lau SP, et al. Intensive and stable blue light emission of Pb(ZrxTi1-x)O3. Appl Phys Lett 2001;79(8):1082–4. [433] Hirata GA, McKittrick J, Devlin D. Growth and analysis of red, green and blue luminescent oxide thin films. Surf Rev Lett 1998;5(1):413–7. [434] Kudo A, Yanagi H, Hosono H, Kawazoe H. SrCu2O2: a p-type conductive oxide with wide bandgap. Appl Phys Lett 1998;73(2):220–2. [435] Jiang YD, Zhang F, Summers CJ, Wang ZL. Synthesis and properties of Sr2CeO4 blue emission powder phosphor for field emission displays. Appl Phys Lett 1999;74(12):1677–8. [436] Price KJ, Sharpe LR, McNeil LE, Irene EA. Electroluminescence in silicon oxynitride films. J Appl Phys 1999;86(5):2638–41. [437] Wang YQ, Zhao TP, Liu J, Qin GG. Near-ultraviolet and near-infrared electroluminescence from an indium-tin-oxide film/native Si oxide/p-Si structure. Appl Phys Lett 1999;74(25):3815–7.

682

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[438] Yuan S, Jagadish C, Kim Y, Chang Y, Tan H, Cohen RM, et al. Anodic-oxide-induced intermixing in GaAs-AlGaAs Quantum-well and quantum wire structures. IEEE J Select Topics in Quantum Elelctronics 1998;4:629–35. [439] Xu CX, Sun XW, Chen BJ, Sun CQ, Tay BK, Li S. Photoemission of aligned ZnO nanofiber array prepared by vapor transport in air. Chinese Phys Lett 2003;20(8):1319–22. [440] Brongersma ML, Polman A, Min KS, Boer E, Tambo T, Atwater HA. Tuning the emission wavelength of Si nanocrystals in SiO2 by oxidation. Appl Phys Lett 1998;72(20):2577–9. [441] Liu A. Microstructure and photoluminescence spectra of porous InP. Nanotechnology 2001;12:L1–3. [442] Ponce FA, Bour DP. Nitride-based semiconductors for blue and green light-emitting devices. Nature (London) 1997;386(6623):351–9. [443] Chambouleyron I, Zanatta AR. Nitrogen in germanium. J Appl Phys 1998;84:1–30. [444] Hammer P, Victoria NM, Alvarez F. Electronic structure of hydrogenated carbon nitride films. J Vac Sci Technol A 1998;16(5):2941–9. [445] Reynolds DC, Look DC, Jogai B, Morkoc H. Similarities in the band edge and deep-center photoluminescence mechanisms of ZnO and GaN. Solid State Commun 1997;101:643. [446] Austin IG, Jackson WA, Searle TM, Bhat PK, Gibson RA. Philos Mag 1985;B52:271. [447] Zanatta AR, Chambouleyron I. Nitrogen in the amorphous-germanium network: from high dilution to the alloy phase. Phys Rev 1993;B48:4560–70. [448] Vilcarromero J, Marques FC. Influence of the deposition conditions on the properties of amorphous germanium nitrogen alloys. Phys Status Solidi B 1995;192(2):543–7. [449] Hasegawa S, Matsuda M, Kurata Y. Bonding configuration and defects in amorphous SiNx:H films. Appl Phys Lett 1991;58:741–3. [450] Theib W. Optical properties of porous silicon. Surf Sci Rep 1997;29:91–192. [451] Collins RT, Fauchet PM, Tischler MA. Porous silicon: from luminescence to LEDs. Phys Today 1997;1:24–31. [452] Canham LT. Silicon quantum wire array fabrication by electrochemical and chemical dissolution of wafers. Appl Phys Lett 1990;57:1046–8. [453] Glinka YD, Lin S-H, Hwang L-P, Chen Y-T, Tolk NH. Size effect in self-trapped exciton photoluminescence from SiO2-based nanoscale materials. Phys Rev B 2001;64:085421. [454] Qin GG, Jia YQ. Mechanism of the visible luminescence in porous silicon. Solid State Commun 1993;86:559–63. [455] Koch F, Petrova-Koch V, Muschik T, Nikolov A, Gavrilenko V. Microcrystalline Semiconductors: Materials Science and Devices vol. 283. Pittsburgh, PA: Materials Research Society; 1993. p. 197. [456] Iwayama TS, Hole DE, Boyd IW. Mechanism of photoluminescence of Si nanocrystals in SiOx fabricated by ion implantation: the role of interactions of nanocrystals and oxygen. J Phys Condens Matter 1999;11:6595–604. [457] Brus JE. On the development of bulk optical properties in small semiconductor crystallites. J Lumin 1984;31:381. [458] Kayanuma Y. Quantum-size effects of interacting electrons and holes in semiconductor microcrystals with spherical shape. Phys Rev B 1988;38:9797. [459] Steigerwald ML, Brus LE. Acc Chem Res 1990;23:183. [460] Luo X, Zhang SB, Wei S-H. Chemical design of direct-gap light-emitting silicon. Phys Rev Lett 2002;89:076802. [461] Sun CQ, Li S, Tay BK, Chen TP. Upper limit of blue shift in the photoluminescence of CdS and CdSe nanosolids. Acta Materilia 2002;50(8):4687–93. [462] Borgohain K, Singh JB, Rao MVR, Shripathi T, Mahamuni S. Quantum size effect in CuO nanoparticles. Phys Rev B 2000;61(16):11093–6. [463] Schmeiber D, Bo¨hme O, Yafantis A, Heller T, Batcherlor DR, Lundstrom I, et al. Dipole mement of nanparticles at interfaces. Phys Rev Lett 1999;83(2):380–3. [464] Nanda J, Sarma DD. Photoemission spectroscopy of size selected zinc sulfide nanocrystallites. J Appl Phys 2001;90:2504. [465] Nanda J, Kuruvilla A, Sarma DD. Photoelectron spectroscopic study of CdS nanocrystallites. Phys Rev B 1999;59:7473.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

683

[466] Balasubramanian T, Andersen JN, Wallde´n L. Surface-bulk core-level splitting in graphite. Phys Rev 2001;B64:205420. [467] Sun CQ, Bai HL, Tay BK, Jiang EY. Length, strength, extensibility and thermal stability of metallic bond in a gold monatomic chain. J Phys Chem B [in press]. [468] Wang X, Huang DM, Ye L, Yang M, Hao PH, Fu HX, Hou XY, Xie XD. Pinning of photoluminescence peak positions for light-emitting porous silicon: An evidence of quantum size effect. Phys Rev Lett 1993;71:1265–7. [469] Delerue C, Lannoo M, Allan G. Properties of porous silicon. London: INSPEC; 1997. p 213. [470] Dorigoni L, Bisi O, Bernardini F, Ossicini S. Electron states and luminescence transition in porous silicon. Phys Rev B 1996;53:4557–64. [471] Sanders GD, Chang Y-C. Theory of optical properties of quantum wires in porous silicon. Phys Rev B 1992;45:9202–13. [472] Hybertsen MS, Needels M. First-principles analysis of electronic states in silicon nanoscale quantum wires. Phys Rev B 1993;48:4608–11. [473] Kanemitsu Y, Uto H, Masumoto Y, Matsumoto T, Futagi T, Mimura H. Microstructure and optical properties of free-standing porous silicon films: size dependence of absorption spectra in Si nanometer-sized crystallites. Phys Rev B 1993;48:2827–30. [474] Schuppler S, Friedman SL, Marcus MA, Adler DL, Xie YH, Ross FM, et al. Size, shape, and composition of luminescent species in oxidized Si and H-passivated Si. Phys Rev 1995;B52:4910. [475] Ohno T, Shiraishi K, Ogawa T. Intrinsic origin of visible light emission from silicon quantum wires: Electronic structure and geometrically restricted exciton. Phys Rev Lett 1992;69:2400–3. [476] Yeh CY, Zhang SB, Zunger A. Confinement, surface, and chemisorption effects on the optical properties of Si quantum wires. Phys Rev 1994;40:14405–15. [477] Read AJ, Needs RJ, Nash KJ, Canham LT, Calcott PDJ, Qteish A. First-principles calculations of the electronic properties of silicon quantum wires. Phys Rev Lett 1992;69:1232–5. [478] Polatoglou HM. J Lumin 1993;57:117–20. [479] Pavesi L, Giebel G, Ziglio F, Mariotto G, Priolo F, Campisano FU, Spinella C. Nanocrystal size modifications in porous silicon by preanodization ion implantation. Appl Phys Lett 1994;65(17):2182–4. [480] Pan LK, Sun CQ, Tay BK, Chen TP, Li S. Photoluminescence of Si nanosolid near the lower end of size limit. J Phys Chem B 2002;106:11725. [481] Zhu W, Kochanski GP, Jin S. Low-field electron emission from undoped nanostructured diamond. Science 1998;282:1471–3. [482] Geis MW, Efremow NN, Krohn KE, Twitchell JC, Lyszczarz TM, Kalish R, et al. A new surface electron-emission mechanism in diamond cathodes. Nature (London) 1998;393:431–5. [483] Xu WS, Tzeng T, Latham RV. Similarities in the ‘cold’ electron emission characteristics of diamond coated molybdenum electrodes and polished bulk graphite surfaces. J Phys D 1993;26:1776–80. [484] Huang ZH, Cutler PH, Miskovsky NM, Sullivan TE. Calculation of local-density of states at an atomically sharp Si tip. J Vac Sci Technol B 1995;13:522–4. [485] Amaratunga GAJ, Silva SRP. Nitrogen containing hydrogenated amorphous carbon for thin-film field emission cathodes. Appl Phys Lett 1996;68:2529. [486] Okano K, Koizumi S, Silva SRP, Amaratunga GAJ. Low-threshold cold cathodes made of nitrogen-doped chemical-vapour-deposited diamond. Nature 1996;381:140. [487] Cheah LK, Shi X, Liu E, Tay BK. Electron field emission properties of tetrahedral amorphous carbon films. J Appl Phys 1999;85(9):6816–21. [488] Lim SC, Stallcup II RE, Akwani IA, Perez JM. Effects of O2, H2, and N2 gases on the field emission properties of diamond-coated microtips. Appl Phys Lett 1999;75(8):1179–81. [489] Sowers AT, Ward BL, English SL, Nemanich RJ. Field emission properties of nitrogen-doped diamond films. J Appl Phys 1999;86(7):3973–82. [490] Sugino T, Yamamoto T, Kimura C, Murakami H, Hirakawa M. Field emission characteristics of carbon nanofiber improved by deposition of boron nitride nanocrystalline film. Appl Phys Lett 2002;80(20):3808–10. [491] Sugino T, Kimura C, Yamamoto T. Electron field emission from boron-nitride nanofilms. Appl Phys Lett 2002;80(19):3602–4.

684

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

[492] Herna´ndez E, Goze C, Bernier P, Rubio A. Elastic Properties of C and BxCyNz composite nanotubes. Phys Rev Lett 1998;80(20):4502. [493] Riege H, Boscolo I, Handerek J, Herleb U. Features and technology of ferroelectric electron emission. J Appl Phys 1988;84(3):1602–13. [494] Stolichnov I, Tagantsev AK, Colla EL, Setter N. Cold-field-emission test of the fatigued state of Pb(ZrxTi1
x)O3 films. Appl Phys Lett 1998;73(10):1361–3. [495] Pickett WE. Negative electron affinity and low work function surface: cesium on oxygenated diamond (100). Phys Rev Lett 1994;73(12):1664–7. [496] Park N, Han S, Ihm J. Effects of oxygen adsorption on carbon nanotube field emitters. Phys Rev 2001;B64:125401. [497] Kung S-C, Hwang KC, Lin IN. Oxygen and ozone oxidation-enhanced field emission of carbon nanotubes. Appl Phys Lett 2002;80(25):4819–21. [498] Wadhawan A, Stallcup II REM, Stephens II KF, Perez JM, Akwani IA. Effects of O2, Ar, and H2 gases on the field-emission properties of single-walled and multiwalled carbon nanotubes. Appl Phys Lett 2001;79(12):1867–9. [499] Briddon P, Jones R. Theory of impurities in diamond. Physica B 1993;185:179–89. [500] Robertson J. Mechanisms of electron field emission from diamond, diamond-like carbon, and nanostructured carbon. J Vac Sci Technol B 1999;17(2):659–65. [501] Rutter MJ, Robertson J. Ab initio calculation of electron affinities of diamond surfaces. Phys Rev B 1998;57:9241–5. [502] Geis MW, Twichell JC, Macaulay J, Okano K. Electron field emission from diamond and other carbon materials after H2, O2, and Cs treatment. Appl Phys Lett 1995;67:1328–30. [503] Bo¨er K. Survey of semiconductor physics vol. II. New York: Van Nostrand Reinhold; 1992. p. 359. [504] Jiang EY, Sun CQ, Liu Y, Li J. The structures and magnetic properties of facing targets sputtered Fe-nitride films. J Appl Phys 1989;65:1659–63. [505] Schaaf P. Laser nitriding of metals. Prog Mater Sci 2002;47(1):1–161. [506] Coey JMD, Sun HJ. Mag Mag Mater 1990;87:L251. [507] Pan H, Yang F, Chen Y, Han X, Tang N, Chen C, Wang Q. Magnetic properties of a new series of rare-earth iron nitrides R3(Fe, Mo)29Nx (R=Ce, Nd, Sm, Gd, Tb, Dy or Y) J. Phys Condens Matt 1997;11:2499–505. [508] Johnson CE, Bennett JM, Nadler MP. Oxidation of diamond windows. J Mater Res 1995;10(10): 2555–63. [509] Joshi A, Nimmagadda R, Herrington J. Oxidation kinetics of diamond, graphite, and chemical vapor deposited diamond films by thermal gravimetry. J Vac Sci Technol 1990;A8:2137. [510] Zhu W, Wang XH, Pickrell DJ, Badzian AR, Messier R. The oxidation of CVD diamond films. Carbon 1990;28(6):796–9. [511] Bobrov K, Shechter H, Hoffman A, Folman M. Molecular oxygen adsorption and desorption from single crystal diamond (1 1 1) and (1 1 0) surfaces. Appl Surf Sci 2002;196(1–4): 173–80. [512] Li J, Zhang Q, Yoon SF, Ahn J, Zhou Q, Wang S, Yang D, Wang Q, Li Z, Wang J, Lei Q. Reaction of diamond thin films with atomic oxygen simulated as low earth orbit environment. J Appl Phys 2002;92(10):6275–7. [513] Theije FKde, van Veenendaal E, van Enckevort WJP, Vlieg E. Oxidative etching of cleaved synthetic diamond {111} surfaces. Surf Sci 2001;492(1–2):91–105. [514] Pehrsson PE, Mercer TW, Chaney JA. Thermal oxidation of the hydrogenated diamond (100) surface. Surf Sci 2002;497(1–3):13–28. [515] Hartman P, Perdok WG, On the relations between structure and morphology of crystals. I. Acta Crys 1955;8(9):49–52; On the relations between structure and morphology of crystals. II. Acta Crys 1955;8(9):52–4; On the relations between structure and morphology of crystals. III. Acta Crys 1955; 8(9):525–9. [516] Hartman P. Non-uniform distribution of faces in a zone. Z Kristallogr 1965;121:78. [517] Kim YK, Lee KY, Lee JY. Texture-controlled diamond films synthesized by microwave plasmaenhanced chemical vapor deposition. Thin Solid Films 1996;272(1):64–70.

C.Q. Sun / Progress in Materials Science 48 (2003) 521–685

685

[518] Angus JC, ArgoitiaA, Gat R, Li Z, Sunkara M, Wang L, Wang Y. In: Lettington A, Steeds JW, editors. The thin film diamond. London: Chapman and Hall; 1994. p. 9. [519] Loh KP. Private communication. May, 2001. [520] Fu YQ, Yan BB, Sun CQ, Du HJ. Carbon turns the tensile surface stress of Ti into compressive. J Phys D 2001;34(24):L129–132. [521] Sun CQ, Fu YQ, Yan BB, Lau SP, Sun XW, Tay BK. Improving diamond–metal adhesion with graded TiCN interlayers. J Appl Phys 2002;91(4):2051–4. [522] Fu YQ, Sun CQ, Du HJ, Yan BB. From diamond to crystalline silicon carbonitride: effect of introduction of nitrogen in CH4/H2 gas mixture using MW-PECVD 2003;424(1):107–14. [523] Tsu R, Babic D, Loriatti Jr L. Simple model for the dielectric constant of nanoscale silicon particle. J Appl Phys 1997;82:1327. [524] Ye HT, Sun CQ, Huang HT, Hing P. Dielectric transition of nanostructured diamond films. Appl Phys Lett 2001;78(13):1826–8. [525] Ye HT, Sun CQ, Hing P. Size control and size effect on the dielectric constant of nanostructured diamond films. J Appl Phys D 2000;33(23):L148–152. [526] Lundgren E, Kresse G, Klein C, Borg M, Andersen JN, De Santis M, et al. Two-dimensional oxide on Pd(111). Phys Rev Lett 2002;88:246103. [527] Lefebvre I, Szymanski MA, Olivier-Fourcade J, Jumas JC. Electronic structure of tin monochalcogenides from SnO to SnTe. Phys Rev 1998;B58(4):1896–906.