Tailoring oxide properties : an impact on adsorption characteristics of molecules and metals

Both density functional theory calculations and numerous experimental studies demonstrate a variety of unique features in metal supported oxide films and transition metal doped simple oxides, which are markedly different from their unmodified counterparts. This review highlights, from the computational perspective, recent literature on the properties of the above mentioned surfaces and how they adsorb and activate different species, support metal aggregates, and even catalyse reactions. The adsorption of Au atoms and clusters on metal-supported MgO films are reviewed together with the cluster’s theoretically predicted ability to activate and dissociate O2 at the AuMgO(100)/Ag(100) interface, as well as the impact of an interface vacancy to the binding of an Au atom. In contrast to a bulk MgO surface, an Au atom binds strongly on a metal-supported ultra-thin MgO film and becomes negatively charged. Similarly, Au clusters bind strongly on a supported MgO(100) film and are negatively charged favouring 2D planar structures. The adsorption of other metal atoms is briefly considered and compared to that of Au. Existing computational literature of adsorption and reactivity of simple molecules including O2, CO, NO2, and H2O on mainly metal-supported MgO(100) films is discussed. Chemical reactions such as CO oxidation and O2 dissociation are discussed on the bare thin MgO film and on selected Au clusters supported on MgO(100)/metal surfaces. The Au atoms at the perimeter of the cluster are responsible for catalytic activity and calculations predict that they facilitate dissociative adsorption of oxygen even at ambient conditions. The interaction of H2O with a flat and stepped Ag-supported MgO film is summarized and compared to bulk MgO. The computational results highlight spontaneous dissociation on MgO steps. Furthermore, the impact of water coverage on adsorption and dissociation is addressed. The modifications, such as oxygen vacancies and dopants, at the oxide-metal interface and their effect on the adsorption characteristics of water and Au are summarized. Finally, more limited computational literature on transition metal (TM) doped CaO(100) and MgO(100) surfaces is presented. Again, Au is used as a probe species. Similar to metal-supported MgO films, Au binds more strongly than on undoped CaO(100) and becomes negatively charged. The discussion focuses on rationalization of Au adsorption with the help of Born-Haber cycle, which reveals that the so called redox energy including the electron transfer from the dopant to the Au atom together with the simultaneous structural relaxation of lattice atoms is responsible for enhanced binding. In addition, adsorption energy dependence on the position and type of the dopant is summarized.


I. INTRODUCTION
Metal oxides have long been considered as potential materials for large variety of applications ranging from gas sensing and protective coatings to electrodes in fuel cells, heterogeneous photo(electro) catalyst, and bio-compatible materials 1 . Compared to any other material class both crystallographic and electronic properties of oxides display diverse behaviour, e.g., electronic conductivity ranges from wide-gap insulators to materials with conductivity comparable to metals 2 . The characteristics of oxides can be tailored to improve desired properties in various ways by introducing structural modifications like steps and grain boundaries, adding impurity atoms as dopants, or removing atoms from the structure 1,2 . In particular, point defects such as oxygen vacancies determine optical, electronic and transport properties of insulating oxides, and they usually dominate the chemistry of its surface 3 . While transition metal oxides are utilized for their catalytic properties, simple oxides such as MgO or CaO are intrinsically inert owing to their very deep valence band and very high conduction band; thus, they are less exploited in applications. However, simple oxides are interesting model systems whose properties have been thoroughly investigated 1,4 and therefore they form an ideal platform to explore the impact of different tailoring strategies to improve their reactivity. One way to achieve this is to prepare oxides as metal grown thin films, which provides an unique approach to modify structural, electronic and chemical properties as a function of film thickness extensively, which is discussed in the reviews of Prof. H.J. Freund, Prof. G. Pacchioni, and Prof. N. Nilius [5][6][7][8] . From the experimental point of view metal supported thin-film systems create specific technical challenges to be tackled. The significant benefit of ultra-thin arrangement of insulating oxides is that they can be studied with scanning tunneling microscopy (STM), which is not possible for their bulk counterparts.
Ultra-thin oxide films such as MgO 6,9 , NiO 10 , CaO 11 , Al 2 O 3 12,13 , FeO 14-16 , and SiO 2 17 have been extensively studied. A comprehensive overview of metal-supported transition metal oxide films can be found in reference 18 . Possible applications of metal supported ultra-thin oxide films can be divided into two groups: support materials and active players in chemical conversions. Gold clusters on metal supported ultra-thin films show distinct features compared to clusters on bulk films 5 and calculations predict that the perimeter of these particles is highly reactive e.g., activating oxygen readily 19 . Supported films can also directly act as a catalyst for CO oxidation 20,21 and dissociate H 2 22 . Furthermore, metal-supported ultra-thin 4 films are reactive towards water dissociation 23 .
Other ultra-thin insulating materials grown on metal surfaces include e.g., NaCl on Cu (111), which is used as a substrate to explore charge transport to nanostructures including Au adatoms [24][25][26] . While the stoichiometry and atomic structure of ultra-thin MgO corresponds that of the bulk MgO, this is not always the case. The most prominent example of this is an ultra-thin alumina film over a NiAl support for which determination of atomic structure turned out to be particularly challenging owing to a complex atomic structure.
From the interplay between STM experiments and density functional theory (DFT) calculations, the peculiar alumina structure was revealed and it corresponds to Al 10  However, dopant atoms can have either lower or higher valence compared to a cation they substitute and this dictates their impact on host oxide characteristics. In the latter case, the 5 influence of a dopant resembles that of a metal support in the ultra-thin oxide film. However, the number of spare electrons depends on the concentration of dopants and the fact that the dopants can be mobile, unlike the metal support. There is a large body of studies, where the impact of substutional doping on catalytic properties of oxides has been examined for example see references in 32,33 . In many cases it has been shown that the reactivity has improved upon doping. Among the studied systems Li-doped MgO has received the most attention owing to its use as a catalyst for oxidative methane coupling to ethane that is often labeled as Holy Grail in catalysis [34][35][36] . Despite extensive research on this topic, many questions have remained open, and there is not a conclusive understanding of the role of the catalyst, let alone its structure. In the case of CaO, STM studies demonstrate that tiny amounts of Mo embedded into the oxide introduce similar features as metal-supported ultra-thin MgO films namely, Au atoms become negatively charged 37,38 and Au aggregates favour the 2D growth mode 39 . Calculations predict enhanced binding of Au adatoms, which is due to electron transfer from Mo to the adsorbate and substantial lattice relaxations near the dopant originating from increased attraction between more positively charged Mo and surrounding anions 40 . Notwithstanding the similarities, TM-doped simple oxides and metal-supported ultra-thin oxide systems show fundamental differences as impact of dopant is more localised compared to metal support, and a number of electrons that the dopant can provide is very limited, which means that the dopant concentration is a new adjustable parameter.

II. PREPARATION AND PROPERTIES OF MODIFIED SIMPLE OXIDES
Here the preparation of oxide systems is briefly discussed and for a more detailed discussion on the preparation and experimental studies of ultra-thin films and simple doped oxides refer to previous reviews 5, 6,8 . Ultra-thin-oxide layers are synthesized by the vaporization of metals such as Mg in a background of molecular oxygen on a metal support and they typically have smaller optimal lattice parameters than bulk oxides 41 . By far the most extensively studied ultra-thin oxide film is MgO for which the best support is either a Ag (100) or Mo (100) surface owing to small lattice mismatch, which amounts to 3% and 5.4 %, for Ag and Mo respectively. In the thin film structure O anions are preferably located above Ag/Mo atoms 42 as displayed in Figure 1. Already a 3ML-thick MgO film on a Ag-support is sufficient to reproduce the band gap of bulk MgO 9 .
Crucial to the characteristics of metal-supported ultra-thin oxide materials is the interaction between a metal substrate and an oxide film, which affects both the geometric and electronic properties of the system. Furthermore, even a small lattice mismatch, like the In the pioneering work of Haruta and Hutchings, it was found that Au nanoparticles can catalyze CO oxidation to CO 2 at low temperature 75,77 and acetylene hydrochlorination 75,78,79 .
Later, Au nanoparticles were found to be active in reactions such as propylene epoxidation 80 ,  Calculations from Honkala and Häkkinen show that Au adsorption energy depends sensitively on film thickness being strongest on the thinnest film, Figure 2 with the exception that a 2ML-thick MgO film binds Au more strongly than a 1ML-thick film 48     image appears because tunneling is not in resonance with any cluster states. Adjusting the bias to appropriate values, flower-like shapes appear and the increase in cluster height is seen. Both effects suggest that the STM contrast is dictated by electronic structure of a cluster whose exact atomic structure remains unknown. Theoretically, the electronic structure of a planar cluster can be predicted employing the 2D harmonic oscillator potential model, which confines the atomic 6s valence electrons 118 . The eigenstates of the 2D harmonic oscillator (HO) can be characterised by their principal quantum number together with the projection of angular momentum on the normal vector of the plane, m. Thus, the states have n-1 radial nodes and |m| angular nodes. The atomic nomenclature of the states is adopted, which means that e.g., the 2P state has one radial node and two angular nodes.
Moreover, states such as 1F exhibit a very flower-like appearance. Again, each Au atom brings in one 6s electron, therefore the number of atoms in a cluster can be obtained by analyzing the nodal structure of frontier orbitals in a cluster within the 2D HO model. To resolve the atomic structure of these clusters, an extensive DFT search for possible structures was carried out at size regime of 6-20 Au atoms per a cluster. The aim was to find clusters having similar HOMO and LUMO states as seen in STM images of unknown species. The candidate cluster sizes were selected so that the 2D HO model predicts that their HOMO and LUMO states resemble those seen in experiments. Furthermore, since flower-like shapes are typical for symmetric planar clusters the computational search focused on these structures. Similar to 1D Au chains Au clusters might have additional electrons not originating from 6s states. Figure 8 collects some of the selected cluster sizes and shapes calculated. All of them experience thin-film effects that is having a planar geometry with formation energies ranging from -2.1 --2.5 eV/atom. The charge analysis suggests that the smallest species are doubly charged, and the large ones can accommodate 3-4 electrons. Enhanced adsorption on metal-supported ultra-thin film is not limited to metal species.

Calculations assign that Au
The phenomenon is more general in nature and present also for molecules with high electronic affinity such as O 2 and NO 2 . Understanding NO 2 adsorption characteristics is important, for example for NO x storage catalysts to control emissions from combustion in oxygen excess 124 .
On bulk-like MgO NO 2 adsorbs weakly and stays neutral 125

23
Relatively little attention has been paid on chemical reactions on MgO due to its inertness compared to more reactive reducible oxides such as TiO 2 . However, as discussed above ultrathin films display higher chemical reactivity compared to bulk MgO binding species with high electron affinity more strongly. The question then arises: can thin films systems facilitate chemical reactions ? difference in kinetics of CO 2 production from CO+O 2 and CO+NO reactions indicate that the healing of oxygen vacancies is the rate-limiting step 21 . DFT calculations demonstrate that this is indeed the case because NO adsorbs N-down on an oxygen vacancy and CO reaction with the "dangling" oxygen is a highly activated process 21 .
The adsorption and activation of O 2 is known to be difficult on Au 131 , yet its pivotal role in oxidation reactions on an Au catalyst is undeniable. Therefore, it is natural that first reactivity studies on ultra-thin film supported Au clusters include molecular oxygen. O 2 adsorption and activation on thin-film supported Au clusters has been theoretically analyzed on  and is strongly linked to bridge the materials and pressure gaps to discover potential morphology changes of the catalyst particle under ambient conditions. The Au 14 cluster has an idealized C 2v symmetry with ten atoms at the edge and four inner atoms forming a rhombus-like structure. All the favourable adsorption sites reside at the edge of the cluster as the adsorption on the top of the cluster is endothermic. For all the studied structures adsorption energies range from from -1.15 eV for a single molecule to -0.6 eV for the 10'th molecule. The best adsorption structures for 1, 2, 6, and 10 oxygen molecules are given in    relative to the orientation as shown in Table I showed evidence of facile hydrolysis [155][156][157] , where the active site was proposed to be at the metal-oxide interface. This was further supported by the DFT study of the system 158 . Interestingly, calculations show that isolated H-bonded water dimers facilitate barrierless water dissociation on a bare and Ag-supported MgO(100) surface 161 . Dissociation is further stabilised on MgO ultra-thin films, which is attributed to two factors. The presence of metal support stabilizes the final state of charged fragments owing to structural relaxations in oxide that induce a strong corrugation of the film. Furthermore, based on the analysis of interaction energies the ion-ion interaction is suggested to change to a dipole-dipole type interaction ensuing a weakly bound ion pair 161 . At monolayer water coverages different structural motifs corresponding to surface cells with different sizes have been employed in calculations 148 .
An example applied structural model is given in Figure 18 The properties of metal supported ultra-thin oxide films can be tuned without varying the film thickness via engineering the interface between the metal support and the oxide. This can be achieved by introducing impurity atoms into the metal oxide interface and by varying the species, which in turn has been predicted to modify the electronic characteristics of the system and can be seen as changes in the work function [163][164][165] . Also, interfacial vacancies can be induced in the supported oxide films. In these systems kinetic effects have been adsorption and dissociation on Ag-supported MgO. In the latter two cases, an Ag atom in the topmost Ag layer is replaced with either an impurity atom or a dopant atom. The oxidemetal interface modifications change neither the reaction mechanism, nor the adsorption geometries of the water molecule, nor the dissociation products OH − and H + ; they remain similar to those on the ideal MgO(100)/Ag(100) surface discussed in chapter IV C. Instead, any kind of the interface defects introduce changes in energetics: smaller for the adsorption energy of water and larger for the transition and final states. The transition state strongly resembles the final state and actually describing it as a transition state is a controversial issue since the H-OH bond is clearly already broken at this state. Comparison of the influence of an interfacial oxygen vacancy and O/Mg impurity atoms shows that the vacancy introduces the largest stabilization effect: the barrier height drops 42 % compared to the ideal MgO/Ag system. Actually, the dissociation does not require external thermal energy atall to proceed, which means that even a small number of hidden vacancies is sufficient to enhance water dissociation substantially. Among the studied 3d TM dopants, the Ti atom is the most reactive one, and it stabilises the initial, transition and final states by 0. 17 including the plane average density analysis, the electronic localization function, and the Bader charge analysis, unravel that while one vacancy electron stays in a support metal atom under the vacancy, the second electron goes to a Au atom; thus the vacancy is more like a F ++ center. The simple thermodynamic analysis -with charge transfer and electrostatic terms-renders the formation of F ++ thermodynamically possible. According to the calculations, the electron transfer process is independent of the film thickness (at the range of 1-3 layers), and takes place similarly on both Ag and Mo support metals.
The fact that Au binding energy to the surface vacancy is always more exothermic than to the site with a buried vacancy, leads to the thermodynamic driving force to extract the vacancy from the metal-oxide interface to the surface. Despite thermodynamic preference, the vacancy extraction from the interface to the oxide surface can be hindered by kinetic factors, that is, too high diffusion barriers. The diffusion pathway of an oxygen vacancy consists of jumps between two nearest neighbour sites. In bulk MgO this has been reported 41 to be highly demanding with activation energy of ∼ 4.25 eV 170 . On two or three layers thick, unsupported MgO slabs the barrier is markedly reduced being still, however, ∼ 2.   Clearly, adsorption becomes weaker with the increasing distance between the Au and the dopant. The term ∆E coul demonstrates approximately 1/r behaviour, where r is the Audopant distance proving further support that energy change ∆E coul describes electrostatic interaction. Owing to strong screening by the polarized oxide, the Coulomb interaction is significant only at very short distances. In experiments the capping region is typically a few undoped CaO layers 37 . If the Mo is in the third layer the Au-Mo distance is approximately 8Å which corresponds the contribution of only ∼-0.5 eV while the adsorption energy is -3.0 eV, thus the electrostatic contribution to Au adsorption is minor on this oxide. The adsorption energy is dominated by the other two terms. The iono-covalent term is independent of a dopant as it describes the interaction between Au and ions in the oxide matrix but depends on the adsorption site. The ∆E redox term represents energy gain related to the electron exchange between Au and the dopant and corresponds to the dominant energy contribution in adsorption energy. It can be divided into two cases. Since the Au gains charge already upon the adsorption on a Ca-site of pristine CaO, the redox energy change is assigned to the system stabilization due to Mo oxidation as the more cationic dopant binds stronger to anionic O 2− . This is manifested as decreased distance between the dopant and the surrounding oxygens. Since this is a local effect, it does not depend on the depth of the dopant as seen in Figure 23. On an O-top site Au is initially neutral and thus the redox process includes charge transfer and stabilization due to structural relaxations around the dopant; again, the redox energy saturates quickly to a constant value. The redox energies differ by 1 eV such that the less exothermic value is found for the O-top site. This is due to Coulomb repulsion between negatively charged Au and negatively charged O, manifested also in a Au-O bond length, which increases ∼ 0.6Å upon Au charging. Figures 23 B) 34 . The reaction has not been thoroughly understood yet and for example, an active site for a homolytic CH 3 -H bond breaking step has remained elusive.
In the case of undervalent doped systems, the Born-Haber cycle analysis can also be applied to unravel the relevance of different energy terms for e.g., Au adsorption, which has been experimentally studied 39

49
Over the past years significant advances have been made in the preparation and experimental studies of tailored oxides as well as in the modeling their properties. These modified oxides have emerged as materials with unique properties, not encountered in their bulk counterparts. The selected examples discussed highlight the potential of simple oxides. They can be made chemically active and can operate as active catalysts or act as supports for catalyst particles. Two different schemes to modify the properties of simple oxides are addressed, namely growing oxides as ultra-thin films over metal supports and doping them with metal impurities. The discussion mainly focuses on theoretical studies for metal supported MgO films and in doped oxides on CaO and MgO systems, while other oxides are occasionally brought in attention for comparison. All the reviewed systems underline the potential and importance of density functional theory calculations and their role to unravel experimentally seen features but even more importantly to predict material properties, especially activity.
Despite the fact that computational methods have their limitations, for certain cases, such as buried vacancies, they are almost irreplaceable in addressing the characteristics and impact of hidden defects on adsorption.
The key adsorbates elaborated include electronegative species such as Au atoms and clusters, molecular oxygen, and NO 2 . The adsorption and dissociation of molecular oxygen is considered both on a bare metal-supported ultra-thin MgO and on an Au cluster over Ag-supported MgO. Interestingly, the calculations predict that the presence of the support metal facilitates simultaneous activation of several oxygen molecules. The adsorption and dissociation of non-electronegative water on ideal MgO/Ag is discussed both at low and high water coverages. In the latter case, the formation of strong hydrogen-hydrogen bonds leads to dissociation of a fraction of adsorbed water molecules. In addition, the modifications at the oxide-support interface are addressed in the two cases; on Au adsorption on a MgO/Ag(Mo) surface with an interfacial oxygen vacancy and water adsorption on MgO/Ag with an interfacial dopant. The role of thin-film systems is not limited to model systems but they have a variety of applications including catalysis, solid oxide fuel cells, gas sensors, corrosion protection, and biocompatible materials, just to mention a few. The spectrum and composition of oxide materials developed for different application is broad; some of these oxides are simple while there is an increasing number of fairly complex oxides. In the 50 future, we need to compute the electronic structure of these complex oxides reliably, which potentially have a variety of different defects, and simultaneously get a description for weak van der Waals interactions. In particular, calculation of line defects increases the system size, which in turn poses challenges for calculations.
On doped CaO and MgO oxides, the adsorption of an Au atom and an O 2 molecule is discussed. While the film thickness is an adjustable parameter in metal-supported thin film systems, in doped oxides the corresponding parameter is the dopant concentration.
Moreover, unlike the metal support, dopants can be mobile and they are also predicted to introduce vacancies into oxide. However, calculations indicate that the impact of over-valent dopants on electronegative adsorbates such as Au atoms and O 2 molecules, is similar to the metal support. Thus adsorption is enhanced, the species become negatively charged, and on both supports Au clusters favour planar geometries. With the help of the Born-Haber cycle the enhanced binding is attributed to energy gain owing to simultaneous electron transfer and lattice relaxations. In an idealized system, the ionization energy of the TM dopant explains the variation of Au adsorption energy from one dopant to the other but in reality dopant induced vacancies have an influence on electron transfer processes and modify this ideal rule. While in metal-supported ultra-thin oxide films systems one can gain better control over structural and electronic characteristics, the atomic structure of doped oxides displays larger plasticity and uncertainty. The model system studies of these materials make the basis to identify the key factors determining the function of the material, which paves the way for examination of less-defined oxides.
Doped oxides also have a variety of applications ranging from catalysis and photovoltaic to chemical sensing, solid oxide fuel cells, and coatings. In many cases the doped oxide is in the thin-film arrangement. The presence of dopants typically affects a number of point defects, such as oxygen vacancies, which then strongly impact on surface chemistry of an oxide. From the theoretical point of view, the concentration and distribution of dopants and vacancies substantially enhance the computational burden and add complexity. Although possible, special care must be taken when the Born-Haber cycle is applied to rationalize chemical reactions on complex, doped oxide surfaces. This is because sufficiently accurate description of the electronic structure is needed and because the impact of vacancies must be included to the analysis. Furthermore, the identification of an active site is usually more demanding than on metal surfaces and nanoparticles owing to a larger number of different 51 factors such as dopant and vacancy concentration, which might influence the nature and characteristics of the active site. One more thing that makes DFT calculations particularly challenging, is the lack of experimental methods, that could reveal the morphology and composition of oxides under reaction conditions. Therefore, there is an urgent need to develop computational methods and concepts to simulate oxide characteristics and establish their key descriptors to advance in one of the most challenging areas of Material Science.