Properties of oxygen sites at the MoO₃(0 1 0) surface: density functional theory cluster studies and photoemission experiments

Abstract

Ab initio density functional theory cluster studies on the MoO₃(0 1 0) surface as well as ultraviolet photoemission (UPS) experiments on well-crystallized single phase Mo oxides are carried out to examine the electronic structure of the oxide systems. In addition, electronic details of different surface oxygen vacancies are studied by appropriate vacancy clusters. Calculations on embedded clusters as large as Mo₁₅O₅₆H₂₂ confirm the mixed covalent/ionic character of the oxide. The computed width of the O 2sp dominated valence band region of MoO₃, about 7 eV, agrees well with the present photoemission data for MoO₃(0 1 0) samples. The overall shape of the computed densities of states (DOS) in the O 2sp region of MoO₃ is rather similar to the measured UPS intensity curves indicating weak energy dependence of corresponding transition matrix elements. Calculated vacancy energies for the different surface oxygen sites at MoO₃(0 1 0) yield rather large values, 6.8–7.6 eV, which shows that oxygen is bound quite strongly to the substrate. Vacancy formation leads to reduction of neighboring molybdenum centers which expresses itself by increased metal d electron occupation and corresponding DOS contributions above the O 2sp region. This is consistent with the experimental UPS data for MoO₃(0 1 0) where oxygen vacancies have been introduced by mild ion bombardment. It is further supported by the present UPS data for well-crystallized intermediate molybdenum oxides, such as Mo₁₈O₅₂, Mo₈O₂₃, or Mo₄O₁₁. These oxides show, depending on the degree of reduction, one or two additional peaks above the valence band. Characteristic changes in the intensity ratios of the O 2sp peaks can be interpreted on the basis of the theoretical DOS results as a preferential loss of bridging oxygen from the MoO₃ lattice. Mild ion bombardment, a technique which is often used to clean surfaces in UHV experiments, results in the case of MoO₃(0 1 0) in considerable surface reduction. The reduced Mo species is comparable to that in MoO₂ as indicated by the UP spectra. Therefore, mild ion bombardment cannot be considered a suitable tool for the preparation of molybdenum oxide surfaces.

Introduction

Molybdenum oxides in combination with other elements, such as bismuth, vanadium, cobalt, or aluminum, are used as highly active and selective catalysts for many reactions of very different type. Examples of commercial processes are isomerization, polymerization, and production of formaldehyde and acronitrile [1], [2], [3]. Due to the high industrial relevance of Mo-based catalysts it is important to improve their reaction performance, for example, by considering ternary or quaternery mixed oxides with V, W, Nb, Ta [4], [5], [6], [7], [8], [9], [10], [11] or by adding Cu or Sb as promoters [12], [13], [14], [15], [16], [17], [18], [19], [20]. In addition, fundamental understanding of microscopic details of the catalysts has to be gained as a pre-requisite for future material science leading to tailor-made selective catalysts.

Catalytic properties of mixed molybdenum oxides can be discussed in terms of compound and surface structure sensitivity where both concepts are interrelated and result from local geometry and electronic properties of the active sites. Compound sensitivity is clearly seen in all phases of molybdenum oxides of intermediate composition, produced by successive reduction of MoO₃, as a result of the different degree of Mo reduction. Other examples are ordered oxides, where Mo ions are stabilized in different oxidation states. They can form series of oxygen deficient materials known as Magneli phases [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], where defects (oxygen vacancies) order and result in extended shear structures. Most of these phases are catalytically active for partial oxidation [32]. Compound sensitivity, which results from chemical composition and special crystal geometry, may be responsible for the fact that only one out of many possible Mo oxide phases exhibits highest selectivity and activity.

It is well known that molybdenum oxides can exhibit pronounced crystallographic anisotropy. As a result, different exposed crystal faces have different catalytic behavior, and, therefore, participate in various elementary steps of catalytic reactions [33], [34]. This phenomenon, known as surface sensitivity, originates from the local environment of active sites and the orientation of the surface where the active sites are located. For example, oxygen basicity is found to increase with decreasing Mo–O bond strength occurring at different surfaces and may lead to different catalytic behavior. The importance of this factor for hydrocarbon oxidation reactions has become a subject of lively discussion in recent years [1], [2], [3], [11], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47]. Here one can mention the selective conversion of propylene to acrolein [35], [36], [37], [38], [39], [40], [41] at molybdenum trioxide surfaces. The two conversion steps, hydrogen abstraction from and oxygen insertion into the hydrocarbon, are found to take place at different crystal faces of MoO₃ [10]. Hydrogen abstraction occurs at the (0 0 1) or (1 0 0) crystal faces, whereas nucleophilic addition of oxygen into the allylic species takes place at the basal (0 1 0) face [39], [40]. The dependence of catalytic properties on the ratio of exposed (1 0 0) and (0 1 0) faces in MoO₃ catalysts was also reported by Volta [41]. MoO double bonds were proposed to be active for formaldehyde formation [48], [49], [50], [51], while bridging Mo–O–Mo sites were responsible for the transformation into dimethyl ether [51]. In all oxidation processes discussed above, MoO₃ surface oxygen participates and leads to the oxygenated species.

All oxide phases may also expose different types of oxygen vacancies at their surfaces. These vacancies may undergo re-oxidation by gaseous oxygen or may give rise to extended (rather than point) defects [21], [54]. In the latter process, known also as crystallographic shear, changes in the metal-to-oxygen ratio (stoichiometry) are accompanied by rearrangements of the elementary metal–oxygen polyhedral units. Haber [10] proposed that the formation of crystallographic shear planes might be one of the factors creating the ability of these structures to insert oxygen into the organic species in the selective oxidation process. The shear mechanism is supported by quantum chemical calculations on the interaction of MoO₆ octahedra in a Mo₂O₁₀ cluster [52] where the edge-linked geometry without a vacancy is energetically more favorable than the corner-linked geometry (with a vacancy). However, the formation of crystallographic shear planes cannot exclude the existence of point defects at MoO₃ surfaces [53], [55].

For partial oxidation reactions, compound and surface structure sensitivity has been discussed in many studies [1], [2], [3], [10], [15], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [57], [58], [59], [60]. The strong anisotropy of Mo–O bonding (see also Fig. 1c) implies the validity of surface structure sensitivity. In many studies it was neglected that surface imperfections and/or the notorious instability of highly oxidized surfaces, even of model MoO₃ samples, lead always to a substantial amount of defects which represent all Mo-bond situations. Thus the controversy about the structure sensitivity may be traced back to a lack of structural definition of samples rather than to a basic problem. In this situation a rigorous study of the surface catalytic properties of MoO₃ would be a cornerstone in any attempt to disentangle the complex functions of technical catalysts. This was attempted before in several surface science-inspired studies [61], [62], [63]. In these studies the notorious instability of the samples and special treatment conditions for maintaining a usually defective and thus conducting surface were extensively discussed. Electronic properties of surfaces were investigated and valence band spectra were tentatively assigned to local structural features of the MoO₃ motif. The relation of these studies to practical catalysis is, however, uncertain.

The degree of reduction or the defect density of the model systems cannot be related to that of practical catalysis because of structural and functional deficits of the model system. Pure binary MoO₃ is for most relevant reactants (except of highly activated alcohol molecules) either catalytically inactive or requires extremely high temperatures such that the structure is not stable and is reduced to metallic oxides (transient total oxidation `activity'), which are also inactive in catalytic conversions. Thus the promoters mentioned above are essential for the function of the selective oxidation systems but are not incorporated into the binary model system. In addition, the structures of real operating catalyst systems are either not known in detail or highly complex [29], [30], [31]. Thus, a direct correlation between the real catalyst and corresponding model systems cannot be made. The present work builds upon the previous studies but utilizes a series of samples prepared by chemical vapor phase transport (CVT), a method which allows to synthesize single phase binary oxide systems in qualities suitable for surface science studies. In this work the spectroscopic results of the new model system are related to literature data and compared with rigorous theoretical analysis comprising the stoichiometric system and the process of defect formation. In future work these studies will be extended to the analysis of ternary oxide systems accessible in the same quality as the binary oxides described here.

In the theoretical part of this work, we study the local electronic structure of the MoO₃(0 1 0) surface using density functional theory (DFT) cluster models. Embedded surface clusters, as large as Mo₁₅O₅₆H₂₂, represent sections of the elementary bi-layer at the substrate surface. In addition, surface oxygen vacancies are considered by appropriate Mo₁₅O₅₆H₂₂–O clusters and their influence on the geometric, electronic, and energetic behavior at the MoO₃(0 1 0) surface is examined. The electronic structure of the surface systems, characterized by corresponding total and partial valence densities of states (DOS), are compared with the present experimental photoemission data.

In the experimental part of this work, ultraviolet photoemission (UPS) measurements are carried out to study the valence band region of single crystal MoO₃ and of well-crystallized (single-phase) intermediate oxides between MoO₃ and MoO₂, such as Mo₄O₁₁, Mo₈O₂₃, and Mo₁₈O₅₂. Further, oxygen vacancies are introduced at the MoO₃ surface by mild ion bombardment, which is often used in UHV experiments for surface cleaning. For both the MoO₃ with oxygen vacancies and the set of intermediate oxides, the intensity ratio of the O 2sp peaks in the valence band region changes and additional intensity is observed in the band gap. The former result can be interpreted by the cluster calculations as due to preferential loss of bridging oxygen upon reduction while the latter is found to be due to Mo 4d contributions originating from reduction of Mo ions near the vacancies. Thus, the UPS results suggest that even mild ion bombardment can reduce the MoO₃ surface substantially leading to photoemission comparable to that of MoO₂ surfaces.