Atomic Resolution Imaging of Nanoscale Structural Ordering in a

Jul 28, 2012 - Core Laboratories of Imaging and Characterization, King Abdullah University of Science and Technology, Thuwal 23955-6900,. Saudi Arabia...
1 downloads 0 Views 2MB Size
Article pubs.acs.org/cm

Atomic Resolution Imaging of Nanoscale Structural Ordering in a Complex Metal Oxide Catalyst Yihan Zhu,†,‡ Qingxiao Wang,§ Lan Zhao,§ Baiyang Teng,‡ Weimin Lu,*,† and Yu Han*,‡ †

Institute of Catalysis, Department of Chemistry, Zhejiang University, Hangzhou 310028, PR China Advanced Membranes and Porous Materials Center, Chemical and Life Science and Engineering Division, King Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia § Core Laboratories of Imaging and Characterization, King Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia ‡

S Supporting Information *

ABSTRACT: The determination of the atomic structure of a functional material is crucial to understanding its “structure-to-property” relationship (e.g., the active sites in a catalyst), which is however challenging if the structure possesses complex inhomogeneities. Here, we report an atomic structure study of an important MoVTeO complex metal oxide catalyst that is potentially useful for the industrially relevant propanebased BP/SOHIO process. We combined aberration-corrected scanning transmission electron microscopy with synchrotron powder X-ray crystallography to explore the structure at both nanoscopic and macroscopic scales. At the nanoscopic scale, this material exhibits structural and compositional order within nanosized “domains”, while the domains show disordered distribution at the macroscopic scale. We proposed that the intradomain compositional ordering and the interdomain electric dipolar interaction synergistically induce the displacement of Te atoms in the Mo−V−O channels, which determines the geometry of the multifunctional metal oxo-active sites. KEYWORDS: complex metal oxide, electron microscopy, crystallography, M2 phase



INTRODUCTION Complex metal oxides constitute one of the most important families of materials for applications in catalysis1−6 and energy conversion and storage.7−9 For example, MoVTe(Nb)O complex metal oxides have been extensively investigated10,11 as selective oxidation catalysts that are capable of directly functionalizing propane to high value-added oxygenates, e.g., acrolein,5 acrylic acid,2,6 and acrylonitrile.1,2 This functionalization allows the replacement of propylene, a conventional feedstock, by abundant and less expensive propane, and it can be regarded as an upgraded BP/SOHIO (British Petroleum/ Standard Oil of Ohio) process.12 The M1 and M2 phases are two most important oxides, with layered structures in the families of MoVTe(Nb)O catalysts. Although the M2 phase alone is inactive in propane (amm)oxidation reaction,13 it has a “synergetic effect” when used together with the active M1 phase, remarkably enhancing the selectivity toward AA/ ACN.13−15 Moreover, the M2 phase can replenish the tellurium loss of the M1 phase during the reaction to avoid its quick deactivation.15 Studies of such catalysts demonstrated that their superior catalytic activity originates from the unique Mo−V− Te multifunctional metal-oxo sites,16,17 but some puzzling structural complexity remains unsolved, due to the intrinsic structural inhomogeneities of this kind of material. X-ray crystallography, one of the most frequently used techniques for © 2012 American Chemical Society

resolving structures, can precisely determine atomic coordinates and occupancy,18−22 but it provides only average structural information on the whole grain. Local structural features, such as nanoscopic structural/compositional ordering,23−25 grain/ domain boundaries,23,25,26 and defects,26 which have essential impacts on the material properties, are usually “invisible” to Xray crystallography. In this sense, electron microscopy is an alternative and complementary solution, which can generate images with sub-angström spatial resolution23,25−29 from extremely small areas in the specimen, allowing the investigation of local structural features. In particular, aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) can provide nanoscopic compositional information through image contrast (Z-contrast).26−29 In this study, we demonstrate that the combination of X-ray diffraction, HAADF-STEM, and quantum mechanics (QM) calculations provides a comprehensive understanding of the structural inhomogeneities in complex metal oxides, using one of the important MoVTe(Nb)O selective oxidation catalysts (M2 phase) as an example. Previous studies of the M2 phase Received: June 12, 2012 Revised: July 27, 2012 Published: July 28, 2012 3269

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials



show that the Te species exhibits off-center displacement in the Mo−V−O channels of hexagonal tungsten bronze (HTB) structures and its partial disorder gives rise to the formation of orthorhombic “domains” with 3-fold twin orientations.27,30 However, the lack of an explicit atomic-scale chemical and structural elucidation of both orthorhombic domains and their boundaries makes it difficult to determine the crystallographic origin of the domain formation and the nanoscopic driving force for Te displacements. To address these issues, which are important for understanding the nature of the catalytically active sites in the M2 phase, powder X-ray diffraction (PXRD) was combined with the maximum entropy method (MEM) to provide an average pattern for the precise orientations and amplitudes of the in-channel Te displacements. QM calculations further indicate that such displacements are actually driven by both the V substitution position and the Te coordination mode. Using atomic-resolution HAADF-STEM imaging, nanoscopic ordering of the Mo/V composition and Te displacements as well as the continuous change in orientation of the Te displacement inside a single domain are directly observed. Upon analysis of the electric dipole induced by the Te displacement, the atomic structures of domain boundaries, and the three-dimensional domain distribution, we propose for the first time that a nanoscopic compositional ordering and electric dipolar interaction synergistically drive the Te displacement, leading to the averaged structure deviating from the local order in such a complex metal oxide catalyst.



Article

RESULTS AND DISCUSSION

The M2 phase catalyst has an HTB-like Mo−V(Nb)−O solid solution framework consisting of one-dimensional hexagonal channels along the [001] direction, which are enclosed by corner-sharing MO6 (M = Mo, V, or Nb) octahedra and partially occupied by {TeOx} species (Supporting Information, SI Figure S1a). Unlike a typical HTB structure in which the inchannel species simply occupy the geometric centers, the {TeOx} species in the MoVTe(Nb)O oxide has an off-center displacement27 and exhibits partial disorder on the macroscopic scale.30 The Te orientation (i.e., the direction of Te displacement) determines the symmetry of the structure as well as the geometry of TeMn (M = Mo or V) multifunctional metal-oxo active sites. An early electron diffraction study of a ́ MoVTeO M2 phase carried out by Garcia-Gonzá lez et al. showed 3-fold twin orientations of orthorhombic lattices (at 60° to one another) that result from the nanoscopic ordering of Te displacement but are “invisible” under the hexagonal global symmetry from PXRD,30 suggesting that MoVTeO oxide has different structural ordering on the nano- and macroscales. The nanoscopic orthorhombic lattice and macroscopic hexagonal lattice have a relationship in unit cell constants as follows: ahex = √3/3aortho = bortho. Interestingly, in the MoVTeNbO isomorph of the M2 phase, the superstructure reflections associated with orthorhombic lattices can be observed and refined by PXRD in DeSanto et al.’s work,19 implying a higher degree of orthorhombic ordering in this phase. They involved these superstructure reflections for structural refinement and consequently proposed an orthorhombic average structural model with Pmm2 symmetry, which is currently accepted for all the MoVTe(Nb)O M2 phases. However, we argue that the superstructure reflections (e.g., d > 6.3 Å) should not be included for resolving the macroscopically averaged structure because they solely carry structural information of the orthorhombic lattices with short-range order. On the other hand, the 3-fold twin orientations result in significant degeneration of diffraction reflections from different orthorhombic lattices. Since the orthorhombic lattice has only shortrange order,19 the degenerated reflections are relatively more intensive than the others and become resolvable by PXRD. On average, it seems as if these reflections arose from a hexagonal lattice with Te disorder. For instance, the (200) and (110) reflections belonging to orthorhombic lattices with 3-fold twin orientations are degenerated with each other, which could be treated as the (100) reflections of the hexagonal lattice extending all over the grain. In this way, however, the nanoscale structural information within the orthorhombic lattice is missing. Hence, both a macroscale structural model concerning Te disorder (i.e., an X-ray crystallographic model regardless of the orthorhombic superstructure) and a nanoscopic one concerning the Te ordering within the orthorhombic lattice (i.e., an electron crystallographic model) are necessary to address the complete chemistry of the MoVTeO catalyst. A crystallographic model of the MoVTeO oxide was resolved from the LPXRD data set with a high signal-to-noise ratio (S/ N) and data resolution of 0.82 Å by a direct phasing procedure (Supporting Information Methods and Figure S1b). Besides a trace amount of VOMoO4 and TeVO4 impurity that frequently accompanies the MoVTeO M2 phase,31 no orthorhombic superstructure reflections were observed, and all reflections were perfectly indexed by a hexagonal cell (a = 7.2748(6) Å, c = 4.0084(3) Å), similar to the case of an earlier report.30 This

EXPERIMENTAL SECTION

Sample Preparation. In a typical synthesis of the MoVTeO M2 phase complex metal oxide, (NH4)6Mo7O24 (Aldrich, >99.98%, 2.5000g), NH4VO3 (Aldrich, >99.0%, 1.3087 g), and H6TeO6 (Aldrich, >99.9%, 1.9539 g) were vigorously stirred and dissolved in 80 mL of DI water at 353 K. The obtained precipitate was isolated, dried in air, and then calcined at 873 K under a 5 mL/min N2 flow in a quartz tube for 2 h to yield the final product. High-Resolution (HR)STEM Imaging. The STEM specimens were prepared by focused ion beam (FIB) using an FEI Helios NanoLab 400S FIB/SEM dual-beam system under the protection of a Pt/C multilayer. HRSTEM images were collected on an FEI aberration-corrected Titan Cubed S-Twin transmission electron microscope operated at 300 kV with a Fischione model 3000 HighAngle-Annular-Dark-Field (HAADF) detector. The inner and outer electron collecting angles of the HAADF detector were set to 80 and 200 mrad, respectively, providing robust Z-contrast images and eliminating Bragg scattering. The simulation of STEM images was carried out with the CrystalKit and MacTempas bundles of programs using the multislice method. PXRD Characterization. Room temperature high-resolution synchrotron powder X-ray diffraction (SPXRD) data were collected at Powder Diffraction Beamline, Australian Synchrotron (λ = 0.8250 Å). A high-resolution laboratory powder X-ray diffraction (LPXRD) data set was collected at 300 K on a Thermo ARL X’TRA polycrystalline diffractometer with vertical θ−θ Bragg−Brentano geometry and Cu Kα radiation (λ = 1.54056 Å). The structure was resolved by EXPO and MAUD programs. The maximum entropy method (MEM) was utilized via the BayMEM code based on the Sakata-Sato algorithm. Theoretical Calculations. The ab initio DFT calculations including geometric optimization and Born effective charge calculations were performed by CASTEP codes based on the plane-wave pseudopotential method. Perdew−Burke−Ernzerhof was used as the exchange-correlation functional with general-gradient approximations. Our detailed experimental and theoretical methods can be found in the Supporting Information. 3270

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

Figure 1. Crystallographic structure of MoVTeO oxide resolved from PXRD using the maximum entropy method. (a) Projected electron density map (along the [001] direction) of a 2 × 2 × 2 hexagonal supercell isosurface at 1.50 e Å−3 upon which a colored crystallographic model is superimposed: purple for mixed occupied Mo/V, orange for Te1, and red for oxygen. (b) Electron density slice at z = 0.583, displaying the Te1 and Te2 sites. The off-center displacements (δTe1 and δTe2) are indicated, and the orientations of Te1 and Te2 (moving toward an “edge” or a “corner” of a M6O6 hexagonal window) are denoted by a red triangle and a green quadrangle, respectively.

Figure 2. Observed (dot), calculated (line), and difference synchrotron X-ray diffraction profiles for the Rietveld refinement of MoVTeO oxide (λ = 0.8250 Å). (a) An enlarged view of fitted profiles at high angles. (b) Observed (top) and simulated low-angle X-ray diffraction profiles using the crystallographic model defined in Figure 3d (middle) and DeSanto’s model (bottom). Red and black indices refer to hexagonal and orthorhombic order, respectively. Carrot and asterisk denote the reflections belonging to traces of crystalline VOMoO4 and TeVO4 impurities, respectively.

the degenerate ones (i.e., reflections from hexagonal order) (see Supporting Information Figure S3). Following the same phasing procedures, a similar initial hexagonal structural model was derived from SPXRD data with slightly different cell constants (a = 7.26966(5) Å, c = 4.00558(5) Å). In the structural model derived by the direct method, the anomalous atomic displacement parameters of oxo-tellurium (Te1, O1) and apical oxygen (O3) of the MO6 sites suggest a remarkable structural disorder (Supporting Information Figure S1), which cannot be directly resolved with the current data resolution. Higher resolution was achieved by the MEM method (Figure 1 and SI Figure S4), which directly reconstructed the electron densities from the phased structure factors with a real-space precision up to 0.05 × 0.05 × 0.05 Å3. Figure 1a shows the

suggests that the ordering of the orthorhombic superlattice is below the detection limits of the LPXRD, as simulated under an instrumental broadening factor of 0.15 (2θ)° and an S/N ratio of 2000 (∼10 nm, Supporting Information Figure S2), and all the reflections used for the average structure determination are within the long-range order in a single grain. By comparison, SPXRD provides much higher reflection intensities and smaller line width, and is therefore more suitable for probing nanoscopic order. The orthorhombic superstructure reflections, such as (01 L), (21 L), (30 L), and (12 L), etc., can be explicitly observed from a 0.62 Å resolution SPXRD data set collected with λ = 0.8250 Å synchrotron radiation (Figure 2). Apparently, these weak reflections that are generated by nanoscopic order have different line widths from 3271

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

Figure 3. HRSTEM image and crystallographic model of a single domain. (a) HRSTEM image taken along the [001] direction. The inset is the fast Fourier transform (FFT). (b) Magnified image of the marked region in part a upon which the derived Mo−O−V framework model is superimposed (refer to part d). The Te1 and Te2 orientations are marked by red triangles and green quadrangles, respectively. (c) Linear intensity profiles along the [010] and [110] directions, corresponding to the regions marked with green and red dashed rectangles in part b, respectively. (d) Orthorhombic structure model for a single domain (projected along the [001] direction) derived from the HRSTEM image with the unit cell outlined and the Te displacements indicated.

sites (O4/O5 and O1) were carefully examined to determine their occupancy (region α, Supporting Information Figure S4). The results show that Te1 moves 0.67 Å toward the edge of the M6O6 hexagonal window by forming a TeM2 ternary active site (denoted by a red triangle in Figure 1b), while Te2 moves 0.58 Å toward a corner by forming a TeM3 quaternary active site (denoted by a green quadrangle), and that TeO3E is the dominant species (TeO3E/TeO4E = 4.02). The anisotropic disorder of the O3 site is also explicitly identified in region β (SI Figure S4). The oxygen disorder originates from either thermal vibration or bond fluctuation, as indicated by the quantum mechanical calculations (SI Figure S6). To understand the driving force behind the Te displacement, the bond valences of Te cations at the geometric center of the channel and its two off-center counterparts (Te1/Te2) were calculated to be 2.5 and 3.1/3.0, respectively. The higher valences of the off-center positions were closer to the observed quadrivalence for Te cations,30,33 suggesting that the Te displacement is driven by atomic bonding deficiency and thus associated with the composition of

reconstructed total electron density map projected along the [001] direction embedded with the identified atoms. The atomic coordinates were extracted from the charge density maxima and then refined upon satisfactory agreement by the Rietveld method on both LPXRD and SPXRD data sets (Figure 2, Supporting Information Figure S5 and Table S1). Those nondegenerate superstructure reflections in the SPXRD data set are simultaneously refined under an orthorhombic cell (a = 12.57695(48) Å, b = 7.27635(12) Å, c = 4.01753(7) Å) by the Le Bail method as well. From the density distribution truncated at a core electron level in a M6O6 hexagonal window (Figure 1b), we determined that disordered Te cations continuously occupy the off-center 12f Wyckof f sites with an integrated total occupancy of 0.88. Although two density maxima can be extracted for Te/Te1 in a TeO4E unit (E: electron lone pair) and for Te2 in a TeO3E unit, it is difficult to assign an accurate electron density between them due to the severe overlapping and ambiguous “zero-flux” boundary of the electron density gradients between their atomic basins (Bader’s model).32 Alternatively, their associated and well-separated apical oxygen 3272

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

the theoretical study discussed earlier, the observed ordering of the Te orientations implies a nanoscopic compositional (e.g., V or VO) ordering in the MO atomic columns. To verify this implication, line intensity profiles were extracted from the HAADF-STEM image. As illustrated in Figure 3c, the MO columns show alternate strong/weak intensities along both the [010] and the [110] directions, implying a possibility of alternating Mo- and V-rich atomic columns in these directions, i.e., a para-substitution of V into the M6O6 window. One may argue that the intensity difference between the neighboring Mo- and V-rich atomic columns is rather small, which is possibly caused by an atomic column width change or the “channeling effect” that could enhance the wave amplitude at certain atomic columns. However, in this study, we tried to avoid the “channeling effect” by using a large inner detector angle of 80 mrad,34,35 and we did not observe remarkable width difference between V-rich and Mo-rich atomic columns. Hence, these possibilities can be ruled out and the observed STEM intensity variation should be attributed to the compositional ordering. Actually, the nanoscopic Mo/V ordering was implied in our recent study,31 which showed that, at high temperature, MoVTeO can transform to the VOMoO4 nanoislands that have an alternating Mo−V cationic order. It is well-known that the Mo/V solid solution framework in the M2 phase lacks a macroscopic Mo−V order,19 so the formation of VOMoO4 nanoislands most likely originates from an intrinsic nanoscopic Mo−V order in the M2 phase as observed here. Additionally, as is shown in Figure 2, the simulated profiles of the in-plane (100), (010), and (210) reflections belonging to the nanosized orthorhombic superlattice with Mo−V order (model proposed in Figure 3d) well match the experimental ones in SPXRD data that solely come from the scattering of orthorhombic superlattices. In contrast, DeSanto’s orthorhombic structural model without Mo/V ordering19 gives a much larger I(210)/I(010) or I(100)/I(010) ratio. It is thus concluded that nanoscopic Mo−V ordering exists in the MoVTeO M2 phase, with Te1 and Te2 cations moving toward a Mo−O−V edge and an O−Mo−O corner, respectively. As Blom et al. observed in the MoVTeNbO isomorph,27 most TeO atomic columns have comparable profile widths to single-atomic MO columns. Additionally, we found that. even in a nanosized region, the superstructure reflections of the orthorhombic lattice (e.g., the (100) and (010) reflections) can be observed by FFT, which indicates that the orthorhombic distortion does not result from the lateral/longitudinal average of either the Te offsets or the Mo/V composition as described in DeSanto et al.’s model,19 but by a spatial distribution of two different single Te orientations. It is not like the Te orientation inside the lattice ́ as proposed by Garcia-Gonzá lez et al.30 either, since it does not result in an orthorhombic but a hexagonal cell. In fact, the alternation of two different Te rows forms a nanoscopically ordered orthorhombic lattice with the same Te orientation (Te1 or Te2) at the corner of the unit cell and another Te orientation (Te1 or Te2) inside the unit cell. There is also a compositional order on the (001) plane within the lattice. Hence, the orthorhombic lattice can be defined as a “domain” (Figure 3d). In the bulk structure, three types of such domains with different lattice orientations (with 60° between one another) constitute a centrosymmetric P6m hexagonal lattice on the (001) plane on the macroscale (SI Figure S11), with Te1/Te2 cations moving toward six corners/edges of an M6O6 window with equal probabilities.

the M6O6 hexagonal window. We also conclude from the QM calculation results (see SI Table S2) that a Te1 cation tends to move toward an edge of the M6O6 hexagonal window while a Te2 cation tends to move toward a corner. This is consistent with the crystallographic model derived by PXRD (Figure 1b). More precisely, the direction of the Te1 off-center displacement is toward either a vanadium cation or oxygen vacancy (VO) located at the “V−O−M”/“M−VO−M” edge while a Te2 tends to move toward an “O−V−O”/“VO−M−O” corner. The identified TeO4E/TeO3E species here are similar to those proposed by DeSanto et al.,19 but different in geometry and location. They also explained the offset magnitude of Te by the average Mo/V site occupancy but not as explicitly as by the atomic-level driving force model described here. Although conceptually it is difficult to very accurately refine the coordinates or fractional occupancy of oxygen from X-ray diffraction data, due to the weak scattering abilities of light atoms at high angles, the bond valence analysis on Te1 and Te2 sites supports the assignment of the oxotellurium coordination modes derived by electron density analysis. Specifically, a Te1 cation is tricoordinated by M−O−M bridging oxygen atoms and has a bond valence of ca. 1.55. In contrast, a bicoordinated Te2 cation only has a bond valence of 1.22. To reach a similar total valence of Te, it is reasonable that Te1 further bonds to two apical oxygen atoms to gain another valence of 1.55 (TeO4E), while Te2 bonds to a terminal oxygen atom for an extra valence of 1.89 (TeO3E). We emphasize here that the driving force for the off-center Te shift is dependent on both composition and coordination within a single structural unit (i.e., a TeO3E/TeO4E unit in a M6O6 hexagon), which means a transition from Te1 to Te2 sites could be accompanied by a coordination change from TeO4E to TeO3E as well. In an average structure resolved by PXRD, Te1 and Te2 occupy two sets of 12f Wyckof f symmetrically equivalent special points with partial disorder, creating a 6-fold axis parallel to the [001] axis. Electron microscopy is capable of uncovering the local shortrange order in disordered matter. Similar to previous reports,27,30 the phenomenon of a “three-fold twin orientation” was also observed in our study. But more straightforward and sufficient evidence for such a model requires atomic real-space resolution. To this end, we utilized HAADF-STEM to acquire easily interpretable Z-contrast images, first along the [001] direction (Supporting Information (SI) Figure S7). Analysis of the resolving power of STEM imaging indicates that both an off-center Te displacement and an alteration between Te1 and Te2 orientations can be explicitly identified, while only the major orientation (i.e., the one with the largest quantity) can be determined if more than one Te orientation is superimposed along the projection direction (SI Figure S8). As shown in Figure 3, hexagonally arranged MO columns form M6O6 windows, among which brighter TeO columns can be easily distinguished due to the larger atomic number of Te than that of Mo or V (see SI Figure S9a for the simulated image). Parallel to the [010] direction, there are two types of Te rows with different orientations, the Te1 row and the Te2 row, which appear alternately with an interspacing of 3.64 Å (Figure 3b and SI Figure S10). Te1 cations move toward the bottom edge of the M6O6 hexagonal window (red triangles), while Te2 cations move toward the bottom right corners (green quadrangles). This observation coincides well with the proposed crystallographic model (Figure 1) and QM predictions (SI Table S2). Given that the Te orientation is determined by the composition of the surrounding M6O6 hexagonal window, as pointed out by 3273

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

Figure 4. Twin domains and their boundary. (a) An HRSTEM image taken along the [001] direction covering two domains (I and II). The reciprocal lattice orientations of the two domains are indicated in the corresponding FFT (inset). (b) The spatial distributions of the two domains are individually derived from the corresponding (200) and (010) spots by inverse FFT. The domain boundary is indicated by a green dashed line. (c) A magnified image of the marked regions in parts a and b that involve two domains. The orthorhombic cell and lattice direction of each domain are indicated. Two types of Te orientations and the resulting spontaneous polarizations are marked by red triangles and green quadrangles (refer to Figure 1b), PS and PS′, respectively. The domain walls are represented by dashed lines consisting of two segments labeled as A and B. (d) Linear intensity profiles of two neighboring rows of atomic columns in part c that start from the green and red arrows, showing V−V antisites at the domain wall.

We notice that the Te orientation in a single domain determined from STEM (Figure 3d) is actually inconsistent with the QM calculations, which predict Te2 movement toward an O−V−O corner. Moreover, the Te1/Te2 ratio should be close to one in perfect orthorhombic domains, as observed in the above-mentioned local region, which is apparently different from that determined from PXRD. These observations imply that bonding deficiency may not be the only driving force for Te displacement. Considering the Mo/V ordering, the offcenter Te displacement breaks the Cmm centrosymmetry of the Mo−O−V orthorhombic lattice on the (001) plane (Figure 3d), separating the centers of positive and negative charges and thus creating an electric dipole. The spontaneous polarizations caused by the displacements of Te1 (PS) and Te2 (Ps′) have magnitudes of 15.0 and 14.4 μC cm−2, respectively, pointing to different directions (see Supporting Information Methods and Table S3). The dipoles in the Te1 and Te2 rows cannot offset each other and thus produce a nonzero net polarization within a single domain. It is worth noting that individual electric dipoles can be identified only from the real-space image based on the Te orientations because an inversion center is imposed in electron diffraction. Meanwhile, as reported by Jia et al.,23,25

the domain walls (DWs) are important boundaries that separate different polarized domains. Another STEM image was thus taken along the [001] direction that covers two neighboring domains with a lattice orientation separation of 60° (Figure 4 and Supporting Information Figure S10). The two domains are individually visualized by inverse FFT using the (200) and (010) spots, showing that they basically complement each other with slight overlap along the domain boundary (DB) (Figure 4b). The intensity profiling of two MO rows clearly shows that the alternating Mo−V atomic columns in domain I along the [010] direction are terminated by dual vanadium (V−V) antisites at the DB (Figure 4c and 4d). This observation supports the proposed models of the 60° lattice rotational order among different domains and the Mo−V compositional order within each domain (Supporting Information Figure S11). In Figure 4c, the domain wall (DW) is contoured by yellow dashed lines. Segment A of the DW contains two atomic layers where the Te orientations are irregular due to the local strain, while segment B does not have such a phenomenon and two domains meet there without a gradual transition. The DW exhibits a mixed character of a neutral “longitudinal domain wall” (LDW) and a charged 3274

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

Figure 5. Direct mapping of the electric dipoles in MoVTeO oxide. An HRSTEM image taken along the [001] direction covering three neighboring domains (refer to Supporting Information Figure S13) is marked with arrows according to the Te orientations to show the electrical dipoles. The [±100] directions of three domains are indicated by three intercrossing double arrows in the upper-left corner. Two domain boundaries are represented by the yellow dashed lines, and one grain boundary is represented by the solid line. In the central domain, electric dipoles are classified into two groups based on the rotation directions using black and orange arrows, while the circles represent unidentifiable electric dipole directions due to the superimposition of equally thick domains as demonstrated in Supporting Information Figure S8. The electrical dipoles in the other two domains are also marked with green and blue arrows, respectively.

“transverse domain wall” (TDW) everywhere, with its normal neither parallel nor perpendicular to the polarization direction.23 The charging of a DW is usually compensated by mobile charges, ionic defects, or valence variations,23,36 whereas imperfect screening of DW charging could induce a long-range depolarization field across the domains that drives the intradomain electric dipoles to continuously rotate and reorganize in vortices or flux-closure structures.25,37−39 Hence, the on-plane interdomain dipolar interactions may play an additional role in driving the Te off-center displacement. Figure 5 and Supporting Information Figure S12 show an HRSTEM image taken along the [001] direction that covers three types of domains (I, II, and III; see Supporting Information Figure S13 for the domain distribution), in which the electric dipoles are plotted based on the Te orientations. It is clear that DB I and II are positively and negatively charged, respectively, due to the mainly head-to-head and tail-to-tail coupled dipoles (Figure 5), creating a net depolarization field across domain II. If only the chemical driving force is accounted for, two types of electric dipoles associated with the Te1 and Te2 orientations should distribute alternately, based on the compositionally ordered single domain structure (Supporting Information Figure S11 and Table S2). However, the identified dipoles in domain II form two groups of flux-closure structures by continuous rotation (Figure 5). In comparison with the intradomain

compositional ordering that introduces a strong correlation between two Te orientations in adjacent rows through “chemical bonding”, the interdomain dipolar interaction provides an “electrostatic” driving force to organize the Te orientations in flux-closure structures. They synergistically lead to a single domain comprising two groups of continuously changed Te orientations, which result in an unequal amount of Te1 and Te2 species and explain why the observed Te orientations (Figure 3d and 5) differ from those predicted by the QM calculations (Supporting Information Figure S11). Notably, the continuous dipole rotation all over domain II is not induced by the strain at the antiphase boundary (e.g., Segment A in Figure 4c), which would only result in a locally irregular distribution of electric dipoles. The electric dipole rotation creates a lot of hexagonal or partially disordered “defects”, as shown schematically in Supporting Information Figure S14, and remarkably decreases the orthorhombic domain size to less than ∼10 nm, which makes the superstructure of an MoVTeO oxide “invisible” by LPXRD. As mentioned above, the observed intensity difference between Mo and V atomic columns (Figure 3c) is less remarkable compared to the expected Z-contrast (i.e., approximately proportional to Z2). This can be attributed to the switching of different domains along the [001] direction, which leads to the contrast difference between Mo/V columns 3275

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

Figure 6. Domain switching along the [001] direction and Te defects. (a) An HRSTEM image of MoVTeO oxide covering three types of domains and the corresponding FFT (inset). The electron beam incidence is along the [010], [110], and [110̅ ] directions for different domains. In the FFT, the (201)/(111)/(111) spots, respectively, belonging to the three types of domains have different intensities, but they overlap, as marked by white triangles. When they are used for inverse FFT, the domain with the [010] projection is visualized with lower contrast and thus distinguished from the other two, as shown in part b. (c) A magnified STEM image with a projected atomic columns model: orange is the TeMO column, and purple is the MO column. (d) Linear intensity profiles for the corresponding rows in part c marked with the same colors. The pink flat lines indicate the constant ⁗) intensity within the single domains while the black steps reflect the intensity change between neighboring domains. Pink triangles indicate (VTe defects. (e) The lattice parameter, c, is measured from the MO atomic column positions in the STEM image, showing a marked change at the domain boundary.

Figure 6b, inverse FFT using the overlapped spots selectively displays the domain of the [010] projection due to the contrast difference. In this way, the interpenetration of different domain segments (1−5 nm) along the [001] direction is evident. An intensity profile shows the transition from VO-rich columns to MoO-rich columns, providing additional proof for the domain switching along this direction (see the black curve in Figure 6d). The switching is coupled with lattice strain due to a change in the local bonding environment (e.g., from Mo−O−Mo to Mo−O−V bonds), as evidenced by the variation in the lattice parameter, c, at the DB (Figure 6e). The three-dimensional domain intergrowth along with inhomogeneous Te orientations results in different compositions and geometries of multifunctional cationic active sites (e.g., TeM2 or TeM3) as well as the nonuniform spatial

being diminished, as simulated in Supporting Information Figure S15. To directly observe the domain switching in the [001] direction, a STEM image was taken along its perpendicular direction (Figure 6 and Supporting Information Figure S7). The lattice parallel to the [001] direction shows alternating strong/weak contrast, corresponding to the TeMO and MO atomic columns (Figure 6c and Supporting Information Figure S9b). Due to the “three-fold twin” relationship among the domains, this projection corresponds to the [010], 11̅0, and [110] directions for different domains. In the FFT, the (201)/(111)/(111) spots, respectively, belonging to the three types of domains overlap (Figure 6a). However, the large intensity difference between the (201) and (111) spots helps us to discriminate one type of domain from the others (Supporting Information Figure S16). As shown in 3276

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

Notes

distribution of the TeO3E and TeO4E oxo-species. The TeO3E species with a short TeO terminal bond (1.82 Å) was reported to attain high activity toward allylic hydrogen.16,17,40 The Mo/V ordering in a single domain provides an ideal configuration of the active sites (i.e., Mo−V−Te) for selective oxidation as well.16,17,41 On the other hand, there are ionic defects, less active sites (Mo−Mo antisites), and overoxidation active sites (V−V antisites16,17) enriched in the DWs to screen the DW charging. The possible enrichment of localized Te defects (V⁗ Te) at the vicinity of DWs can also be viewed from an abrupt intensity decrease in the linear profiles of the TeMO atomic columns (Figure 6c and d). The high energy electron bombardment also gives rise to the beam-induced Te sublimation or migration, which is observed before28,31 but is less likely to take place only at a localized Te atomic column upon a raster-scanned electron probe. Since the presence of Te defects and antisites at the DWs decreases the number of the optimal (Mo−V−Te) active sites of MoVTeO oxide, the domain size that determines the density of the DWs would have a direct impact on the reaction and larger domains favor higher selectivity. This suggests the possibility of controlling the reaction pathway by tailoring the domain size and explains the reported selectivity difference between different MoVTe(Nb)O hexagonal phases. For example, due to the stable +5 valence state of Nb, it is not surprising that doping Nb5+ in the MoVTeO phase markedly destabilizes the DWs and enlarges the orthorhombic domain sizes to be macroscopically recognizable by PXRD.19 Consequently, the MoVTeNbO isomorph exhibits much higher selectivity (e.g., toward acrylonitrile) than the Nb-free one,42,43 although Nb does not offer active sites on its own.16,17

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This research was supported by baseline research funds from King Abdullah University of Science and Technology. We appreciate the support of Dr. Qinfen Gu from Australian Synchrotron research center for carrying out the powder X-ray diffraction experiments. We thank Prof. Guanglie Lv (National Science Park, Zhejiang, PR China) for useful discussions.





CONCLUSIONS New insights into the atomic structures and nature of active sites in a complex metal oxide catalyst have been provided by this study, including the identification of nanoscale structural ordering and the reorganization of cationic sites by electric dipolar interactions. Our results not only explain the current catalytic phenomenon but also provide new routes for designing more selective catalysts. We also demonstrated that a simple periodic or local crystallographic model may be insufficient to describe the inhomogeneity of atomic structures in a complex metal oxide. In a broader sense, other mixed metal oxide materials possessing partial disorder may also have hidden nanoscopic periodicities owing to the nanoscopic compositional ordering.44 The aberration-corrected STEMHAADF technique works as a powerful tool for direct atomiclevel imaging including Z-contrast composition information, while PXRD as a complementary technology extends the resolution to the subatomic level for precise atomic position determination and electron density study.



ASSOCIATED CONTENT

S Supporting Information *

Detailed experimental and theoretical methods and supporting figures (Figures S1−S16) and tables (Tables S1, S2, and S3) (PDF). This material is available free of charge via the Internet at http://pubs.acs.org.



REFERENCES

(1) Ushikubo, T.; Koyasu, Y.; Nakamura, H. U.S. Patent 5750 760, 1998. (2) Tsuji, H.; Koyasu, Y. J. Am. Chem. Soc. 2002, 124, 5608. (3) Pradhan, S.; Bartley, J. K.; Bethell, D.; Carley, A. F.; Conte, M.; Golunski, S.; House, M. P.; Jenkins, R. L.; Lloyd, R.; Hutchings, G. J. Nat. Chem. 2012, 4, 134. (4) Gai, P. L.; Kourtakis, K. Science 1995, 267, 661. (5) Zhu, Y. H.; Lu, W. M.; Li, H.; Wan, H. L. J. Catal. 2007, 246, 382. (6) Ushikubo, T.; Nakamura, H.; Koyasu, Y.; Wajiki, S. U.S. Patent 5380 933, 1995. (7) Mor, G. K.; Prakasam, H. E.; Varghese, O. K.; Shankar, K.; Grimes, C. A. Nano Lett. 2007, 7, 2356. (8) Choi, W. S.; Chisholm, M. F.; Singh, D. J.; Choi, T.; Jellison, G. E.; Lee, H. N. Nat. Commun. 2012, 3, 689. (9) Gutierrez, J.; Llordes, A.; Gazquez, J.; Gibert, M.; Roma, N.; Ricart, S.; Pomar, A.; Sandiumenge, F.; Mestres, N.; Puig, T.; Obradors, X. Nat. Mater. 2007, 6, 367. (10) Wagner, J. B.; Timpe, O.; Hamid, F. A.; Trunschke, A.; Wild, U.; Su, D. S.; Widi, R. K.; Abd Hamid, S. B.; Schlogl, R. Top. Catal. 2006, 38, 51. (11) Sanchez, M. S.; Girgsdies, F.; Jastak, M.; Kube, P.; Schlögl, R.; Trunschke, A. Angew. Chem., Int. Ed. 2012, 51, 7194. (12) Grasselli, R. K.; Tenhover, A. A. In Handbook of Heterogeneous Catalysis, 2nd ed.; Ertl, G., Knözinger, H., Schüth, F., Weitkamp, J., Eds.; Wiley-VCH: Weinheim, 2008; Vol. 7, p 2303. (13) Holmberg, J.; Grasselli, R. K.; Andersson, A. Appl. Catal., A: Gen. 2004, 270, 121−134. (14) Korovchenko, P.; Shiju, N. R.; Dozier, A. K.; Graham, U. M.; Guerrero-Perez, M. O.; Guliants, V. V. Top. Catal. 2008, 50, 43−51. (15) Baca, M.; Aouine, M.; Dubois, J. L.; Millet, J. M. M. J. Catal. 2005, 233, 234−241. (16) Grasselli, R. K.; Burrington, J. D.; Buttrey, D. J.; DeSanto, P.; Lugmair, C. G.; Volpe, A. F.; Weingand, T. Top. Catal. 2003, 23, 5. (17) Grasselli, R. K.; Buttrey, D. J.; DeSanto, P.; Burrington, J. D.; Lugmair, C. G.; Volpe, A. F.; Weingand, T. Catal. Today 2004, 91−2, 251. (18) Sadakane, M.; Watanabe, N.; Katou, T.; Nodasaka, Y.; Ueda, W. Angew. Chem., Int. Ed. 2007, 46, 1493. (19) DeSanto, P.; Buttrey, D. J.; Grasselli, R. K.; Lugmair, C. G.; Volpe, A. F.; Toby, B. H.; Vogt, T. Z. Kristallogr. 2004, 219, 152. (20) Nishimura, S.; Kobayashi, G.; Ohoyama, K.; Kanno, R.; Yashima, M.; Yamada, A. Nat. Mater. 2008, 7, 707. (21) Collins, D. M. Nature 1982, 298, 49. (22) Harris, K. D. M.; Tremayne, M. Chem. Mater. 1996, 8, 2554. (23) Jia, C. L.; Mi, S. B.; Urban, K.; Vrejoiu, I.; Alexe, M.; Hesse, D. Nat. Mater. 2008, 7, 57. (24) Xu, Z. K.; Gupta, S. M.; Viehland, D.; Yan, Y. F.; Pennycook, S. J. J. Am. Ceram. Soc. 2000, 83, 181. (25) Jia, C. L.; Urban, K. W.; Alexe, M.; Hesse, D.; Vrejoiu, I. Science 2011, 331, 1420. (26) Pyrz, W. D.; Blom, D. A.; Sadakane, M.; Kodato, K.; Ueda, W.; Vogt, T.; Buttrey, D. J. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 6152. (27) Blom, D. A.; Pyrz, W. D.; Vogt, T.; Buttrey, D. J. J. Electron. Microsc. 2009, 58, 193. (28) Pyrz, W. D.; Buttrey, D. J.; Blom, D. A.; Vogt, T. Angew. Chem., Int. Ed. 2008, 47, 2788.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; [email protected]. 3277

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278

Chemistry of Materials

Article

(29) Zhang, W.; Trunschke, A.; Schlogl, R.; Su, D. S. Angew. Chem., Int. Ed. 2010, 49, 6084. (30) García-González, E.; Nieto, J. M. L.; Botella, P.; GonzalezCalbett, J. M. Chem. Mater. 2002, 14, 4416. (31) Zhu, Y. H.; Wang, Q. X.; Zhao, L.; Han, Y. Angew. Chem., Int. Ed. 2012, 51, 4176. (32) Bader, R. F. W. Atoms in Molecules: A Quantum Theory; Clarendon Press, Oxford Science Publications: Oxford, 1994. (33) Baca, M.; Millet, J. M. M. Appl. Catal., A: Gen. 2005, 279, 67. (34) Pennycook, S. J. Ultramicroscopy 1989, 30, 58. (35) Gemmell, D. S. Rev. Mod. Phys. 1974, 46, 129. (36) Wu, X. F.; Vanderbilt, D. Phys. Rev. B 2006, 73, 020103. (37) Naumov, I. I.; Bellaiche, L.; Fu, H. X. Nature 2004, 432, 737. (38) Naumov, I.; Bratkovsky, A. M. Phys. Rev. Lett. 2008, 101. (39) Pan, X. Q.; Nelson, C. T.; Winchester, B.; Zhang, Y.; Kim, S. J.; Melville, A.; Adamo, C.; Folkman, C. M.; Baek, S. H.; Eom, C. B.; Schlom, D. G.; Chen, L. Q. Nano Lett. 2011, 11, 828. (40) Goddard, W. A.; Liu, L. C.; Mueller, J. E.; Pudar, S.; Nielsen, R. J. Top. Catal. 2011, 54, 659. (41) Grasselli, R. K.; Tenhover, A. A. In Handbook of Heterogeneous Catalysis, 2nd ed.; Ertl, G., Knözinger, H., Schüth, F., Weitkamp, J., Eds.; Wiley-VCH: Weinheim, 2008; Vol. 7, p 3489. (42) Andersson, A.; Haggblad, R.; Wagner, J. B.; Deniau, B.; Millet, J. M. M.; Holmberg, J.; Grasselli, R. K.; Hansen, S. Top. Catal. 2008, 50, 52. (43) Holmberg, J.; Grasselli, R. K.; Andersson, A. Appl. Catal., A: Gen. 2004, 270, 121. (44) Blom, D. A.; Li, X.; Mitra, S.; Vogt, T.; Buttrey, D. J. ChemCatChem 2011, 3, 1028−1033.

3278

dx.doi.org/10.1021/cm301828n | Chem. Mater. 2012, 24, 3269−3278