Atomic-level imaging of Mo-V-O complex oxide phase intergrowth, grain boundaries, and defects using HAADF-STEM

Abstract

In this work, we structurally characterize defects, grain boundaries, and intergrowth phases observed in various Mo-V-O materials using aberration-corrected high-angle annular dark-field (HAADF) imaging within a scanning transmission electron microscope (STEM). Atomic-level imaging of these preparations clearly shows domains of the orthorhombic M1-type phase intergrown with the trigonal phase. Idealized models based on HAADF imaging indicate that atomic-scale registry at the domain boundaries can be seamless with several possible trigonal and M1-type unit cell orientation relationships. The alignment of two trigonal domains separated by an M1-type domain or vice versa can be predicted by identifying the number of rows/columns of parallel symmetry operators. Intergrowths of the M1 catalyst with the M2 phase or with the Mo₅O₁₄-type phase have not been observed. The resolution enhancements provided by aberration-correction have provided new insights to the understanding of phase equilibria of complex Mo-V-O materials. This study exemplifies the utility of STEM for the characterization of local structure at crystalline phase boundaries.

Selective catalytic oxidation of light hydrocarbons is of tremendous commercial importance for the production of a variety of key industrial organic chemicals and intermediates. Existing processes, valued in the billions of US dollars per annum, are predominantly based on oxidation of olefin feeds; however, there are significant economic and environmental benefits to replacement by more energy- and carbon-efficient paraffin-fed processes. Pursuit of active and selective catalysts for use in such replacement processes is a currently very active area of research. Mixed-metal oxides in the system Mo-V-Nb-Te-O, when prepared under mildly reducing conditions, may consist of one or more of several network “bronze” structures with mixed valences for Mo and V (1–4). Materials from this system show promise as catalysts for selective oxidation of ethane to ethylene, propane to acrylic acid, and, in the presence of cofed ammonia, propane ammoxidation to acrylonitrile. One phase in particular, commonly designated as “M1,” has been found to be essential as a catalyst component for selective paraffin oxidation, and under some conditions seems to be promoted by coexistence with another phase designated as “M2” (1–18). Our interest in chemical and structural inhomogeneities observed in this system is driven by the apparent need for composite M1/M2 phase coexistence to realize the most active, selective, and stable catalysts.

The M1 phase has an orthorhombic unit cell with a structure comprised of a network of interconnected pentagonal {Mo₆O₂₁} rings joined together by linking octahedra, resulting in the formation of nanoscale hexagonal and heptagonal channels (4, 14, 19–21). The M2 phase is belongs to the hexagonal tungsten bronze family, but exhibits a small orthorhombic distortion (4). In its pure form, the M2 phase is efficient in the conversion of propylene to the same products (11, 13, 22) by selective oxidation and ammoxidation. Using propane as a feedstock, the composite M1/M2 system catalyzes the production of acrylonitrile with yields up to 62% (23) and in the case of the synthesis of acrylic acid at yields up to 53% for selective oxidation (24).

Following the initial studies, significant effort has been focused on improvement of this system through elemental substitutions or the development of new structural variants (8, 18, 25–29). Compositional variants of the M1 phase have been successfully prepared ranging from the simplest two component Mo-V-O case to more complex formulations that commonly incorporate combinations of Te, Sb, Nb, and/or Ta (8, 18, 25–29). Alternatively, several other Mo-V-O-based phases related to the M1 catalyst have been identified and are comprised of similar structural units, involving common pentagonal Mo building blocks {Mo₆O₂₁} which consist of five {MoO₆} octahedra that each share one edge with a {MoO₇} pentagonal bipyramidal center. These structurally related materials are either pseudotetragonal and similar to the well-known Mo₅O₁₄-type structure originally reported by Kihlborg et al., or a recently discovered phase with trigonal symmetry (27, 30, 31). Fig. 1 depicts an idealized representation of each of these phases, where the common pentagonal unit is comprised of metal-oxygen octahedra (blue) surrounding a pentagonal bipyramidal site (orange).

Fig. 1.

Ideal two-dimensional 2 × 2 unit cells for three molybdenum vanadate structures: M1-type (Left), Mo₅O₁₄-type (Center), and trigonal (Right).

We use aberration-corrected high-angle annular dark-field (HAADF) imaging to study various hydrothermally prepared Mo-V-O phases at the atomic-level. A major advantage of using HAADF imaging is that the scattered electrons used to generate the image are mainly incoherent, and the image contrast scales approximately with the square of the atomic number (Z) (32, 33). The enhanced spatial resolution and image contrast provided by an aberration-corrected STEM electron probe, coupled with the chemical sensitivity of the HAADF technique provides a situation in which crystalline phases can be distinguished based on their chemical and/or structural differences. In our prior work, we have taken advantage of the Z² relationship to analyze chemical differences between various orthorhombic M1 formulations such as Mo-V-Te-O, Mo-V-Te-Nb-O, and Mo-V-Te-Ta-O (34). In one case, contrast variations within HAADF images suggested that a single Ta-doped crystallite was made up of isostructural orthorhombic domains distinguished by two distinct levels of Ta content in the pentagonal biprismatic center sites (34), consistent with a solid state miscibility gap. These phases are chemically, but not structurally, distinct.

In this study, we utilize aberration-corrected HAADF-STEM imaging to characterize the nature of registry between structurally distinct intergrowth phases in the Mo-V-O system. We report, to the best of our knowledge, previously undescribed evidence of coherent intergrowths between the orthorhombic M1-type phase and the related trigonal Mo-V-O phase. Based on the atomically resolved images, structural models describing the nature of the phase boundaries are developed. Using these structural models, it is possible to predict phase/antiphase alignment of the next-nearest neighbor grains of the same phase based on the width of the intervening domain. In addition to the coherent phase intergrowth, we also present and describe additional defect structures and grain boundaries commonly observed in the M1-type phase. The intergrowths observed in this study are unique examples of lattice registry in this important complex oxide bronze family.

Models

Ideal model structures for the M1-type, Mo₅O₁₄-type, and trigonal Mo-V-O oxides are shown in Fig. 1. The DeSanto et al. model for the refined MoVTeNbO structure was modified and idealized to represent the M1-type Mo-V-O by removing Te from the channels and replacing Nb with Mo in the pentagonal center (4). The model for the Mo₅O₁₄-type Mo-V-O was derived from the single component tetragonal subcell for Mo₅O₁₄ presented by Yamozoe and Kihlborg (30). For the trigonal phase, we have adopted the model proposed earlier by Sadakane et al. based on the analysis of powder x-ray diffraction data (27). Each phase is based on a network of pentagonal {Mo₆O₂₁} building blocks that contain an {MoO₇} pentagonal bipyramidal center (orange) surrounded by five edge-shared {MoO₆} octahedra (dark blue) (4). The pentagonal units are interconnected by several {MO₆,M = Mo or V} linking octahedra (4). Controlling the number and alignment of these linking sites dictates the final structure.

Results

HAADF Imaging and Characterization of Trigonal-Orthorhombic Phase Intergrowths.

In the process of screening two-component Mo-V-O samples by electron microscopy, impurity domains of the M1-type phase were observed on occasion within crystallites that were otherwise trigonal (Fig. S1). The trigonal phase is easily identified in HAADF images of [001] projections by its triangular cluster of three heptagonal rings (henceforth called the heptagonal triplet). For this crystallite, however, visual inspection of this image reveals several domains of the trigonal phase bounded by narrow intergrowths of the orthorhombic M1-type phase. It is also worth mentioning that the edges of the crystallite parallel to the electron beam preferentially terminate with the M1-type phase, suggesting that termination with the M1 phase may be energetically favored over the trigonal phase for these lateral faces.

To identify the nature of internal boundaries between the intergrowth domains, higher magnification HAADF images were acquired; an example is shown in Fig. 2. The upper left portion of this image shows the trigonal phase, whereas the lower right portion displays the M1-type structure. Idealized models of both structures show good qualitative agreement when superimposed on the image. Geometric representations of the plane group symmetry operators in the [001] orientation for each phase, p3 for the trigonal phase, and p2gg for the M1-type phase, were also superimposed on top of the image. For each operator, different colors distinguish the unique positions. For the trigonal molybdenum vanadate, the unique operators represented with orange and blue triangles are positioned within chemically equivalent hexagonal channels. For the remainder of figures within this text, the green 3-fold rotational operator within the p3 model will always be aligned with the heptagonal triplet and the light-blue binary operators within the p2gg model will always be aligned with the light-blue positions of the M1-type unit cell.

Fig. 2.

High-resolution HAADF image enlarging one of the twin boundaries. The left side of the twin boundary is the trigonal Mo-V-O phase and the right side of the boundary is the orthorhombic M1-type Mo-V-O phase. Idealized representations of the trigonal and the M1-type unit cell along with their respective plane groups are superimposed onto the image. Good qualitative agreement exists between the idealized models and the HAADF image. The arrangements of symmetry operators for the p2gg and p3 plane groups are included for reference. Different colors are used to represent the crystallographically distinguishable symmetry operators.

Additional high-resolution HAADF images of the crystallite with intergrowths are shown in Fig. S2. Within the hybrid crystallite, trigonal domains separated by M1-type intergrowths are observed in the same orientation (Fig. S2B) or they are aligned such that the heptagonal triplet is rotated by 180° (Fig. S2 A, C, and E). Analysis of the entire particle shown in Fig. S1 indicates that the same types of registry patterns occur when M1 domains are interrupted by a trigonal intergrowth domain. The four possible intergrowth combinations are T-M1-T, T-M1-T_rot, M1-T-M1, and M1-T-M1_rot. In this notation, T represents the trigonal phase, M1 represents the M1-type phase, and the subscript “rot” indicates a phase that is rotated with respect to its phase-equivalent partner on the opposite side of the intergrowth. These triple domain sequences are represented in Figs. 3–6; the corresponding idealized structural models are presented in Figs. 3B, 4B, 5B, 6D. In all cases, the atomic-level registry between the domains is coherent. A set of triple domain intergrowth representations constructed using only the symmetry operators from the p2gg (M1-type) and the p3 (trigonal) plane groups is shown sequentially in Figs. 3C, 4C, 5C, and 6E.

Fig. 3.

(A) HAADF image of a T-M1-T_rot phase intergrowth. (B) Idealized representation of the T-M1-T_rot using the M1-type and trigonal unit cells displayed in Fig. 1. (C) Reconstruction of the intergrowth using the plane group symmetry operators. Operators have been removed if the immediate local symmetry is broken as a result of proximity to the interface.

Fig. 4.

(A) HAADF image of a T-M1-T phase intergrowth. (B) Idealized representation of the T-M1-T using the M1-type and trigonal unit cells displayed in Fig. 1. (C) Reconstruction of the intergrowth using the space group symmetry operators. Operators have been removed if the immediate local symmetry is broken as a result of proximity to the interface. Note that one of the M1-type unit cells have been removed for simplicity.

Fig. 5.

(A) HAADF image of a M1-T-M1_rot phase intergrowth. (B) Idealized representation of the M1-T-M1_rot using the M1-type and trigonal unit cells displayed in Fig. 1. (C) Reconstruction of the intergrowth using the plane group symmetry operators. Operators have been removed if the immediate local symmetry is broken as a result of proximity to the interface.

Fig. 6.

(A) HAADF image of the particle shown in Fig. 2 highlighting two M1-type intergrowths separated by a large trigonal domain. (B) High-resolution HAADF image showing the M1-domain labeled “b.” (C) High-resolution HAADF image showing the M1-domain labeled “c.” (D) HAADF image showing the trigonal intergrowth with a width of 18 heptagonal triplets separating the two M1-type domains. (E) Simplified and idealized representation of the M1-T-M1 phase intergrowth using the M1-type and trigonal unit cells displayed in Fig. 1. (F) Reconstruction of the intergrowth using the plane group symmetry operators. Operators have been removed if the immediate local symmetry is broken as a result of proximity to the interface.

For the T-M1-T_rot domain structure (Fig. 3 A and B), the a axis of the left trigonal grain is antiparallel to the a axis of the M1-type intergrowth and the b axis of the right trigonal domain is parallel to the a axis of the M1-type phase. Because the center M1 domain was less than a single unit cell in width, only one column of binary operators remain from the plane group representation of the junction in Fig. 3C. Alternatively, for the T-M1-T domain structure (Fig. 4 A and B), the b axes of both of the trigonal domains are antiparallel to the a axis of the M1-type intergrowth. In this case, two columns of binary operators remain in the symmetry representation for the junction in Fig. 4C. In general, for an M1 intergrowth separating two trigonal domains, if there are 2n (n: integer) columns of binary operators present in the M1 domain, then the trigonal phase on either side of the M1 intergrowth will have the same orientation. If there are 2n + 1 columns of binary operators, then the two trigonal phases separated by an M1 intergrowth will be rotated by 180° about the a axis and by 60° in the counterclockwise direction about the c axis.

For the M1-T-M1_rot domain structure (Fig. 5 A and B), the a axis of the left M1-type grain is parallel with the a axis of the trigonal intergrowth, and the a axis of the right M1-type grain is antiparallel with the trigonal intergrowth. The symmetry operator representation shown in Fig. 5C depicts a junction in which there are three staggered columns of 3-fold rotational operators separating the two M1-type domains. For the M1-T-M1 case, it was necessary to use multiple images to describe the intergrowth structure (Fig. 6 A–D). Fig. 6 B and C show high-resolution images from the areas labeled in Fig. 6A. The image in Fig. 6D shows that the two M1-type intergrowths from Fig. 6 B and C are separated by a trigonal band with a width of 18 heptagonal triplet units. In this case, analysis of the two M1 domains confirms that this is an M1-T-M1 intergrowth. This intergrowth can be simplified and idealized by removing the excess trigonal unit cells as represented in Fig. 6E. For this case, the a axes of both M1-type domains are antiparallel to the a axis of the trigonal phase. The representation of symmetry operators in Fig. 6F shows that there are six staggered columns of 3-fold rotational operators separating the two M1-type domains. In general, for a trigonal intergrowth separating two M1-type domains, if there are 3n (n: even) columns of 3-fold rotational operators present in the trigonal domain, then the M1 on either side of the trigonal phase intergrowth will have relative orientations without rotational relationships. If there are 3n (n: odd) columns of 3-fold rotational operators, then the two M1 phases separated by a trigonal intergrowth will be rotated by 180° about b. This can be further simplified by counting the number of heptagonal triplets separating the two M1-type domains. If there is an even number of heptagonal triplets, then the M1 intergrowth will have the same orientation. If odd, then the two M1 phases separated by a trigonal intergrowth will be rotated by 180° about b.

Despite the differences in the unit cell orientations that create the phase boundaries, the atomic-level registry between the two phases is the same. Based on the idealized schematics for the junctions in Figs. 3–6, the overlap of the phases occurs through the sharing of a string of pentagonal units connected by mixed Mo/V octahedra. This is also confirmed when comparing the junctions using symmetry operator representations.

Defect Structures Within the Mo-V-O M1-Type Phase.

In the Mo₅O₁₄-type sample prepared from M1-type phase, several M1-type impurity crystallites were observed in addition to the Mo₅O₁₄-type phase. The coexistence of this impurity phase is a clear indication of a successful, but incomplete, transformation during the final synthetic step. A HAADF image of one of the M1-type impurities is shown in Fig. 7A. In this crystallite, three distinct step changes in the image contrast are observed as the image is viewed left to right. These abrupt changes in contrast are indicative of discrete steps in the crystallite thickness. Starting on the left, there is coherent atomic-scale registry across the first step. However, there is a drastically different structure present at the second boundary, which is enlarged and displayed in Fig. 7B. Close inspection of the grain boundary reveals features that are very similar to the heptagonal triplet that is commonly observed in the trigonal Mo-V-O phase. The difference in this case is the number of atomic columns that are resolved in the center of this triplet. In the trigonal case, the current structural model predicts only three atomic columns separating the three heptagonal rings (27). In this grain boundary, the heptagonal rings are separated by four or five atomic columns. The consequence of these additional columns is a compression in the typically round heptagonal rings. A similar compression is observed for the hexagonal rings in the Mo₁₇O₄₇ structure (31).

Fig. 7.

(A) HAADF image showing a M1-type crystallite with a trigonal-like grain boundary separating two M1-type crystalline grains. (B) Enlarged HAADF image of the grain boundary. The difference between the trigonal-like defect and the heptagonal triplet within the trigonal phase is the observation of more than three atomic columns separating heptagonal channels.

Using HAADF imaging, we have observed several poorly ordered crystallites in which there are voids, intergrowth precipitates, and disordered fragments. An example of this incomplete ordering is shown in Fig. 8; this disordered crystallite was found within a mixture of other crystallites showing much better order and derived from the synthetic procedure for trigonal phase preparation. The ordered regions within this crystallite predominantly display M1-type local order with a variety of azimuthal orientations (around a common c axis) from one local domain to another. This crystallite also contains small trigonal-like strips (indicated by the white arrows) or clusters (black arrows), and several highly disordered regions.

Fig. 8.

HAADF image showing a highly defective M1-type impurity within the trigonal sample showing small pockets of the trigonal phase (black arrows) and trigonal-like grain boundaries (white arrow).

Discussion

The formation of intergrowths is consistent with the proposed growth mechanism of the self-assembly of complex Mo₆O₂₁ pentagonal units in solution with smaller VO_x and MoO_x groups to form the larger Mo-V-O structures (27, 35). It was proposed that the final phase, orthorhombic M1-type or trigonal, is highly dependent on solution conditions, and pH has been identified to be a significant variable (27). The formation of these intergrowths is likely to be the result of an initial 2D self-assembly of these structural units in the a–b plane followed by growth along the c direction. We suggest that 2D self-assembly precedes growth because the crystal thickness is reasonably constant for entire crystallites and it is very unlikely that domains would be able to grow independently in solution to the same thickness and subsequently meet to form an intergrown crystallite. Furthermore, in the case of M1-trigonal intergrowths, we propose that several “raft-like” clusters of pentagonal groups linked by VO_x and MoO_x units, some with M1-type symmetry and others with trigonal-symmetry, joined to form sheets. Following the formation of these sheets, growth in the c direction is likely to occur by a condensation of pentagonal units that approach the initial 2D self-assembled structure in a favorable orientation. The small bands of either phase are likely the result of trapped/disorganized structural units that order through diffusion (VO_x and MoO_x units) or rotation (pentagonal Mo₆O₂₁ units) to an energetically favorable position during the heat treatment. The formation of crystalline bands of the M1-type phase versus the trigonal phase is likely controlled by the local ratio of pentagonal units to the smaller MoO_x and VO_x units. We believe that a higher concentration of the smaller units would lead to the M1-type phase because the unit cell contains far more linking sites.

An alternative explanation for the formation of an intergrown domain structure could involve the merger of multiple growth fronts. In such a case, a loosely associated cluster of pentagonal Mo₆O₂₁ units with appended MoO_x and VO_x groups could slowly assemble under hydrothermal conditions at various locations within this cluster. These nucleation centers with either trigonal or M1-type symmetry slowly grow from within embryonic clusters until they eventually meet. If like domains improperly align due to relative rotation about an axis, a metastable configuration may result in which the second phase mediates the orientational mismatch.

So far, we have not observed intergrowths of the Mo₅O₁₄-type phase with either the trigonal or the M1-type phases. It is interesting to note that the M1-type phase has been a precursor used for synthesis of the Mo₅O₁₄-type phase, yet the unit cell does not contain features that can be superimposed on top of features from either of the other two phases. The Mo₅O₁₄-type phase also lacks the heptagonal channels found in the M1-type and trigonal phases.

Using aberration-corrected HAADF-STEM imaging, we report the observation and structural characterization of intergrowth phases and defects within the Mo-V-O system. Within a single crystallite, atomic-level imaging clearly shows domains of the orthorhombic M1-type phase intergrown with those of the trigonal phase. All forms of registry require the b axis from the M1-type phase to be perpendicular with the a or b axis from the trigonal phase. Despite the variety of possible combinations of orientations, the atomic-scale registry at the interface is the same. HAADF-STEM images also revealed crystallites with significant disorder containing M1-type precipitates intergrown with narrow bands or clusters of the trigonal phase. From this study we conclude that, although the M1 phase does intergrow with the trigonal phase and may contain significant disorder in some preparations, there is no evidence for intergrowth with the M2, proposed as a promoter for M1 conversions (9, 10), or with the Mo₅O₁₄-type phase.

The Ångstrom-level resolution provided by aberration-corrected STEM imaging has provided unique insights into the phases present in the Mo-V-O oxide system. Combining these resolution enhancements with the Z² compositional sensitivity of the HAADF imaging technique allows characterization of crystalline phases based on both composition and structure. We believe that the identification and characterization of structural intergrowths, compositional miscibility gaps (34), and phase boundaries, as well as nonintergrowth coexisting phases in complex catalyst formulations, provides critical information that can assist in the understanding and continued development of new and improved catalytic materials.

Materials and Methods

Catalyst preparation procedures for each of the phases are described in SI Text.

Aberration-corrected high-resolution STEM operated at an accelerating voltage of 200 kV was used to image the materials with a JEOL 2100F equipped with a Corrected Electron Optical Systems GmBH spherical aberration corrector on the illumination system at the Electron Microscopy Center of the University of South Carolina. The geometrical aberrations were measured and controlled to provide less than a π/4 phase shift of the incoming electron wave over the probe-defining aperture of 15.4 mrad. HAADF-STEM images were acquired on a Fischione Model 3000 HAADF detector with a camera length such that the inner cutoff angle of the detector was at least 75 mrad. The scanning acquisition was synchronized to the 60 Hz ac electrical power to minimize 60 Hz noise in the images and a pixel dwell time range between 7 and 32 μs was selected. Each sample was prepared for STEM by finely grinding the as-prepared catalyst specimen and then dipping a holey-carbon coated Cu grid into the powder.