Letter pubs.acs.org/NanoLett
Controlled Growth of Ceria Nanoarrays on Anatase Titania Powder: A Bottom-up Physical Picture Hyun You Kim,† Mark S. Hybertsen,*,‡ and Ping Liu*,‡ †
Department of Materials Science and Engineering, Chungnam National University, 99-Daehakro, Daejeon 34134, Republic of Korea Center for Functional Nanomaterials, Brookhaven National Laboratory, Upton, New York 11973, United States
‡
S Supporting Information *
ABSTRACT: The leading edge of catalysis research motivates physical understanding of the growth of nanoscale oxide structures on different supporting oxide materials that are themselves also nanostructured. This research opens up for consideration a diverse range of facets on the support material, versus the single facet typically involved in wide-area growth of thin films. Here, we study the growth of ceria nanoarchitectures on practical anatase titania powders as a showcase inspired by recent experiments. Density functional theory (DFT)-based methods are employed to characterize and rationalize the broad array of low energy nanostructures that emerge. Using a bottomup approach, we are able to identify and characterize the underlying mechanisms for the facet-dependent growth of various ceria motifs on anatase titania based on formation energy. These motifs include 0D clusters, 1D chains, 2D plates, and 3D nanoparticles. The ceria growth mode and morphology are determined by the interplay of several factors including the role of the common cation valence, the interface template effect for different facets of the anatase support, enhanced ionic binding for more compact ceria motifs, and the local structural flexibility of oxygen ions in bridging the interface between anatase and ceria structures. KEYWORDS: Oxide interfaces, nanostructure growth, epitaxial growth, formation energy, density functional theory calculations, ceria, titania
M
model catalyst structures, the role of the CeOx NPs has been extensively studied: they affect the dispersion of the metal NPs, the electronic structure of the metal NPs, and the slow kinetics of reaction steps on metal sites.29,31−33 How do CeOx nanostructures form when depositing Ce on anatase titania nanostructured supports in an oxygen atmosphere? High-resolution transmission electron microscopy (HRTEM) studies revealed diverse CeOx nanostructures on several facets of A-TiO2 crystallites in the powder: 0D clusters, 1D chains, 2D plates, to 3D NPs.30 Such a distribution of nanostructure motifs was not observed previously for CeOx nanoparticles synthesized directly using a hydrothermal method, where substrate template effects were absent.34 It also substantially differed from the growth mode on model systems using either metal surfaces35,36 or model oxide surfaces29,33,37 as substrate, including rutile TiO2(110). Furthermore, oxide size and shape can have a significant effect on catalytic performance.38,39 Thus, for practical catalysts on a powder support, such as the CeOx/TiO2 system, the fundamental structure−function understanding must be investigated directly. This starts with insight into the growth mechanisms and the factors that control structures. This is both
aterial interfaces are ubiquitous. Optimizing the physicochemical properties of interfaces continues to be one of the primary tasks in rational design of functional materials.1−4 For example, utilizing the interplay between nanostructured materials and the interface with the substrate to form highly effective catalysts is just emerging.5−7 Understanding the resultant structure−property relationships and developing methods to form diverse heterointerfaces on demand promise great potential for tuning chemical and physical properties.8−13 Further progress will rely on systematic strategies for rational control of the growth and evolution of interfaces. This opens the broader potential for impact from breakthroughs in the growth of heterostructures.1,14−16 In recent catalysis research, high activity for a variety of catalytic processes (e.g., CO oxidation, oxygen evolution, and methane activation) was achieved by controlling supported catalyst structure motifs at the nanoscale.17−28 In an exciting example, catalysts have been developed from ceria (CeOx) nanoparticles (NPs) grown on a single crystal rutile-TiO2(110) model surface,8,29 or more practically on anatase titania (ATiO2) powders,30 with the addition of metal NPs. These catalysts, acting as an integrated system, have shown significant promotion of the water−gas shift (WGS) reaction, photocatalytic water dissociation and CO2 activation reactions in comparison to the combination of the metal NPs with each individual oxide as support separately.6,8,29,31−33 Utilizing © XXXX American Chemical Society
Received: October 7, 2016 Revised: November 23, 2016 Published: December 5, 2016 A
DOI: 10.1021/acs.nanolett.6b04218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 1. Interface formation of cubic-fluorite CeO2 with anatase TiO2. (a,b) Wulff construction and the morphology of the specific facets involved in the interface formation of anatase TiO2 and cubic-fluorite CeO2. (c) Schematic illustration of the surface dependent formation of CeO2 architectures on TiO2(101) and (112) facets, as reported in ref 30. (d) The zigzag oxygen rows of TiO2(112) that interlock with a corresponding surface motif of CeO2(100) to form a coherent interface structure.
of the fluorite CeO2(100) facet, the distorted oxygen rows interlock and the stacking of nominally neutral layers can be seen to be continuous from O−Ti−O to O−Ce−O trilayer motifs. Similar logic, based on Tasker’s classification scheme,42 leads to the same interfacial structure. This was used in a DFT study directed to understand the relative stability of oxygen vacancies at the epitaxial interface, including the impact of cation intermixing.30 Overall, we seek to understand the driving forces that control these diverse expressions of nanostructure epitaxial growth in this exemplary complex oxide system. Our approach is to explore the development of nanostructures from the bottom-up, using the calculated total energy and relaxed structures from DFT-based calculations as a guide. Total energies and forces are computed with the gradientcorrected functional PW91,43 supplemented with the DFT+U approach for the Ti and Ce ions,44 using the projector augmented wave method45 as implemented in the VASP code.46,47 The details are documented in the Supporting Information.8,43−50 The formation energies reported here are relative to gas phase species. To model the experimental growth procedure, atomic Ce is deposited in the presence of oxygen. Figure 2a illustrates our bottom-up method to model the experimental growth procedure, by alternating Ce deposition and oxygen adsorption at the Ce sites. When a Ce atom is deposited, the most stable position is selected. For instance, on A-TiO2(101), a Ce atom prefers to bridge two surface oxygen atoms. The adsorbed Ce also generates two Ti3+ sites (Figure 2a), which act as active centers for the adsorption (Figure 2b) and dissociation of O2 (Figure 2c). This leads to the formation of a compact CeO2 motif that remains coordinated to the two surface bridging oxygen atoms. Introduction of a second Ce atom in the supercell (Figure 2d) followed by a third O atom gives a structure that suggests the formation of a compact chain (Figure 2e). However, with a fourth O atom, the lowest energy structure results in the formation of a second CeO2 motif that is essentially identical to the first one (Figure 2f). This is the early stage in the formation of chain structures based on CeO2 moieties; by comparison, other local organizations are modestly higher in energy (Figure S1). As illustrated in Figures S2 and S3, this procedure systematically extends to explore diverse routes to assemble nanostructures.
a challenge and an opportunity to explore growth beyond the traditional scope of thin film epitaxy, in the diversity of both the morphology of epitaxial structures and the facets of the support. In this Letter, we use density functional theory (DFT) to perform a bottom-up study of the origin of diverse growth modes for CeOx nanoarchitectures on two facets of A-TiO2: the low energy (101) facet and the (112) facet. We find that detailed energetic considerations can rationalize the growth mechanism and the specific nanostructures that are observed. The exposed TiO2 facet plays the essential role of template during this process. The surprising emergence of regular, 3D fluorite CeO2 nanostructures selectively on the higher surface energy (112) facet points to the complex factors at play at this interface. These include an abrupt change in lattice structure (although the two oxides share a common 4+ cation valence), strain induced by alignment of the ceria motifs to the facet template, the tendency for ionic bonding to drive compact motifs, and the structural flexibility of interfacial oxygen ions from both oxides. Our results pave the way toward the rational design of practically relevant nanomaterials with enhanced stability and functionality. As illustrated by the Wulff construction in Figure 1a, the lowest surface energy (101) facets dominate the A-TiO2 nanocrystal surface.40 Turning to CeO2, its cubic fluorite crystal structure is quite distinct (Figure 1b) and its low energy (111) facet offers no obvious match in terms of lattice structure to the A-TiO2(101) template. Interestingly, clear trends emerged from the experiments,30 as summarized in Figure 1c. On what was expected to be the most stable (101) facets, CeOx clusters, chains, and plates were observed. Similar structures were also observed on (100) facets, with details depending on the amount of CeOx deposited. Strikingly, well-ordered CeOx NPs were uniquely formed on (112) facets, which were less stable and smaller in surface area than that of (101) facets (Figure 1a and Table S1). HRTEM measurements exhibited epitaxial interfaces between fluorite CeO2(100) facets and ATiO2(112) that were atomically ordered, but with some Ti/Ce interdiffusion.30 For these higher energy and polar facets, there is a route to a near-lattice matched epitaxial interface, as illustrated in Figure 1d. Consider that each bare surface is stabilized by a reconstruction with half of a monolayer (ML) of oxygen removed surface oxygen.41 Then upon 45 deg rotation B
DOI: 10.1021/acs.nanolett.6b04218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 2. Illustration of bottom-up methodology exploring CeO2 growth. Sequence modeling the early steps of Ce deposition on ATiO2(101) in an oxygen atmosphere. (a) First Ce deposition. Black dots mark Ti3+ sites. (b) First O2 adsorption. (c) First O2 dissociation. (d) Second Ce deposition. (e,f) Second O2 dissociative adsorption in two steps: the impact of one additional O atom (e), and then the lowest energy structure in the presence of both additional O atoms (f).
Figure 3. Growth of CeO2 architectures on anatase TiO2(101). (a) Energy of formation (Eform) as a function of number of CeO2 clusters incorporated into finite CeO2 structures or CeO2 coverage for periodic arrays. (b) Morphology of selected, representative CeO2 architectures on TiO2(101) facet.
coverage. The fundamental motif is a periodic structure consisting of four CeO2 units (Figure 3b). This unusual, vertically modulated chain structure displays higher density along the chain than that of the diagonal chain, which enhances the ionic interactions through the formation of extra Ce−O bonds between neighboring CeO2 monomers. It is thermodynamically preferred over the more loosely organized diagonal type of chain. The interchain interactions between the infinite [101̅] chains are relatively weak. As a result, with increasing CeO2 coverage, packing the isolated chains to form a 2D plate with 1 ML of coverage does not gain much energy (Figure 3a). Nonetheless, at sufficiently high coverage, the epitaxial 2D CeO2 overlayer consisting of [101̅] chains should naturally emerge on A-TiO2(101). By comparison, the pathway to assemble the 2D finite plate from the diagonal chains is hindered by severe distortions upon formation (Figure S5), resulting in lower stability. None of the chain and plate-like structures that emerge from our calculations for A-TiO2(101) appear to serve as an appropriate template for the NP growth with bulk-like fluorite structure of CeO2. Following our bottom-up methodology, two CeO2 NPs are built up on A-TiO2(101) through successive addition of CeO2 units (Ce10O20, Figure 3a) and direct deposition of a bulk-like NP (Ce14O28, Figure 3b and Figure S6), which shows high stability on A-TiO2(112) (see below). However, they are either energetically less favorable than the corresponding [101̅] chain structure at similar average coverage of CeO2, or completely collapse on A-TiO2(101). By following the formation energy, we can arrive at monomers and finite chains by limiting the amount of CeO2 deposited. With the coverage increasing, the biclusters, the [101̅] chains and finally arrays of those chains form plates as the preferred nanostructures of CeO2 deposited on the ATiO2(101). This agrees with the HRTEM observations over all, specifically regarding the chain orientation and the lack of stable 3D NPs.30 The stabilization of the dominant [101̅] chain motif is a combination of the strong template effect following the atomic arrangement of the TiO2 support and the enhanced ionic bonding achieved in this buckled chain configuration. The key factor preventing 3D NP growth is the mismatch between bulk TiO2 and bulk CeO2 at this interface (Figure 1). This induces severe structural disordering upon NP formation and
A sampling of the most significant nanostructures that emerge at varying CeO2 coverage is shown below for the ATiO2 (101) and (112) facets. As noted, the fundamental unit is the CeO2 monomer, which is strongly bound to the surface for both facets. Energetically, each adsorbed CeO2 monomer is likely to act as a nucleation site and drive the sequential growth of various CeO2 patterns. In the present modeling, we will not consider kinetics, but focus specifically on the relative stability of competing nanostructures. For all the structures studied, the Ce atoms are fully oxidized as Ce4+, indicating that bonding at the interface is predominately ionic (Figure S4). Motifs with an odd number of O atoms are always found to be less stable than a corresponding stoichiometric motif. In the experiments, some reduced Ce3+ was observed, but largely attributed to oxygen vacancy formation facilitated by Ce/Ti ion intermixed at the interface with the CeOx nanoparticles,30 an effect beyond the scope of the present study. Yet, as shown in the following, our current focus on otherwise defect-free structures is appropriate to understand the main forces that control the growth of different nanostructures in this system. We first consider the low energy A-TiO2(101) facet. The growth of CeO2 starts with a single CeO2 cluster. There is a very negative Eform (−4.67 eV/CeO2) due to a strong and ionic interaction (Figure 3a). Once formed, the CeO2 monomer is immobile (diffusion barrier of 2.75 eV), and the structure evolution is controlled by Ce deposition and oxidation, rather than ripening following exposure. In the presence of an additional single CeO2 cluster next to the monomer, a Ce2O4 cluster is formed (Figure 2); there is a small electrostatic attraction (0.14 eV/CeO2, effectively quadrapolar) for a neighboring site in the next row. With the coverage of CeO2 increasing to 1/4 ML, this interaction eventually favors the formation of a diagonal infinite chain, which is 69.3° off from the [010] direction of A-TiO2 (Figure 3a). However, this is not the most stable structure for chains at high coverage of CeO2. With the size of CeO2 clusters increasing to Ce4O8, a 1D finite chain oriented along the [101̅] direction is more favorable than the corresponding diagonal chain ([101̅] chain, Figure 3b and Figures S2). With the coverage increasing longer single chains form, here shown by infinite, separated chains at 1/2 ML C
DOI: 10.1021/acs.nanolett.6b04218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
significant energy gain is observed due to monomer interactions. For a Ce2O4 cluster, two diagonally arranged CeO2 monomers share one oxygen atom (ΔEform= −0.63 eV/ CeO2, VC2 in Figure 4). From this nucleus, two distinct, extended structures can be formed. Expansion of the Ce2O4 cluster along the diagonal direction across the oxygen rows results in the formation of infinite diagonal chains at 1/3 ML of coverage (Figure S3a), although the local rearrangement costs energy relative to the root VC2 cluster (Figure 4a). These chains do have an attractive interaction between them, leading to relative stability for epitaxial plates up to 1 ML (Figure S3a). Alternatively, the Ce2O4 cluster nucleus can be expanded along the vertical direction by filling the gaps between the two nearby oxygen rows (Figures 4a and S3b). This leads to the formation of a series of progressively more stable vertical chains (VCn) and ultimately infinite vertical chains, as illustrated at 1/3 ML (Figure S3b). However, VCs have a mildly repulsive interchain interaction. While an epitaxial monolayer can be formed, it is relatively less stable. This monolayer structure also extends to thin films at higher coverage (Figure S7). This fundamental zigzag CeO2 motif naturally maximizes the interaction between the zigzag oxygen rows on A-TiO2(112) and the added CeO2 as well as that between the two neighboring CeO2 units. As a result, the VC structure is more stable than the diagonal chain at the same coverage (Figures S3a and S3b). Since formation of a full monolayer is not the most stable structure formed, we explore alternative routes to form 2D plates. In particular, square and rectangular fragments of the full ML structure, oriented with their sides along the diagonal, can be formed by association of a set of finite VCs and monomers (Figure S3c). Due to mild repulsion, they are less stable than VCs, e.g., formation of a Ce6O12 plate out of two VC3 costs 0.13 eV/CeO2 monomer. However, such a diagonal plate naturally forms a perfect template for the assembly of 3D NPs with low energy (111) facets, as shown in Figures 4b and S3d. The smallest NP is built on a diagonal Ce4O8 plate by adding the fifth CeO2 unit in the next layer, which significantly lowers the energy and forms a square base 3D NP. Similarly, addition of two CeO2 monomers to the Ce6O12 plate forms a rectangle base 3D NP. However, at this size scale, the NPs are not as low in energy as the VCs (Figure 4a). With increasing coverage, the 3D NPs become more energetically favorable. Starting from the rectangular Ce40O80 NP, the 3D NPs show the lowest energy among all growth modes studied on A-TiO2(112). Taken together, the results suggest that with the increasing exposure of CeO2 on A-TiO2(112), the growth starts with the biclusters, then 1D finite vertical chains, and is eventually taken over by 3D NPs on reaching a critical size that may be as small as Ce20O40, based on Figure 4a (but likely depending in detail on the trends for the finite VCn chains longer than VC5). Overall, this agrees well with the HRTEM study.30 By comparison of the CeO2−CeO2 interaction strength measured by the bulk cohesive energy to the smaller CeO2−TiO2 binding energy, the Volmer−Weber regime applies to the NP growth.51 To get more insight, we analyzed a simplified model of the epitaxial NP energy. See the SI text, Table S2 and Figures S8 and S9 for details.49,50 The key factor that destabilizes the plate structures and inhibits thin film growth is the relatively high formation energy of the CeO2(100) facet. The basal plane structure is not stable enough to allow a layer-by-layer growth, but addition of CeO2 in the next layers fundamentally stabilizes the growing NPs as the much lower energy CeO2(111) facets emerge. The volume strain energy is the least important term
therefore destabilizes CeO2 NPs in bulk-like structures (see Figures S5 and S6). The structural incompatibility at the TiO2− CeO2 interface for this facet inhibits the formation of thicker films or the emergence of 3D NPs. We now turn to the higher energy A-TiO2(112) surface (Figure 4). A CeO2 monomer adopts a similar configuration to
Figure 4. Growth of CeO2 architectures on anatase TiO2(112). (a) Energy of formation (Eform) as a function of number of CeO2 clusters incorporated in finite CeO2 structures and coverage for periodic arrays. (b) Morphology of selected representative CeO2 architectures on TiO2(112).
that described for A-TiO2(101), but with a slight twist in structure induced by the zigzag oxygen chains protruding on the surface. It has an even stronger binding energy (−5.23 eV/ CeO2). In an environment where CeO2 is growing on the facets, this strong preference of CeO2 for A-TiO2(112) relative to A-TiO2(101) (ΔEform = −0.56 eV/CeO2) can overcome the difference in the intrinsic surface energy between A-TiO2(112) and A-TiO2(101) (Table S1). According to our calculations, for 0.17 ML of CeO2 the total energy crosses over to favor the (112) facet. When going beyond CeO2 monomers, a very different picture emerges compared to the case of A-TiO2(101). A D
DOI: 10.1021/acs.nanolett.6b04218 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 5. Role of flexible interface oxygen in the interface adhesion of CeO2 on anatase TiO2(112) and subsequent NP formation. (a) TiO2(112) template effect which drives interface and surface oxygen rows of CeO2(100) into zigzag arrangement. (b) Subsequent restoration of straight oxygen rows during subsequent growth due to the strong CeO2−CeO2 interaction. (c) Energy cost associated with the oxygen displacement on the surfaces of CeO2(100) and TiO2(112).
oxygen rows in the interior straighten out. As shown in Figure 5a,b, the specific distortion involved has a very low energy per unit area. It also brings out another key factor for high interface adhesion: the energy cost to distort the oxygen by a small amplitude (
