D. Pravarthana*ab,
O. I. Lebedevb,
A. Davidb,
A. Fouchetb,
M. Trassinc,
G. S. Rohrerd,
P. A. Salvadord and
W. Prellier*b
aCAS Key Laboratory of Magnetic Materials and Devices, Ningbo Institute of Materials Technology and Engineering, Chinese Academy of Sciences, Ningbo 315201, People's Republic of China. E-mail: dpravarthana@gmail.com
bLaboratoire CRISMAT, CNRS UMR 6508, ENSICAEN, Normandie Université, 6 Bd Maréchal Juin, F-14050 Caen Cedex 4, France. E-mail: wilfrid.prellier@ensicaen.fr
cDepartment of Materials, ETH Zurich, Vladimir-Prelog-Weg 4, 8093 Zurich, Switzerland
dDepartment of Materials Science and Engineering, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, Pennsylvania 15213, USA
First published on 25th June 2019
Using the Combinatorial Substrate Epitaxy (CSE) approach, we report the stabilization of Dy2Ti2O7 epitaxial monoclinic, layered-perovskite phase Dy2Ti2O7 thin films. To achieve this, the films are deposited on high density, polished La2Ti2O7 polycrystalline ceramic substrates, which are stable as monoclinic layered-perovskites, and were prepared by conventional sintering. Microstructural analysis using electron backscatter diffraction (EBSD), electron diffraction (ED), and high-resolution transmission electron microscopy (HRTEM) support this observation. Further, they reveal that the cubic pyrochlore phase is observed far from the interface as films are grown thicker (100 nm), confirming the importance of substrate-induced phase and space group selection. This works reinforces the vast potential of CSE to promote the stabilization of metastable phases, thus giving access to new functional oxide materials, across a range of novel material systems including ferroelectrics.
Since the substrates commercially available are often limited in terms of structure, symmetry, lattice parameter, and orientation, we have recently developed an alternative approach, the so-called Combinatorial Substrate Epitaxy (CSE), where a polished ceramic is used as substrate. In that case, each grain of the ceramic can be viewed as a single-substrate with a particular orientation, and one can screen the entire orientation space map in one single experiment.4–9 It enables a high-throughput way to investigate the structure–property relationships as a function of orientation.10,11 This approach can also be used to synthesize metastable phases.12 For example, Ln2Ti2O7 (lanthanide (Ln3+) = Sm3+, Gd3+, Dy3+) compounds normally stabilized in the bulk as pyrochlore, were recently prepared in the 110-layered perovskite phase using similarly structured, [110] layered perovskite Sr2Nb2O7 ceramic substrates.
The [110] layered perovskite system is a homologous series in the family of oxides with the general formula of LnmBmO(3m+2), where m is the number of perovskite units within a single layer, and m = 4 and B = Ti, Nb, Ta for the oxides of interest herein. In this structure, the four distorted perovskite units are stacked along the [110] direction of LnTiO3 perovskite, with an extra (110) perovskite O2 layer inserted between the perovskite slabs. In the [110] layered perovskite structure, the layered direction become the [100] axis (Fig. 1a).13 One important difference between cubic pyrochlore and monoclinic layered perovskite structures is the connectivity of the corner-sharing TiO6 octahedra network. In the monoclinic layered perovskite structure, the TiO6 network is infinitely extended along linear chains that run along the a-axis, and they terminate at the extra O2 layer in the two orthogonal directions of the perovskite sub-cell. In the cubic pyrochlore structure (Fig. 1b), the network is made of zig-zag chains lying along 〈110〉 direction with a Ti–O–Ti angle of 130°.14 This difference between the two polymorphs can lead to different properties such as high temperature ferroelectricity in the monoclinic layered perovskite, which is not found in the cubic pyrochlore structure.15
The two different polymorphs of Ln2Ti2O7 in bulk also depends on the ratio between the radii (r) of Ln3+ and Ti4+ cations.14,15 While the CSE has already been used to stabilize Dy2Ti2O7 in a metastable layered perovskite structure, its symmetry and the mechanism of growth have not been detailed previously.12 For these reasons, we have grown Dy2Ti2O7 (DTO) thin films on monoclinic [110] layered perovskite La2Ti2O7 (LTO) ceramic substrates using Pulsed Laser Deposition (PLD). We investigated their structure and microstructure, and our results are reported in this article. We find that the DTO films crystallize in a monoclinic structure, which is different from the bulk cubic one, with a large number of anti-phase boundaries, and propose a mechanism for its stabilization.
Such substrates (0.4 mm thick) were used to grow DTO thin films by Pulsed Laser Deposition (PLD). Briefly, the polycrystalline DTO target was irradiated by the excimer KrF laser (λ = 248 nm) under an oxygen pressure of 10−4 mbar and the substrate was kept at 700 °C. The frequency used was 1 Hz, and the fluence was approximately 1.5 J cm−2. After deposition, the samples were cooled down to room temperature in the same pressure at a rate of 10 °C min−1.
Structural and microstructural characterization of the ceramics and films were first carried out using electron backscatter diffraction (EBSD). The samples were typically mounted at a 70° tilt angle from horizontal in a scanning electron microscope (FEG-SEM Carl ZEISS SUPRA 55) operated at 20 kV. Gold was evaporated along the edges of the samples to avoid charging effect on the surface during the experiments. Transmission Electron Microscopy (TEM) investigations were also carried out on cross-section and plan-view samples using a FEI Tecnai G2 30 UT microscope operated at 300 kV (point resolution 1.7 Å). The plan-view and cross-section TEM sample were prepared by conventional mechanical polishing to the thickness of ≈20 μm, followed by Ar ion milling. Image simulations were made with CrystalKit and MacTempas software.
To investigate the crystalline structure of the surface, EBSD patterns were recorded on both the DTO film and the LTO substrate (prior to growth). 21 pairs of grains (film/substrate) were thusly analyzed to understand the epitaxial relationship.4,5,7–10,12 The LTO pattern (Fig. 2b) shows intense, sharp and low symmetry features of the bands, confirming the good crystalline quality, and a monoclinic structure. In Fig. 2b, intense bands are crossed with an angle close to 90° on a zone axis identified as [010]LTO, which confirms the presence of two fold (C2) symmetry axis compatible with a P21 space group. The DTO film grown on this grain exhibits a similar EBSD pattern (Fig. 2c), suggesting the same 110-layered perovskite structure with a similar orientation, and also consistent with a monoclinic structure, which latter was confirmed by SAED. The film pattern is slightly more diffuse, as reported for other films.10 Similar observations were made for all 21 grain pairs discussed herein. As discussed previously,12 automatic indexing of patterns from this complicated low-symmetry structure, is challenging with the commercial software, and improved indexing methods are needed, as discussed elsewhere,9 for generating inverse pole figure maps to illustrate grain over grain growth. However, the patterns collected on individual grains were consistent with the substrate patterns throughout individual grains, indicating reasonable grain over grain growth.12
The epitaxial relationship recorded for 21 pairs of grains is plotted in standard stereographic representations of orientations (see the inverse pole figure in Fig. 2d). All the 21 grain pairs between the LTO substrate and the DTO film exhibit similar orientations, confirming a unit-cell over unit-cell growth (a few examples of grain pairs are labeled 1, 2 and 3 in Fig. 2d). A small misorientation angle is observed, ranging from 3° to 6°, which is mainly attributed to misalignment of the grains resulting from the different positioning of the sample between EBSD runs,5 but may have contributions from film relaxation and growth as well. All these observations are consistent with prior observations of DTO growth on Sr2Nb2O7 substrates,12 which is an orthorhombic 110 layered perovskite. Our observations indicate that high-quality, monoclinic substrates of La2Ti2O7 can be prepared and used to grow epitaxial thin films of metastable Dy2Ti2O7.
Besides the clear monoclinic angle (β ∼ 98°) in the [010] zone axis which is different from the orthorhombic one for this structure (β = 90°), the [010]DTO ED pattern displays typical characteristics of twinning along the a-axis [100]DTO.18,20 It should be noted that twinning is due to the presence of a mirror plane perpendicular to the c-axis. Strong (s) and weak (w) diffraction rows alternate along the [100] in Fig. 3a and are labelled s and w, which corresponds to diffraction from the cation sublattice and oxygen sublattice, respectively.20 The strong and weak diffraction correspond to miller indices where h = 2n and h = 2n + 1, respectively, with n being the order of reflection.20 The [010] SAED pattern (Fig. 3b) shows sharp diffractions spots and does not show any diffuse streaks along the c-axis. All SAED patterns recorded along [010] and [100] zone axes exhibit sharp diffraction spots, confirming the high degree of structural ordering, confirming the quality of the film confirming latter by HRTEM measurements.
These observations suggest that the contrast associated with the boundary is determined by the stress filed around misfit dislocations along LTO/DTO interface. Therefore, the boundaries are low misorientation angle grain boundaries (LAGB). It is well established that small difference between substrate and thin film is accommodated by misfit dislocations. In fact, above a critical thickness, it becomes energetically favorable to accommodate stress by misfit dislocations along film/substrate interface.24
The schematic 3D representation of the boundary planes is given in Fig. 3f. In this figure, the dislocations lines forming LAGBs propagate along the [001] and [100] axes. The boundary abruptly stops propagating along [010]-axis in some regions, while in other regions it bends, and connects with neighbor boundaries. To understand the directional features of the boundary and origin of these boundaries, HRTEM images were recorded in cross-sectional view (Fig. 4a) and plan-view (Fig. 4b). As already mentioned, the [100] and [010] ED patterns are well indexed in a P21 space group. Recorded cross-section HRTEM image and corresponding [120] ED patterns again confirmed the structure of DTO film to be monoclinic. The difference in the unit cell parameters is evident from ED spots (Fig. 4a inset upper right corner) where weak spots belong LTO substrate and strong spots DTO film. The DTO/LTO interface is quite sharp and free of any secondary phase or amorphous layers. Further, the boundary of region (indicated with downward tilted white arrow) does not exhibit stacking faults as viewed from interface to surface of the film in vertical direction. In the HRTEM images, the geometry and position of the white dots, which correspond to projections of the atomic columns on the plane, generally remain undisturbed when crossing the boundary region except for a change in contrast. All these data confirm our suggestion that the contrast change results from the strain field of a LAGB and the presence surrounding strain field. The strain region is imaged as relatively dark contrast because of its slightly different orientation from neighboring grains.
Fig. 4 (a) Bright field cross-section [120] HRTEM image of c-axis oriented DTO film over the region in Fig. 3(c) and (e) correspondingly DTO/LTO interface depicted with white arrows heads. The corresponding cross-section SAED pattern is given as inset at left hand corner and indexed based on P21 monoclinic structure (a = 7.816 Å, b = 5.5412 Å, and c = 13.0067 Å with β = 98.698°). The magnified ED pattern is shown in right top corner as inset and indicates the superposition of the ED pattern of LTO substrate (weak spots) and the DTO film resulting in splitting diffraction spots. Noticed presence of misfit dislocation along LTO/DTO interface. (b) Plane-view [100] HRTEM image of selected region in Fig. 3e. The white arrow cap indicates the APBs region. The enlarged HRTEM image of DTO film together with a computer simulated image based on monoclinic structure is given as inset at right hand top corner. |
The [100] plan-view HRTEM images gives more information on the film structure (Fig. 4b). The enlarged HRTEM image and overlaid calculated image based on the determined DTO structure shows a good agreement. However, the positions of atomic planes are displaced in some region at 1/3c along [001], direction suggesting of anti-phase boundaries (APBs) at LAGBs. APBs are common defects in layered oxides and have been extensively studied in related layered oxides such as YBa2Cu3O7−δ, Aurivillius and Ruddlesden–Popper phases.25,26 The observed APBs runs over an average thickness of 6 nm. The origin of APBs may be attributed to terraces on the LTO substrate surface, as this has previously been reported for APB.27–29 The terraces can be better seen in Fig. 3c across the LTO/DTO interface. In the films, the APBs propagate up to the surface of the film. The red tilted arrows in Fig. 3c indicate the terrace in the surface of DTO film. A similar view of a terrace is also given in Fig. 4a. Thus, terraces in the substrate surface can lead to stabilization and propagation of these APBs. Further, APB introduce strain owing to local variations in the structural parameters. We propose that the APBs can be related to a misalignment of the extra oxygen layer separating two DyTiO3 perovskite slabs along adjacent TiO6 octahedra. In a perfect crystal there is a perfect periodic stacking of the extra oxygen layer after every four DyTiO3 perovskite blocks along the c-axis19 but at the APB, the extra oxygen layer terminates, intersecting an adjacent perovskite block.
The inset of Fig. 4b shows a bright field HRTEM image of DTO region. According to the simulated image, two dark spots of different sizes, which correspond to Dy (larger dark spot) and Ti (smaller dark spot) can be seen. The interlayer shear boundary is seen as a zig-zag band (Dy–O linkage) between the layers. Again, the simulated image using the monoclinic DTO structure (see its superimposition in the inset) fits well with the experimental data. The presence of internal references from the LTO substrate in the SAED pattern (see the inset of Fig. 3d and 4a) enables one measure of the lattice parameters of the DTO film with a high accuracy a = 7.53 Å, b = 5.31 Å, and c = 13.03 Å and α, β, γ = 90°, 98.65°, 90°. When compared to the LTO substrate, the calculated a and b lattice parameters of DTO film is smaller by 3.6% and 4.1%, respectively due to the chemical difference in cation size of La3+ = 103.2 pm and Dy3+ = 91.2 pm. The monoclinic angle β, and the calculated c lattice parameters are less perturbed. The DTO film lattice accommodates the smaller cation Dy3+ in comparison to La3+ by tilts and/or rotations of the TiO6 octahedra, resulting in slightly different lattice parameters from LTO, and an accommodation via misfit strains with the observed LAGBs/APBs.30,31
In order to shed more light on the strain in the DTO film, Geometric Phase Analysis (GPA) was performed on the Fourier transform (FT) of the filtered HRTEM image (Fig. 5a). The corresponding FT pattern is shown in Fig. 5b; it displays diffuse Bragg intensity, which generally occurs due to the presence of strain and defects in the crystal structure.30 The obtained color-coded GPA strain map (Fig. 5c) of the in-plane [001] lattice parameter shows uniform color when moved from LTO substrate to the film, which confirms the epitaxial growth of DTO with a small strain state and identical lattice d-spacing. The small color variation at the interface of LTO/DTO is due to strain at the interface. The APBs in this region are indicated with vertical white arrows, and the region between APBs appears in bright in color. This indicates that APBs are highly strained in the (ac) plane because they have shorter distances between the adjusted atomic columns. In the GPA strain map, recorded in the out-of-plane direction (Fig. 5d), the film and substrate have large lattice difference resulting in splitting of diffraction spots compared to the in-plane lattice parameter, which are almost equal (see Fig. 5b the corresponding FT pattern). This GPA analysis gives clear indication that DTO film is under compressive strain. The APBs appear yellow, which indicates less strain in comparison to the strain of in-plane lattice parameter in the boundary plane. This indicates that the APBs are strained along the c-axis, and relaxed along a and b axis. Conversely, the crystal region adjacent to LAGBs are relaxed along the c-axis, and strained along the a and b axes.
Fig. 5 (a) Fourier transform (FT) of the filtered HRTEM image (Fig. 4a) and corresponding (b) FT pattern. The colour coded GPA map shows the strain variation along (c) in-plane [001] and (d) out-of-plane [20] directions with reference to LTO substrate. In out-of-plane GPA map the yellow contrast boundaries are clearly visible. It should be noticed that position of this boundaries corresponds to dark region contrasts in corresponding HRTEM image (marked with white arrows). DTO/LTO interface depict with white arrow heads. |
Depending on the orientation and substrate surface state, after some thickness, the cubic DTO forms (Fig. 6b). In Fig. 6b, both surfaces of the substrate and the film appear wavy which is attributed to the roughness of the substrate, due to the polishing and chemical etching process used. It should be noted that the film may grow epitaxially with the monoclinic DTO structure along the c-axis, but small areas of cubic structure may grow incoherently at the surface. The color coded GPA strain map (Fig. 6d) taken along the in-plane shows uniformity of colors when moved from LTO substrate to the film, which confirms epitaxial growth and the presence of epitaxial strain in the film. A sharp variation at the interface (dark red) could be observed likely arising from the roughness of the substrate–film interface. The strain variation in the region of marked cubic DTO phase is seen. In the out-of-plane GPA strain map of lattice parameter (Fig. 6e), the film exhibits uniform color that indicates uniform undeformed d-spacing over this region. In this image, the film is more strained compared to the in-plane lattice image, confirming that the film relaxes along the c-axis. Moreover, there are not APBs unlike DTO film grown on smoother LTO substrates as shown in Fig. 3c.
This journal is © The Royal Society of Chemistry 2019 |