High-Quality Heteroepitaxial Growth of Thin Films of the Perovskite Oxynitride CaTaO2N: Importance of Interfacial Symmetry Matching between Films and Substrates.

Perovskite oxynitrides have been studied with regard to their visible light-driven photocatalytic activity and novel electronic functionalities. The assessment of the intrinsic physical and/or electrochemical properties of oxynitrides requires the epitaxial growth of single-crystalline films. However, the heteroepitaxy of perovskite oxynitrides has not yet matured compared to the progress realized in work with perovskite oxides. Herein, we report the heteroepitaxial growth of CaTaO2N thin films with (100)pc, (110)pc, and (111)pc crystallographic surface orientations (where the subscript pc denotes a pseudocubic cell) on SrTiO3 substrates using reactive radio frequency magnetron sputtering, along with investigations of crystallinity and surface morphology. Irrespective of surface orientation, stoichiometric CaTaO2N epitaxial thin films were grown coherently on SrTiO3 substrates and showed clear step and terrace surfaces in the case of low values of film thickness of approximately 20 nm. A (110)pc-oriented film was also more highly crystalline than (100)pc- and (111)pc-oriented specimens. This relationship between crystallinity and surface orientation is ascribed to the number of inequivalent in-plane rotational domains, which stems from the symmetry mismatch between the orthorhombic CaTaO2N and cubic SrTiO3. A CaTaO2N thin film grown on a lattice- and symmetry-matched orthorhombic DyScO3 substrate exhibited a significant crystallinity and a clear step and terrace surface even though the film was thick (∼190 nm). These results are expected to assist in developing the heteroepitaxial growth of high-quality perovskite oxynitride thin films.

Introduction

During the last decade, perovskite oxynitrides were studied with regard to their novel optical, electrical, and electrochemical characteristics, including visible light absorption,[1−3] promotion of photocatalytic solar water splitting,[2−4] colossal magnetoresistance,[5] high dielectric constants,[6] and ferroelectric properties.[7−9] These compounds are typically synthesized via the partial nitridation of oxide precursors under a high-temperature ammonia flow (i.e., by thermal ammonolysis). As a consequence, the resulting oxynitrides are often obtained as low-density fine powders, which prevent detailed characterization of their functional properties. As an example, it is difficult to evaluate the intrinsic electrical properties of these materials, such as carrier mobility, dielectric constant, and photocarrier diffusion length, because of the presence of grain boundaries. In addition, although surface crystallographic orientation is known to be an important factor affecting the activity of photocatalysts/photoelectrodes,[10−14] there is very little control over surface orientation in powder samples of oxynitrides.

The heteroepitaxial growth of oxynitride thin films on lattice-matched single-crystalline substrates is a promising synthetic strategy for overcoming the problems noted above associated with conventional thermal ammonolysis. This is because the single-crystalline epitaxial oxynitride films produced in this manner have fewer grain boundaries and well-controlled crystallographic surface orientations. Indeed, the intrinsic electrical transport and dielectric properties of various oxynitrides, including perovskites such as ATaO2N (A = Ca, Sr, or Ba),[7,15−17] SrNbO3–N,[18] LaTiO2N,[19] LaVO3–N,[20] TaON,[21] and NbON,[22] have been assessed using epitaxial thin films. It has also been reported that the photocurrents generated during water oxidation using photoanodes made of LaTiO3–N epitaxial thin films are affected by the surface orientations of the films,[23] demonstrating that the control of this orientation is of practical importance when considering oxynitrides for photocatalytic applications.

Despite its benefits, heteroepitaxial growth has been applied to perovskite oxynitrides very rarely as compared to the use of this technique with perovskite oxides. This difference is primarily due to the difficulty in controlling the N content of oxynitrides. As an example, it has been reported that the N concentrations of LaTiO3–N thin films with different surface orientations were not the same.[23] However, recent advances in thin-film growth technology, such as the development of nitrogen plasma-assisted pulsed laser deposition or reactive sputtering, have enabled the control of N content through the precise adjustment of the N supply during film growth.[17,24] Another technical issue associated with the heteroepitaxial growth of perovskite oxynitrides is the large lattice mismatch of these materials with commercially available single-crystalline oxide substrates. Specifically, perovskite oxynitrides generally have larger lattice constants (typically apc > ∼3.95 Å, where the subscript pc denotes a pseudocubic cell) compared to oxides (Supporting Information Figure S1) as a result of the larger ionic radius of N3– relative to that of O2– as well as a larger negative charge of N3–, which tends to require larger cations with high valences. In addition, there have been very few reports of the formation of atomically flat epitaxial thin films of perovskite oxynitrides with clear step and terrace structures,[16,17] which are desirable when studying surface/interface chemistry or fabricating heterostructures.

The present work focused on the heteroepitaxial growth of perovskite CaTaO2N thin films with various surface orientations. This compound was selected because it has an orthorhombic structure belonging to the Pbnm space group, which is the most commonly encountered structure in perovskite oxynitrides (Supporting Information Figure S1). In addition, the lattice constants of CaTaO2N (ao = 5.547 Å, bo = 5.624 Å, co = 7.895 Å, where the subscript o denotes an orthorhombic cell, apc = 3.949 Å)[25] are relatively well matched to those of commercially available oxide single crystals (Supporting Information Figure S1). In fact, the epitaxial growth of a (100)pc-oriented CaTaO2N thin film with good crystallinity and an atomically flat surface has been realized on SrTiO3 (cubic, a = 0.3905 nm, lattice mismatch = −1.1% with a pseudocubic approximation).[17] Last, CaTaO2N is a promising visible-light-active photocatalyst capable of promoting overall water splitting.[26]

In this study, stoichiometric CaTaO2N epitaxial thin films were fabricated on the (100), (110), and (111) planes of SrTiO3 by radio frequency (RF) reactive magnetron sputtering. While 20 nm thick CaTaO2N thin films exhibited high crystallinity and atomically flat surfaces irrespective of the surface orientation, the crystallinity and surface morphology were both degraded as the film thickness was increased to 190 nm. This degradation was more significant in (100)pc- and (111)pc-oriented thin films than in (110)pc-oriented specimens because of the formation of multiple domains arising from a symmetry mismatch between the orthorhombic CaTaO2N thin films and the cubic SrTiO3 substrates. Based on these results, we successfully synthesized a thick (100)pc-oriented CaTaO2N thin film with a high degree of crystallinity as well as a clear step and terrace surface on a symmetry- and lattice-matched orthorhombic DyScO3 substrate.

Results and Discussion

In initial trials, the N partial pressure, PN was optimized so as to obtain stoichiometric CaTaO2N thin films, and Figure a shows the θ–2θ X-ray diffraction (XRD) patterns obtained from a series of CaTaON thin films grown on SrTiO3(100) substrates under various PN. Each of these patterns confirms the epitaxial growth of (100)pc-oriented perovskite in a cube-on-cube manner, while the N/Ta ratios vary from 0 to 1 depending on PN that was applied (Figure b). The perovskite structure was evidently maintained even in the case of the film containing no N (PN = 0), possibly because of the formation of cation vacancies.[28] Notably, the N/Ta ratio increased with increases in PN and plateaued at approximately 1 (i.e., at an almost stoichiometric value) at PN ≥ 0.075 Pa. This result suggests that the N concentration was self-limited by crystallization into the perovskite structure.

Figure 1

(a) θ–2θ XRD patterns of CaTaON thin films grown on SrTiO3(100) substrates under various PN (asterisks denote diffraction peaks from the substrate). (b) N/Ta ratios, (c) fwhms of the 200 diffraction rocking curves, (d) unit cell volumes based on a pseudocubic approximation, and (e) rms roughness, rrms, values of films plotted against PN.

The cell volume per formula unit, Vpc, of the CaTaON thin films increased with increases in PN to a maximum of 0.0610 nm3 at PN = 0.075 Pa (Figure c). This Vpc value is slightly smaller than that of bulk CaTaO2N (0.0616 nm3)[25] but is in good agreement with the value for biaxially strained CaTaO2N as calculated assuming a Poisson’s ratio of 0.30 (a typical value for oxynitrides[29]). Further increases in PN caused reductions in Vpc, suggesting the formation of defects. The crystallinity and surface morphology of each CaTaON thin film were evaluated based on the XRD rocking curve and atomic force microscopy (AFM) measurements, respectively. Figure d,e plots the relationships between PN and the full width at half-maximum (fwhm) of the rocking curve for the 200pc diffraction (Δω002) and the root-mean-square (rms) roughness of the CaTaON thin film, respectively. Both values decreased (i.e., crystallinity and surface roughness were improved) as PN was increased from 0 to 0.075 Pa, and reached a minimum at PN = 0.075 Pa. We speculate that the degradation of crystallinity and volume reduction at PN higher than 0.075 Pa originate from slight off-stoichiometry such as excess nitrogen, which could not be detected by the compositional analysis. The minimum value of Δω002 (0.013°) was smaller than that reported for a CaTaO2N thin film grown by nitrogen plasma-assisted pulsed laser deposition (Δω002 = 0.02°).[17] On the basis of these results, we concluded that a PN value of 0.075 Pa is optimal when fabricating stoichiometric CaTaO2N thin films with high crystallinity and smooth surfaces. We hereafter refer to the films grown under this optimized pressure simply as CaTaO2N.

Figure a presents the θ–2θ XRD patterns of 20 nm thick CaTaO2N thin films grown on the (100), (110), and (111) planes of SrTiO3 substrates, while Figure b–d shows the reciprocal space maps (RSMs) of these films. Both the XRD patterns and RSMs confirm the epitaxial growth of CaTaO2N thin films with the same orientations (in a pseudocubic approximation) as the substrate surfaces. The RSMs also indicate that the in-plane lattice constants of the CaTaO2N were locked to those of the SrTiO3 substrates (i.e., coherent growth occurred). Furthermore, atomically flat surfaces with step and terrace structures were observed in these CaTaO2N thin films (Figure e–g), for which the step heights were in good agreement with the interplanar distances of CaTaO2N with the corresponding crystallographic orientations (Figure h–j).

Figure 2

(a) θ–2θ XRD patterns of 20 nm thick CaTaO2N thin films grown on SrTiO3(100), (110), and (111) substrates (asterisks denote diffraction peaks from the substrates). (b–d) XRD RSMs, (e–g) AFM images, and (h–j) cross-sectional line profiles obtained from AFM images along the white lines in (e–g) for the films grown on SrTiO3 (b,e,h) (100), (c,f,i) (110), and (d,g,j) (111) substrates.

Comparing the RSMs of the CaTaO2N films having different crystallographic orientations (Figure b–d), the films grown on the SrTiO3(100) (the 301pc diffraction) and the (111) (the 312pc diffraction) substrates generated broader diffraction spots along the in-plane direction than that on the SrTiO3(110) substrate (the 221pc diffraction). Therefore, the (110)pc-oriented film was more crystalline than the films with other orientations. This tendency became more evident in the case of the thicker CaTaO2N thin films presented in Figure a–c. That is, the diffraction spots of the films grown on the (100) and (111) substrates (Figure a,c and Supporting Information Figure S2a,c) were much broader than those of 20 nm thick films (Figure b,d). In contrast, the (110)pc diffraction spot remained sharp even for the 190 nm thick film (Figure b and Supporting Information Figure S2b), although the step and terrace surface structure disappeared (Figure d–f). Because these films all had the same N/Ta ratios within the experimental error (Supporting Information Table S1), we attribute this evident variation in crystallinity not to deviations from stoichiometry but rather to reflected symmetry-matching, as discussed below.

Figure 3

(a–c) XRD RSMs and (d–f) AFM images of 190 nm thick CaTaO2N thin films grown on SrTiO3 (a,d) (100), (b,e) (110), and (c,f) (111) substrates.

We subsequently focused on the symmetry mismatch between CaTaO2N (orthorhombic) and SrTiO3 (cubic) as a factor that could degrade the crystallinity.[30−33] Note that when analyzing the structure and orientation of the orthorhombic CaTaO2N cell, we hereafter use orthorhombic notation instead of a pseudocubic approximation. Superstructure XRD peaks (i.e., half-order Bragg peaks), which are characteristic of an orthorhombic perovskite structure,[34,35] were observed for all the CaTaO2N films regardless of their surface orientation. An analysis of the out-of-plane lattice constants and the superstructure peaks (Supporting Information Note) demonstrated that each film consisted of orthorhombic cells with a single out-of-plane orientation: (110)oCTON//(100)STO, (010)oCTON//(110)STO, and (011)oCTON//(111)STO (where the superscripts CTON and STO denote CaTaO2N and SrTiO3, respectively). However, the in-plane orientation of the orthorhombic cells was not unique except in the case of the CaTaO2N film on the SrTiO3(110) substrate. The (100)pc-oriented [i.e., (110)o-oriented] CaTaO2N film on SrTiO3(100) evidently contained fourfold rotational domains for which the azimuthal axes differed by 90° along with the epitaxial relationship [001]oCTON//[010]STO, [001]STO (Figure a and Supporting Information Figure S5) because of fourfold symmetry of SrTiO3(100). This fourfold rotational domain structure has been previously reported for epitaxial thin films of orthorhombic SrRuO3 on SrTiO3(100) substrates.[30,31] Similarly, the (111)pc-oriented [i.e., (011)o-oriented] CaTaO2N thin films comprised threefold rotational domains in which the azimuthal axes were rotated by 120° with the epitaxial relationship [0-11]oCTON//[−211]STO (Figure c and Supporting Information Figure S7), reflecting the threefold symmetry of SrTiO3 (111). Interestingly, the (110)pc-oriented [i.e., (010)o-oriented] CaTaO2N thin film on the SrTiO3 (110) consisted only of a single domain with the epitaxial relationship [001]oCTON//[001]STO (Figure b and Supporting Information Figure S6) in spite of the twofold symmetry of the SrTiO3(110) substrate. This was because the (010)o-oriented thin film of CaTaO2N (having the Pbnm space group) had a twofold screw axis parallel to the surface normal. Based on these XRD analyses, we concluded that the inferior crystallinity of the (100)pc- and (111)pc-oriented CaTaO2N thin films was attributable to the presence of in-plane rotational domains. It should be noted that translational domain boundaries, at which the octahedral tilt connection is discontinuous (Figure d) as in the case of a stacking fault, can exist even in a single in-plane rotational domain. We therefore propose that the formation of translational boundaries such as these degraded the crystallinity and surface flatness of the 190 nm thick (110)pc-oriented CaTaO2N thin film, which consisted of a single in-plane rotational domain.

Figure 4

Schematic illustrations of the domains of the CaTaO2N thin films. Note that the orthorhombicity of the cells is enhanced to show structural differences. (a–c) In-plane rotational domain(s) of the CaTaO2N thin films grown on the (a) (100), (b) (110), and (c) (111) planes of SrTiO3 substrates. (d) Example of the translational domain boundary of a (100)pc-oriented CaTaO2N thin film, in which Ta(O,N)6 octahedra having the same color have the same octahedral tilt.

The foregoing discussion of domain structures suggests that not only lattice matching but also symmetry matching between the film and the substrate could improve the crystallinity of CaTaO2N films. This possibility was assessed by synthesizing a thicker (100)pc-oriented CaTaO2N thin film on a symmetry-matched DyScO3 substrate that also provided better lattice matching (orthorhombic, ao = 5.54 Å, bo = 5.71 Å, co = 7.89 Å, lattice mismatch of ∼−0.2% with a pseudocubic approximation). Figure a,b shows an RSM around the 301pc diffraction and an AFM image, respectively, of a 190 nm thick (100)pc-oriented CaTaO2N thin film grown on a DyScO3 (110)o substrate. The 301pc diffraction peak from the CaTaO2N thin film (Figure a and Supporting Information Figure S2d) is much sharper than that produced by a 190 nm thick film on SrTiO3(100) (Figure a) because of the single domain growth (Supporting Information Figure S8). At the same time, a clear step and terrace surface was maintained even with the 190 nm film thickness (Figure b,c), whereas this surface was not evident on the SrTiO3(100) substrate (Figure d). At present, it is unclear which is a more significant factor for improving the crystallinity of the CaTaO2N film, lattice matching or symmetry matching. This question will be answered by comparing the films grown on two kinds of substrates, which indicate good lattice matching to CaTaO2N but have different symmetries, although such a substrate pair is not commercially available. By using a perovskite oxynitride thin film with a smaller lattice constant such as NdTiO2N (Supporting Information Figure S1), contributions of lattice matching and symmetry matching to crystallinity would be evaluated in more detail.

Figure 5

(a) RSM and (b) AFM image of a 190 nm thick (100)pc-oriented CaTaO2N thin film grown on a DyScO3(110)o substrate. (c) Cross-sectional line profile of the AFM image along the white line in (b).

Conclusions

(100)pc-, (110)pc-, and (111)pc-oriented stoichiometric CaTaO2N thin films were heteroepitaxially grown on SrTiO3 substrates by RF reactive magnetron sputtering. These films showed good crystallinity and atomically flat surfaces when fabricated with low thickness values (t = 20 nm). The crystallinity and surface flatness both deteriorated as the thickness increased. This effect was associated with the crystallographic orientation of the films primarily because of the formation of in-plane rotational domains that, in turn, stemmed from a symmetry mismatch between the orthorhombic CaTaO2N and cubic SrTiO3. The (110)pc-oriented CaTaO2N thin films consisting of single rotational domains showed a higher degree of crystallinity and improved surface morphology compared to the (100)pc-oriented (with fourfold rotational domains) and (111)pc-oriented (with threefold rotational domains) thin films. On the basis of these results, we concluded that not only lattice matching but also symmetry matching is vital during the heteroepitaxial growth of high-quality CaTaO2N thin films. The observation that a (100)pc-oriented CaTaO2N thin film deposited on a lattice- and symmetry-matched orthorhombic DyScO3(110)o substrate exhibited high crystallinity and an atomically flat surface even though the film was thick (t = 190 nm) confirms this hypothesis. This finding should also be applicable to the heteroepitaxy of various other perovskite oxynitrides with orthorhombic cells, which are the most common type (Supporting Information Figure S1), and will be helpful in the investigation of the intrinsic properties of these materials.

Experimental Section

CaTaON thin films were deposited by RF reactive magnetron sputtering (Eiko Co., Ltd.) using a 50 mm diameter CaTaO ceramic target (Toshima Manufacturing Co., Ltd.). In this manner, (100)pc-, (110)pc-, and (111)pc-oriented thin films were grown on the (100), (110), and (111) planes of SrTiO3 single-crystalline substrates. In addition, (100)pc-oriented thin films were fabricated on the (110)o planes of DyScO3 (orthorhombic) single-crystalline substrates. Prior to each deposition, the SrTiO3 and DyScO3 substrates were annealed in air under the following conditions to obtain step and terrace surfaces: 2 h at 1050 °C for SrTiO3 (100), 5 h at 1000 °C for SrTiO3 (110) and (111), and 4 h at 1000 °C for DyScO3(110)o. During the thin-film deposition process, the substrate temperature was maintained at 630 °C and gaseous Ar and N2 were introduced into the chamber, in which the base pressure was 1.0 × 10–5 Pa. The total gas pressure and total gas flow rate were set at 0.75 Pa and 20 standard cm3·min–1, respectively. The N partial pressure, PN, was varied between 0 and 0.3 Pa to adjust the N content in the films. The input RF power and the target–substrate distance were set at 100 W (with a power density of 5.09 W·cm–2) and 12 cm, respectively.

The N to Ta ratios (N/Ta) of the thin films were calculated from the N amounts determined by heavy ion elastic recoil detection analysis (ERDA) and the Ta amounts determined by Rutherford backscattering spectrometry with a 38.4 MeV 35Cl beam generated by a 5 MV tandem accelerator (Micro Analysis Laboratory, The University of Tokyo [MALT]).[27] The surface morphologies of the thin films were characterized by AFM (SII-nanotechnology, SPI4000 with SPA400). The crystal structures were analyzed using XRD (Bruker AXS D8 DISCOVER) with Cu Kα radiation, employing a scintillation counter, a one-dimensional array detector (VANTEC1), and a two-dimensional area detector (VANTEC500). Film thicknesses, t, were measured with a stylus surface profiler (Veeco, Dektak 6M).