Self-regulated growth of [111]-oriented perovskite oxide films using hybrid molecular beam epitaxy

The growth of perovskite oxides has long been focused on [001]-oriented films due to the low surface energy of the {001} facets and the inherent stability of the growth front. Other film orientations can have a striking effect on material’s properties. For example, perovskite films grown along the [111] have been predicted to display nontrivial band topologies,^1,2 ferroelectricity,^3,4 Dirac semimetal phases,⁵ spin-nematic phases,¹ the quantum spin Hall effect,^6,7 and the quantum anomalous Hall effect.^8–11 Although a plethora of exotic phenomena have been predicted to exist in [111]-oriented perovskite layers, few have been experimentally realized due to the inherent difficulties in growing [111]-oriented perovskite films and superlattices.

Unlike [001]-oriented perovskites such as SrTiO₃ that have alternating charge neutral Sr²⁺O²⁻/Ti⁴⁺O₂⁴⁻ layers, [111]-oriented SrTiO₃ has alternating Sr²⁺O₃⁶⁻/Ti⁴⁺ layers creating a stacking sequence with highly charged atomic planes. The polar stacking sequence gives rise to the emergence of electric dipoles detrimentally affecting the growth, potentially driving the formation of unexpected phases, undesired orientations, and unwanted exposed facets as the film attempts to minimize its surface energy.^12,13 Typically, high energy growth techniques operating far from thermodynamic equilibrium were required to overcome the effects of this high energy surface and achieve atomically smooth surfaces, such as pulsed laser deposition (PLD), which has been used for the growth of SrRuO₃,^14–16 SrFeO₃,¹⁷ La_0.7Sr_0.3MnO₃,¹⁸ AlFeO₃,¹⁹ NdNiO₃,²⁰ SrIrO₃,²¹ LaFeO₃,²² BaTiO₃,²² LaAlO₃,²³ PbTiO₃,²⁴ and Ba_0.94La_0.04SnO₃²⁵ or such as sputtering for the growth of SrTiO₃,^12,26 NdNiO₃,²⁷ and GdMnO₃.²⁸ 

While [111]-oriented perovskites have been successfully synthesized using these methods, it is highly desirable to establish growth using a low energy deposition technique, such as molecular beam epitaxy (MBE), where unintentional defect concentrations can be minimized. Currently, the homoepitaxial growth of SrTiO₃ is the only reported successful growth of [111]-oriented perovskites by MBE.²⁹ The highly desirable heteroepitaxial growth to take advantage of materials design degrees of freedom in thin film growth, specifically, epitaxial strain, atomically abrupt interfaces and layer thickness, and layering sequence, has remained elusive. The inherent difficulties of utilizing MBE for heteroepitaxial growth along unfavorable directions have been observed for the growth of [111]-oriented rocksalt MgO³⁰ and CaO³¹ on GaN. Under regular conditions, where the cations were supplied by evaporation and subsequently oxidized on the growth front, octupolar surface reconstructions were found to serve as nucleation points for lower energy {100} facets, which resulted in 3D film growth after just a few monolayers. In contrast, the co-exposure of the surface to water vapor enabled the continuous layer-by-layer growth of [111]-oriented rocksalt films. During growth, hydroxyl groups bonded to the positively charged cations on the surface, thereby reducing the polar character of the surface and thus stabilizing the growth front. This effect was further confirmed through density functional theory calculations.³¹ The situation for the growth of [111]-oriented perovskites is similar, and it is therefore hypothesized that agents aiding in the charge reduction of the interface are beneficial to the growth. This effect might also be achieved through the use of other chemical groups such as polar ligands liberated by the thermolysis of the metalorganic precursors used in hybrid MBE (hMBE),^32,33 thus enabling the growth of perovskites along highly polar directions with high crystalline quality.

This hypothesis has been tested for the heteroepitaxial growth of the correlated metal SrVO₃, which has been extensively studied recently^34–38 due to its potential use as a transparent electrode material.³⁹ Since SrVO₃ is a metal, the material quality can be determined by the simple, widely used metric of the residual resistivity ratio (RRR), which is defined as the ratio of the room temperature and low temperature (2 K) resistivities, thus making SrVO₃ an ideal material choice to determine the effectiveness of [111]-oriented growth using hMBE. PLD grown [111]-oriented SrVO₃ has been reported but the low RRR of 5 confirmed a high defect concentration.⁴⁰ Recently, it was shown that hMBE, which exploits a self-regulated growth mode to achieve stoichiometric films, is ideally suited to grow ultraclean SrVO₃ thin films in the [001]-orientation.^41,42 Here, we present a detailed growth study of [111]-oriented SrVO₃ thin films by hMBE. A self-regulated growth window was discovered for [111]-oriented growth, and atomically abrupt surface and interfaces were found. The RRR value of 21 for hMBE grown films significantly exceeded the RRR quality metric of previously grown [111]-oriented SrVO₃ by a factor of 4.

Thin films of SrVO₃(111) were grown using a DCA M600 hMBE equipped with a Sr thermal effusion cell and a molecular oxygen plasma source. Vanadium was supplied in the form of the metalorganic vanadium oxytriisopropoxide (VTIP) using a heated gas injector connected to a heated gas inlet system. The VTIP flux was controlled by maintaining a constant gas inlet pressure, P_VTIP. Additional details of the hMBE setup can be found elsewhere.^33,42 Growth was performed on the closest lattice matched substrate, [111]-oriented (La_0.3Sr_0.7)(Al_0.65Ta_0.35)O₃ (LSAT), resulting in a 0.67% tensile strain. Stoichiometric growth conditions were mapped by choosing a fixed Sr flux of 2.50 × 10¹³ cm^–2 s^–1 using a quartz crystal microbalance and growing a series of samples using different P_VTIP pressures.

Immediately after growth, films were cooled in vacuum to 100 °C and reflection high-energy electron diffraction (RHEED) images were taken along the 〈10-1〉 and 〈11-2〉 azimuths to examine the surface morphology as a function of VTIP gas inlet pressure, P_VTIP, as shown in Fig. 1. For VTIP inlet pressures between 78.0 mTorr and 84.0 mTorr, the RHEED intensity spots were confined to Laue circles and the Kikuchi lines were sharp, both signifying a smooth highly crystalline film surface similar to what has been observed for [001]-oriented SrVO₃ films.⁴² RHEED taken from SrVO₃ films grown outside this region exhibited a more streaky RHEED pattern with intensity modulations along the second order diffraction rods, indicative of a slightly more corrugated growth front, yet still two-dimensional. The occurrence of these RHEED features marked the boundaries of the self-regulated growth window, which ranged from 78 mTorr to 84 mTorr, as will be discussed below. Reconstructions of the polar (111) perovskite surface are known to be complex, and manifold atomic arrangements have been observed depending on cation termination and oxidation conditions.⁴³ For example, [111]-oriented SrTiO₃ has been observed to form reconstructions including n × n where n = 1–6, $7×7R19.1°$⁠, and $13×13R13.9°$ depending on surface termination, oxidation conditions, and annealing temperature and time.^43–45 For SrVO₃ films grown here, surface reconstructions were found along both azimuths, which were sometimes difficult to index. For example, along the 〈10-1〉 azimuth, additional reflections from surface reconstructions appeared somewhat blurred, suggesting that the surface did not assume one specific reconstruction. Instead, multiple surface reconstructions coexisted. Remarkably, it was observed that an odd numbered reconstruction, referred to as (2n + 1)×, was present for all Sr-rich films grown with a gas inlet pressure P_VTIP ≤ 78 mTorr, while an even reconstruction, referred to as (2n)×, was present inside the growth window, as shown in Fig. 1(a). Reconstructions along the 〈11-2〉 azimuth were more sharply defined and followed a similar trend, going from a 7× reconstruction on the Sr-rich side to a 6× reconstruction inside the growth window. RHEED features arising from surface reconstructions vanished with increasing P_VTIP to become exceptionally blurry, likely due to a superposition of odd and even surface reconstructions, and finally vanished for V-rich films (P_VTIP > 84 mTorr).

Films were further characterized using x-ray diffraction. High-resolution 2θ–ω scans around the [111]-diffraction peak of LSAT and SrVO₃ are shown in Fig. 2(a). Unlike the case of commensurately strained SrVO₃(001) films, where the x-ray film peak position reflected the films’ cation stoichiometry due to the associated lattice parameter expansion,^42,46 the film peak position for [111]-oriented films did not change as a function of VTIP pressure. Only interface and surface quality of the films seemed to be affected by the growth conditions, as evidenced from the presence of Kiessig fringes around the [111] film peaks. X-ray scans taken from films grown at VTIP foreline pressures in the range of 78 mTorr–84 mTorr (labeled in orange) had pronounced Kiessig fringes indicating abrupt film surfaces, which faded away for larger (P_VTIP = 86 mTorr) and smaller (P_VTIP = 74 mTorr, 76 mTorr) VTIP foreline pressures (labeled in blue). A reciprocal space map taken around the 303 asymmetric diffraction peak of the SrVO₃(111) film grown at P_VTIP = 81 mTorr revealed that the films inside the growth window were coherently strained, as shown in Fig. 2(b). Films grown outside the growth window were found to partially relax, as shown in the supplementary material. Each high-resolution x-ray 2θ–ω scan was fit using GenX⁴⁷ to determine the film thickness and the interplanar spacing d₁₁₁. The interplanar spacing was directly compared to those of [001]-oriented SrVO₃ films, which were grown immediately after the [111]-oriented growth series. The trends of d₁₁₁ and d₀₀₁ are shown as a function of P_VTIP in Fig. 2(c). The interplanar spacings extracted for both [111]-oriented and [001]-oriented films were also compared to the calculated lattice parameter of strained SrVO₃ using the experimentally determined Poisson ratio of 0.24 for SrVO₃.^41,48 For [001]-oriented films, a clear growth window was observed as d₀₀₁ reached a minimum out-of-plane interplanar spacing for P_VTIP values from 78 mTorr to 83 mTorr, similar to previous studies and near the calculated lattice parameter of 3.824 Å for strained SrVO₃(001).^41,42 On the other hand, the out-of-plane interplanar spacing of [111]-oriented films near the calculated value of d₁₁₁ = 2.209 Å remained relatively unchanged as a function of VTIP pressure, suggesting that defect accommodation mechanisms incorporating excess Sr and V into the films were present in [111]-oriented SrVO₃ that did not increase the out-of-plane lattice parameter and that were unavailable for [001]-oriented films.

To obtain further insights as to why the out-of-plane interplanar spacing remained unchanged for such a wide range of growth parameters, atomic force microscopy (AFM) images were taken to map the film surface morphology as a function of P_VTIP. The rms roughness values as a function of P_VTIP along with representative AFM images are shown in Figs. 2(d) and 2(e). Films grown at VTIP pressures between 78 mTorr and 84 mTorr had an atomic terrace morphology with a rms surface roughness of ∼(0.2 ± 0.1) nm. It is remarkable that this favorable surface morphology was aligned with the SrVO₃ growth window for [001]-oriented films. SrVO₃(111) films grown outside the SrVO₃(001) growth window revealed regularly shaped, rather deep holes with lateral dimensions on the order of several tens of nanometers. These holes were separated by otherwise atomically smooth film areas on the order of hundreds of nanometers. These regularly shaped holes were primarily oriented in the same direction, suggesting that specific crystallographic directions gave rise to the facets defining the holes’ circumference. Films grown under Sr-rich conditions had hexagonal shaped holes, while films grown under V-rich conditions had primarily triangular shaped holes that were about half the lateral size compared to those found in Sr-rich films. These holes allowed films grown under Sr-rich and V-rich conditions another avenue to release stress induced by increased levels of cation nonstoichiometry in the film plane rather than by increasing the out-of-plane interplanar spacing. Therefore, these hole formations rendered the out-of-plane interplanar spacing irrelevant for the determination of stoichiometric growth conditions for [111]-oriented SrVO₃ films. Instead, the pronounced increase in film roughness due to the hole formation gave rise to a growth-window-like relation, as shown in Fig. 2(d). The growth window from surface roughness data of SrVO₃(111) films closely matched the width and position of the growth window found for SrVO₃(001) films using the interplanar spacing in the out-of-plane direction d₀₀₁ [see Fig. 2(c)]. The close resemblance of the growth windows for SrVO₃(111) and SrVO₃(001) films suggested that the kinetic processes at play were not affected by the film surface orientation.

While at first, it seemed strange that a surface sensitive technique such as RHEED was not able to identify a film surface with nanometer deep corrugations, the specific, rather seldom film surface morphology found for SrVO₃(111) grown under non-stoichiometric conditions, namely, holes formed from inward oriented facets with lateral dimension on the order of tens of nanometers and hundreds of nanometers apart can indeed give the false sense of a smooth coalesced film. The streaky RHEED pattern occurred since the surface was quite smooth overall. Spotty RHEED patterns typically arise if electrons transmit through the crystal, e.g., when protrusions are present or the film growth occurs in island mode. If holes are present in the film, the majority of the electrons get diffracted on the overall smooth surface and only a small portion transmits through the edges of the hole that are sufficiently electron transparent. The contribution of these electrons gave rise to the intensity modulation along the diffraction rods, which occurred in the second order diffraction rod in the RHEED pattern for Sr-rich and V-rich growth conditions along both azimuths. Therefore, stoichiometric growth can still be monitored using this in situ technique, and the Sr-rich growth window edge is found by a change in surface reconstruction from sixfold to sevenfold along the 〈11-2〉 azimuth accompanied with a pronounced intensity modulation along the diffraction rods, while at the V-rich growth window edge, the intensity modulations emerged in addition to the blurred surface reconstructions.

To gain further insights how SrVO₃(111) films accommodated large non-stoichiometries, Sr-rich, stoichiometric, and V-rich films were imaged using HAADF-STEM performed in a ThermoFisher Titan³ S/TEM. In addition, energy-dispersive x-ray spectroscopy (EDS) scans were taken for each film and are shown in the supplementary material. Low magnification survey scans of the Sr-rich, stoichiometric, and V-rich film are shown in Figs. 3(a)–3(c), and higher magnification images are shown in Figs. 3(d)–3(f). Low magnification images of the Sr-rich film revealed large regions of uniformly spaced Ruddlesden–Popper (RP)-like stacking faults that are known to occur in Sr-rich SrVO₃(001) films. They appeared to be similar to those observed in Sr₃V₂O₇;^49,50 however, in this case, the orientation of the stacking faults was along the [111] direction. Rocksalt-like AO stacking faults oriented along the [111] direction in ABX₃ perovskites have been previously observed in halide perovskites, such as Cs₃B₂X₉ (B = Sb, In, Bi and X = Cl, Br, I),^51–53 as well as organic perovskites such as [NH(CH₃)₃]₃Sb₂Cl₉,^54,55 whereby long-range vacancy order and sizable bond distortion were observed in the former case, and where the local degrees of freedom of organic molecules stabilized the layered structure in the latter case. Given the nominal stacking sequence of the Sr-rich phase, namely, two perovskite blocks with layers B–AX₃–B–AX₃ followed by the stacking fault AX₃, the general chemical formula of A₃B₂X₉ is expected; however, unlike in the two cases of halide and organic perovskites, the X anion is doubly charged for the oxide case suggesting the presence of an oxygen vacancy ordered phase of either Sr₃V₂O₈ with vanadium assuming a highest possible oxidation state 5+ or Sr₃V₂O₇ with V⁴⁺ on the B-site.

STEM images taken from stoichiometric SrVO₃(111) films are shown in Fig. 3(b). The stoichiometric films were found to have the expected cubic perovskite structure, and no large-scale defects or structural domains were observed. The higher magnification images revealed virtually no defects, indicating a phase-pure, stoichiometric, and high-quality film. The interface between the film and substrate was observed to be abrupt due to the sharp Z-contrast within one unit cell. Finally, films grown under V-rich conditions were found to exhibit wedge-shaped holes lined with amorphous material, as shown in Fig. 3(c). The higher magnification image shown in Fig. 3(f) revealed crystalline facets that contained 〈010〉 and 〈101〉 directions. The triangular shape of the holes seen in AFM suggested that a specific family of planes formed the facets. We suspect that the crystal facets formed are {001} as they are the lowest energy facets in perovskites; however, more in-depth analysis is required to unambiguously determine the geometry of the wedge-shaped holes. From the EDS measurements shown in Fig. S2, the amorphous material covering the facets primarily consisted of V, which is suspected to act as an amorphizing agent, forming a cap which prevented the advancement of epitaxial growth along the out-of-plane direction and instead gave rise to faceted growth. Aside from these two prominent features in Sr-rich and V-rich films, no other extended defects were observed. Again, abrupt interfaces with the substrate were formed, as shown in the supplementary material, indicating that an increasing degree of non-stoichiometry was accumulated on the growth front first before it was incorporated into the film. The STEM results suggested that Sr-rich films accommodated excess Sr into RP-like phases, while V-rich films incorporated excess V into amorphous regions forming {001} facetted growth, giving rise to an out-of-plane interplanar spacing that did not vary as a function of P_VTIP as was observed in XRD measurements in Fig. 2(c).

While it has been shown that films grown inside the growth window had no visible defects, the material quality was further characterized by temperature dependent resistivity measurements to determine the RRR. The dominant scattering mechanism at low temperature arises from defects in the film, while phonon scattering limits the carrier mobility at room temperature; therefore, the RRR is an ideal metric for determining material quality. The temperature dependent resistivity of the [111]-oriented SrVO₃ grown at P_VTIP = 81 mTorr is shown in Fig. 4(a) in comparison to the highest quality [001]-oriented SrVO₃ grown by MBE³⁴ and PLD.³⁵ The resistivity of [111]-oriented SrVO₃ was found to be lower at all temperatures, ranging from a room temperature resistivity of 2.8 × 10⁻⁵ Ω cm to a residual resistivity of 1.3 × 10⁻⁶ Ω cm, resulting in a RRR of 21. The RRR of [111]-oriented SrVO₃ is shown in comparison to [111]-oriented SrVO₃ grown by PLD,⁴⁰ the highest quality [001]-oriented SrVO₃ grown by MBE³⁴ and PLD,³⁵ and ultraclean SrVO₃ grown by hMBE⁴¹ in Fig. 4(b). It is not clear at this point why the RRR of hMBE grown SrVO₃(001) was significantly higher compared to [111]-oriented films. However, it is noted that the RRR of these [111] films favorably compares to [001] films grown by other techniques. Many factors could have affected the unintentional incorporation of defects for the films grown in this study. The higher surface energy of the (111) face could lead to a higher unintentional incorporation of the metalorganic ligands. In addition, a different VTIP precursor charge as well as the aging effects of the metalorganic supply system are potential defect sources. Furthermore, the use of oxygen plasma was necessary to stabilize the growth for [111]-oriented films, which was not required for [001] films.

Optical properties were also investigated and compared to SrVO₃(001) films. The refractive index and extinction coefficient for [111]- and [001]-oriented SrVO₃ thin films grown by hMBE are shown in Figs. 4(c) and 4(d), respectively. Two samples of equivalent thickness (47 nm and 46 nm, determined by x-ray diffraction) and a significantly thinner SrVO₃(111) film (17 nm) were compared to a 50-nm-thick SrVO₃(001) film. No significant sample to sample variations were found for the refractive index or extinction coefficient in the wavelength interval from 200 nm to 1700 nm. As expected from the isotropy of the cubic system, the extinction coefficient followed the same trend as [001]-oriented SrVO₃ where a region with low k was confined by increases in the infrared region due to free carrier reflection and in the ultraviolet range due to interband absorption, leaving a spectral window of high optical transparency in the visible range.³⁹ The extinction coefficient for [111]-oriented samples was significantly lower in the visible range from 450 nm to 750 nm suggesting a higher optical transmission. This effect was likely due to different strain states of the films. This lower extinction coefficient in the visible range coupled with nearly the same room temperature resistivity suggests that SrVO₃ retains a high performance as a transparent conductor in the [111]-orientation.

In conclusion, we have demonstrated that SrVO₃(111) can be grown in a self-regulated growth regime for a wide range of VTIP pressures using hMBE. The growth window was found to be independent of the growth direction. The formation of large-scale defects such as holes and layered Sr-rich phases under non-stoichiometric growth conditions rendered the out-of-plane lattice parameter to be an ineffective measure to determine the growth window position. The growth window boundaries were instead identified by RHEED and AFM measurements. Stoichiometric films inside the growth window were found to be virtually defect-free in STEM. Stoichiometric [111]-oriented SrVO₃ was found to have a high RRR of 21 and optical properties with a slightly reduced extinction coefficient in the visible range compared to SrVO₃(001). The ability to grow highly crystalline perovskite materials along an unfavorable growth direction using a self-regulated, low energy growth technique opens up opportunities to investigate the electronic, magnetic, and topological properties of [111]-oriented perovskites and heterostructures without obstructions from defects.

Thin films of SrVO₃(111) were grown under a 0.67% tensile strain on [111]-oriented (La_0.3Sr_0.7)(Al_0.65Ta_0.35)O₃ (LSAT) substrates using a DCA M600 hMBE equipped with a Sr thermal effusion cell and a molecular oxygen plasma source. Vanadium was supplied in the form of the metalorganic vanadium oxytriisopropoxide (VTIP) using a heated gas injector connected to a heated gas inlet system. The VTIP flux was controlled by maintaining a constant gas inlet pressure, P_VTIP. Additional details of the hMBE setup can be found elsewhere.^33,42 Stoichiometric growth conditions were mapped by choosing a fixed Sr flux of 2.50 × 10¹³ cm⁻² s⁻¹ using a quartz crystal microbalance and growing a series of samples using different P_VTIP pressures. LSAT substrates were prepared by sonication in acetone and isopropyl alcohol followed by a 5-min UV/ozone clean. Prior to film growth, substrates were heated to a temperature of 900 °C and exposed for 15 min to an oxygen plasma supplied from an RF plasma source (Oxford Applied Research) operated at 250 W resulting in a background pressure of ∼5 × 10⁻⁷ Torr. Prior to growth, no surface reconstructions were observed on the LSAT substrates. SrVO₃ films were grown under the same conditions for ∼1 h, resulting in film thicknesses of about 50 nm.

High-resolution x-ray diffraction measurements were taken with a Philips X’Pert³ Panalytical MRD four-circle diffractometer using Cu–K_α1 radiation. Film surface morphologies were measured using a Bruker Dimension Icon AFM in peak force tapping mode. Resistivity was measured using a Quantum Design Physical Properties Measurement System in the van der Pauw geometry. Optical properties were determined by measuring the optical spectra with a J.A. Woollam M2000 rotating-compensator spectroscopic ellipsometer and extracting the optical constants using a least-squares regression analysis and an unweighted error function to fit the experimental ellipsometric spectra based on an optical model consisting of a semi-infinite LSAT substrate (obtained by measuring the ellipsometric spectra of a bare [111]-oriented LSAT substrate from the same batch), [111]-oriented SrVO₃ film, and surface roughness. The surface roughness was represented by a Bruggeman effective medium approximation of 0.5 void +0.5 film material fractions.

The annular bright-field (ABF) STEM and the energy-dispersive x-ray spectroscopy (EDS) were performed in a ThermoFisher Titan³ S/TEM equipped with double spherical aberration correctors. The operating voltage was 300 kV with a beam current of 0.1 nA, and the probe convergence angle was 30 mrad. The aberration correctors were tuned to achieve a spatial resolution of about 70 pm. The EDS mapping was acquired with a beam current of 0.09 nA and a dwell time of 0.1 µs over 256 × 256 pixels in each map. Each map was acquired with the automatic drift-correction turned on in the Bruker software, and the acquisition time for each map was between 8 min and 20 min to ensure a good signal-to-noise ratio.

See the supplementary material for additional XRD and STEM analysis.

We would like to thank Professor Jon-Paul Maria for helpful discussions. J.R., T.K., A.P., and R.E.-H. acknowledge support from the National Science Foundation through DMREF Grant No. DMR-1629477. L.M. and N.A. acknowledge support from the Penn State Center for Nanoscale Sciences, a National Science Foundation MRSEC under Grant No. DMR-2011839. This material is based on the work supported, in part, by the NSF Graduate Research Fellowship Program under Grant No. DGE-1255832 (J.R.). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

The data that support the findings of this study are available from the corresponding author upon reasonable request