Thin-film stabilization of LiNbO₃-type ZnSnO₃ and MgSnO₃ by molecular-beam epitaxy

Awide spectrum of functional properties exhibited by polar crystals¹ are at the heart of many devices including pyroelectric sensors, optical wavelength converters, and ferroelectric memories.² One of the representative polar systems is perovskite-derivative ABO₃; the ideal perovskite structure is cubic and nonpolar. The lattice distortion from cubic perovskite and appearance of polar phases are guided with the tolerance factor,^3,4 t, given by $t=rA+rO/2rB+rO$⁠, where r_A, r_B, and r_O are the ionic radii of the A-site cation, B-site cation, and oxygen ion, respectively. The Goldschmidt diagram,⁴ presented in Fig. 1(a), indicates that the deviation of t from unity (ideal cubic structure) involves the BO₆ octahedral tilting. In the presence of second-order Jahn-Teller active ions or lone-pair ions, the cubic instability increases and, at low temperature, often gives rise to a structural transformation to a low-symmetry polar phase. Archetypal ferroelectric oxides like PbTiO₃ and BiFeO₃ are classified into this category. In the low-t region (t < 0.8), corundum-derivative ABO₃ structures become predominant, known as LiNbO₃-type and ilmenite-type. The LiNbO₃-type structure is polar (space group: R3c), while the ilmenite-type one is nonpolar (⁠$R3¯$⁠). Recently, a series of new LiNbO₃-type polar ABO₃ such as ZnSnO₃, MnSnO₃, and FeTiO₃ have been synthesized by high-pressure techniques.^5–11 In those oxides, in addition to ferroelectric polarization as large as those of LiNbO₃ and LiTaO₃,¹² control of electronic and magnetic properties by utilizing various combinations of two cations is anticipated. Couplings of charge and spin degrees of freedom via the low symmetry character would also lead to candidates for multiferroic materials.^{6,10,11,13,14}

One important step toward further exploration of electronic functionality in relevant LiNbO₃-type oxides is the thin-film growth.^13–15 Oxide film growth generally begins with considerations of oxygen sublattice matching at the film/substrate interface.¹⁶ The averaged oxygen-oxygen distance and oxygen sublattice symmetry are compared between substrate and overgrown film materials. In Fig. 1(b), three-fold symmetry planes of typical oxide substrates, SrTiO₃(111), Al₂O₃(0001), and LiNbO₃(0001), are depicted. For simplicity, only the topmost oxygen ions in AO₆ and BO₆ octahedra are shown by small red spheres. These crystal planes have similar oxygen frameworks and are suited to thin-film growth of oxides with triangular lattices. In particular, corundum-type A₂O₃ and LiNbO₃-type ABO₃ have identical arrangement of oxygen sublattice and octahedral vacancy sites. As seen in the schematic cross sections, however, periodic cation occupation is inherent to the LiNbO₃-type structure. In the film growth of (111)-oriented cubic perovskite-type oxides, such a cation geometry has not been considered much. One of the reasons for this may be that their large A and small B cations are coordinated by twelve and eight oxygen ions, respectively, and thus hardly occupy the anti-sites. In the case for LiNbO₃-type oxides, the situation is however not so clear; the similar octahedral structural units as well as comparable ionic radii might give rise to disordered cation arrangement [Fig. 1(b)]. By molecular-beam epitaxy on LiNbO₃(0001) substrates, we recently succeeded in stabilizing the LiNbO₃-type ZnSnO₃ film and demonstrated its high-mobility transistor operation.¹⁷ In this paper, we describe the decisive role of substrate in the crystallization and the thin-film synthesis of a new LiNbO₃-type oxide, MgSnO₃. These polar stannates not only are transparent oxide semiconductors exhibiting good electrical conduction but also can be a candidate for exploring a polar metal phase.

The films were grown on SrTiO₃(111), Al₂O₃(0001), LiNbO₃(0001), and LiNbO₃-type LiTaO₃(0001) substrates by oxygen-plasma assisted molecular-beam epitaxy. Two-side polished, pre-poled, congruent LiNbO₃ and LiTaO₃ single crystal substrates were mainly used, and stoichiometric LiNbO₃ substrates were also used for scanning transmission electron microscopy (STEM). Prior to the film growth, SrTiO₃, Al₂O₃, congruent LiNbO₃, and stoichiometric LiNbO₃ substrates were annealed in an oxygen ambient at 900 °C, 1000 °C, 900 °C, and 950 °C, respectively. LiTaO₃ was not annealed to keep the pre-poled state because the ferroelectric transition temperature is as low as 600 °C. Elemental Sn, Zn, Mg, and Mn were supplied from effusion cells with high-purity metal sources. Beam equivalent pressure was monitored with a nude ion gauge and was controlled by the cell temperature. Typical pressures were 5.0 × 10⁻⁶ Pa, 1.1 × 10⁻⁵ Pa, 3.5 × 10⁻⁶ Pa, and 1.1 × 10⁻⁵ Pa for Sn, Zn, Mg, and Mn, respectively. Atomic oxygen radicals were generated from O₂ gas using a radio frequency plasma cell. All the films presented below were fabricated at a substrate temperature of 500 °C. The Zn/Sn, Mg/Sn, and Mn/Sn atomic ratios in the films were adjusted to be ∼1 by compositional analysis with X-ray photoelectron spectroscopy and energy dispersive X-ray spectroscopy (see Fig. S1 and Methods in the supplementary material).

In Fig. 2(a), out-of-plane X-ray diffraction (XRD) patterns for Zn-Sn-O films on LiNbO₃ (negatively poled −Z surface), Al₂O₃, and SrTiO₃ substrates with Zn/Sn atomic ratios close to unity are compared. An intense ZnSnO₃(0006) reflection with thickness fringes is observed for the film grown on LiNbO₃(0001). The c-axis length is the closest to the bulk value of c = 14.00 Å around the nearly stoichiometric condition (Fig. S2), verifying the validity of our scheme for growth condition optimization. The film surface has a step-and-terrace structure [inset of Fig. 2(a)], and the step height is approximately one sixth of the c-axis length of ZnSnO₃ [Fig. 1(b)]. The c-axis oriented ZnSnO₃ film can also be grown on the positively poled + Z surface of LiNbO₃ [the top panel in Fig. 2(b)] and isostructural LiTaO₃(0001) (Fig. S3). In stark contrast, no traces of crystalline phases are discerned on Al₂O₃(0001) and SrTiO₃(111). Note that this distinct growth behavior was not merely due to substrate-dependent optimal growth conditions; even in the examinations of the Zn/Sn flux ratio and substrate temperature, the film growth was not successful.

We examined the effectiveness of LiNbO₃-type substrates on the growth of other polar stannate candidates, MgSnO₃ and MnSnO₃. While LiNbO₃-type MnSnO₃ is synthesized by a high-pressure technique,⁶ little has been known about structural variants of MgSnO₃; only the ilmenite-type form is known experimentally.^18,19 As shown in the middle panel in Fig. 2(b), a highly crystallized phase is obtained on LiNbO₃(0001) with a Mg/Sn flux ratio of 0.7. The XRD pattern satisfies the extinction rule for the LiNbO₃-type structure (see Fig. S4 for the wide scan), the peak angle of which is close to that of ZnSnO₃. A characteristic step-and-terrace structure, suggestive of common growth behavior, is also observed (inset). More importantly, the crystallization again proceeds only on LiNbO₃-type substrates but not on Al₂O₃ or SrTiO₃ substrates (Figs. S3 and S5). The Mn–Sn–O film also shows a weak but clear peak, which is consistent with the high-pressure LiNbO₃-type MnSnO₃ phase⁶ [the bottom panel in Fig. 2(b)].

On the (0001) plane of LiNbO₃-type crystals where two distinct cations and metal vacancy exist, the oxygen-oxygen distance L_O-O consists of three different values of 3.358, 2.884, and 2.723 Å for LiNbO₃, and 3.334, 3.022, and 2.784 Å for ZnSnO₃. By contrast, there are two L_O-O (2.862 and 2.523 Å with a ratio of 2:1) on corundum-type Al₂O₃(0001) and only one L_O-O (2.761 Å) on perovskite-type SrTiO₃(111). Using the averaged L_O-O value, the lattice mismatch between ZnSnO₃(0001) and the substrate is calculated to be 2.0%, 10.8%, and 10.4% for LiNbO₃(0001), Al₂O₃(0001), and SrTiO₃(111), respectively. In preceding studies using pulsed-laser deposition, it was reported that highly crystalline films of high-pressure phase NiTiO₃^13,14 and ZnSnO₃¹⁵ as well as ambient phase LiNbO₃ (Refs. 20 and 21) can be obtained on Al₂O₃(0001) and SrTiO₃(111) despite of the large lattice mismatch. Therefore, the observed substrate dependent growth points to the different growth mechanism of high-pressure phase stannates by molecular-beam epitaxy. We speculate that periodic modulation of L_O-O and surface potential might act as a seed for the LiNbO₃-type order.

Lattice parameters were evaluated for the highly crystalline stannate films by XRD reciprocal space mapping (RSM).Figure 3(a) displays RSM around LiNbO₃$(112¯12̲)$⁠. This ZnSnO₃ film on LiNbO₃(0001) (−Z surface) is the same to that used for the out-of-plane XRD measurement [Fig. 2(a)]. The film is partially relaxed, and the lattice parameters are calculated to be a = 5.236 Å and c = 14.25 Å, which are comparable to a = 5.262 Å and c = 14.00 Å reported for bulk.⁵ A locally modified ilmenite-like atomic arrangement at the domain boundaries¹⁷ is probably the origin of the c-axis elongation of 2% from the bulk value of ZnSnO₃. A similar result is obtained for the Mg–Sn–O film [Fig. 2(b)], shown in Fig. 3(b), yielding a = 5.211 Å and c = 14.24 Å. The shrinkage of lattice parameters as compared to those of ZnSnO₃ is consistent with the smaller ionic radius of Mg²⁺ (0.72 Å) than that of Zn²⁺ (0.74 Å), which supports the introduction of Mg into the A-site. Based on these results, we conclude that LiNbO₃-type MgSnO₃ is stabilized in the thin-film form. From the inspection of in-plane rotational symmetry by ϕ-scan measurement [Fig. 3(c)], these c-axis oriented films were found to contain two domains of three-fold symmetry, rotated by 60° in the (0001) plane.

Atomically resolved STEM imaging gives further evidence for the formation of LiNbO₃-type ZnSnO₃.Figure 4(a) shows a high angle annular dark field (HAADF) STEM image of a ZnSnO₃ film viewed along LiNbO₃$[11¯00]$⁠. For this experiment, a stoichiometric LiNbO₃ (−Z surface) was used as the substrate; high-quality ZnSnO₃ similar to those on congruent LiNbO₃ was obtained with identical growth conditions (Fig. S6). It is clear that the film is composed of highly crystallized and c-axis oriented domains over a large area. HAADF and annular bright field (ABF) images around the film/substrate interface are displayed in Figs. 4(b) and 4(c), respectively. In Fig. 4(d), we compared representative patterns, area A in the film and area B in the substrate, together with structural models, indicating the same orientation as observed by XRD. In addition, in the ABF STEM images, where light elements are more clearly resolved than in HAADF STEM ones, it is seen that the asymmetrical contrast due to oxygen ions in the LiNbO₃ substrate (area B) is partly maintained in the ZnSnO₃ film region (area A). Closely inspecting the image contrast inFigs. 4(a)–4(c), we also find an inhomogeneously bright lattice (area C), which is likely related to crystallographically different nanoclusters. One possible orientation that could explain the irregular pattern is SnO₂[001](010)∥LiNbO₃$[11¯00](0001¯)$⁠, though this was not detected in the macroscopic XRD characterization presented above. Suppression of such nanoclusters, which might be segregated in the boundaries of two domains of ZnSnO₃, would be important for the single-domain formation.

According to first-principles calculation,^19,22,23 ZnSnO₃ is a wide bandgap semiconductor with a highly dispersed conduction band of Sn 5s character. We determine the optical bandgap E_g of ZnSnO₃ and MgSnO₃ by transmittance measurement. We adopted LiTaO₃(0001) substrates (+Z surface) with a large E_g of 4.6 eV for this measurement because the bandgap of LiNbO₃ (3.7 eV) is not sufficiently large as compared to that of the stannate films (Fig. S7). The XRD pattern of the films on LiTaO₃(0001) is comparable to that on LiNbO₃ (Fig. S3). An extrapolation of the absorption versus photon energy (hν) plot, shown in Fig. 5(a), yields E_g ∼ 3.7 eV for ZnSnO₃ and ∼ 4.0 eV for MgSnO₃. Although the type of optical absorption is necessary for the detailed analysis, these results underpin the conjectured wide bandgap nature. We then performed photoconductivity measurement for a ZnSnO₃ film on LiNbO₃(0001) (−Z surface) using Xe lamp irradiation. As shown in Fig. 5(b), the pristine ZnSnO₃ sample under dark condition is highly resistive, with a sheet resistance R_s over 1 GΩ at temperature T = 300 K. Photoexcitation at T = 20 K induces persistent photoconductivity [inset ofFig. 5(b); also see Fig. S8] associated with a substantial (several orders of magnitude) decrease in R_s. Hall effect measurement revealed the electron-type carrier, carrier density of 1.5 × 10¹⁸ cm⁻³, and Hall mobility μ_H of 4.9 cm² V⁻¹ s⁻¹ at T = 20 K. On heating above T ∼ 200 K, R_s gradually recovers its original high value, indicating the primally electronic origin of the photo-induced conduction.

In the photo-excited state, R_s exponentially increases below 20 K. This is possibly due to a carrier localization and a subsequent reduction in μ_H in the region next to the substrate. Since we previously observed a high field-effect mobility of 45 cm² V⁻¹ s⁻¹ at T = 300 K in a ZnSnO₃ field-effect transistor,¹⁷ electrical transport at the film surface under electrostatic charge accumulation is worthy of investigation. To induce a high carrier density, we fabricated an electric-double-layer transistor (EDLT) on the ZnSnO₃ surface with an ionic liquid, N,N-diethyl-N-methyl-N-(2-methoxyethyl)ammonium bis(trifluoromethanesulfonyl)imide (See Methods in the supplementary material).Figure 5(c) displays T dependent R_s characteristics at various gate voltages (V_G). The application of a positive V_G decreases R_s (Fig. S9), as being consistent with the n-type conduction observed in the photoconductivity measurement. A large sheet carrier density close to 10¹⁴ cm⁻² at V_G = 6 V (Fig. S10) induces virtually T-independent behavior with a μ_H of as high as 20 cm² V⁻¹ s⁻¹ [inset of Fig. 5(c)], demonstrating the high-mobility nature of the Sn 5s derived conduction band. By fabricating a single-domain ZnSnO₃ film, a two-dimensional metallic state may emerge at the polar surface.

In summary, we demonstrated that the use of LiNbO₃-type substrates is the key to stabilizing crystalline ZnSnO₃ in the thin-film form and discovered a new structural analog MgSnO₃ by molecular-beam epitaxy. The results manifest the potential of these wide bandgap stannates for applications in transparent electronics. One of the obvious targets is to fabricate In-free transparent conducting films with appropriate donor elements. A more nontrivial approach is to induce free carriers at the film surface by utilizing its own ferroelectric polarization. Observation of polarization reversal by an electric field¹⁵ should be tackled for such studies. The duality of ferroelectric and high-mobility properties in wide bandgap polar stannates may bring ferroelectric materials research into a new stage.

See supplementary material for compositional analysis of Zn-Sn-O films by X-ray photoelectron spectroscopy (Fig. S1), optimization of the Zn/Sn flux ratio based on XRD results (Fig. S2), out-of-plane XRD patterns of ZnSnO₃ and MgSnO₃ films on congruent LiTaO₃(0001) (Fig. S3), wide scan XRD data on congruent LiNbO₃(0001) (Fig. S4), substrate dependent growth of MgSnO₃ (Fig. S5), out-of-plane XRD patterns of ZnSnO₃ on stoichiometric LiNbO₃(0001) (Fig. S6), transmittance spectra (Fig. S7), the dependence of persistent photoconductivity on light source power (Fig. S8), transfer characteristics measured at T = 220 K (Fig. S9), and μ_H data (Fig. S10) for a ZnSnO₃ EDLT, for experimental procedures of compositional analysis and device fabrication (Methods in the supplementary material).

The authors thank Y. Inaguma for his insightful advice, T. Matsuoka and S. Kuboya for the use of a spectrophotometer, NEOARK Corporation for the use of a maskless lithography system PALET, and I. Narita, F. Sakamoto, S. Ito, and K. Miura for their experimental assistance. This work was performed under the Inter-University Cooperative Research Program of the Institute for Materials Research, Tohoku University (Proposal Nos. 17G0415 and 17G0417), and also conducted in part at Advanced Characterization Nanotechnology Platform of the University of Tokyo, supported by the “Nanotechnology Platform” of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan. This work was supported by JSPS KAKENHI (No. JP15H02022) and Kato Foundation for Promotion of Science (No. KJ-2607).