Formation of ilmenite-type single-crystalline MgTiO₃ thin films by pulsed-laser deposition

Ilmenite-type (IL-type) crystal structure is closely related to corundum-type and LiNbO₃-type (LN-type) with chemical formula ABO₃ (A and B are cations and O is oxygen).^1–3 While all three structures are composed of layered honeycomb lattices, the combination of valences of A and B cations and whether inversion symmetry exists or not are different. Each structure is classified based on these differences. By considering the ionic radii of those cations, the Goldschmidt tolerance factor⁴ provides an empirical guide to predict stable crystal structures. Some materials are known to be stabilized as both IL-type and LN-type structures though the cation combination is identical; these two structures are controlled by growth methods, for example, using high-pressure synthesis^1,2,5–8 or thin-film techniques.^9,10 The IL-type structure is constructed by alternating stacking of honeycomb lattices composed of octahedral units AO₆ and BO₆ [Fig. 1(a), drawn by VESTA¹¹]. As a famous example, ATiO₃ (A = Mn, Fe, Co, and Ni) has been investigated for the rich physics of magnetism with a variety of intriguing magnetic orders of spins on the A sites.^12–15 Recently, thin-film studies on ATiO₃ are extended to magnetic devices,¹⁶ dielectric devices,^17–19 photochemical reactions,^20,21 and gas sensors.^20,22 Interestingly, IL-type compounds with honeycomb structures of IrO₆ octahedra such as MgIrO₃ (Refs. 23 and 24) have attracted great attention for quantum magnetism in the relationship with Kitaev’s honeycomb model.²⁵ Recently, to stabilize honeycomb lattices of IrO₆ octahedra in thin films, a superlattice approach has been proposed.²⁶ The two-dimensional features of the IL-type structure will open a great arena to investigate interactions among the honeycomb magnets of the 3d metals and the Kitaev candidates.

MgTiO₃ will be a useful IL-type material as a non-magnetic d⁰ insulator.¹² Previous studies on MgTiO₃ films have focused on the dielectric properties for low-loss optical applications. In these studies, samples were prepared by chemical vapor transport,^27,28 sol–gel method,^29–35 metalorganic solution deposition,^36,37 sputtering,^38–49 pulsed-laser deposition (PLD),^50,51 and the other methods.^52–54 The single-crystalline structure was obtained by annealing process in both sol–gel²⁹ and sputtering⁴³ methods, indicating IL-type MgTiO₃ is thermodynamically stable under appropriate conditions. In addition, another stable phase of Mg₂TiO₄ was also detected in polycrystalline films.⁴³ By applying PLD, oxygen pressure strongly affects the formation of IL-type thin film or other phases, relating to oxidation and lattice symmetry.¹⁰ At the surface of the corundum structure of Al₂O₃, the triangular lattice of oxygen inevitably provides two energetically degenerated configurations, leading to twin-domain formation.^55–57 Suppression of the twin-formation is crucially important to improve crystalline quality. In this study, we optimized growth conditions of oxygen pressure and growth rate for MgTiO₃ deposition, resulting in the synthesis of IL-type single-crystalline MgTiO₃ thin films. The bandgap of the MgTiO₃ film was evaluated to be about 4.4 eV from absorption spectra, indicating a transparent insulator.

MgTiO₃ thin films were fabricated on Al₂O₃(0001) substrates by PLD at 875 °C in various conditions of oxygen pressure (P_O2: 10⁻⁶ to 10⁻² Torr). The surfaces of the substrates were miscut by 0.2° from the (0001) plane along the [$1̄$100] direction, which were annealed in a furnace at 900 °C for 2 h in 1 atm oxygen flow before installation into the vacuum chamber. Sintered MgO (99.9%) and TiO₂ (99.99%) powders were pelletized by spark plasma sintering at 900 °C and 80 MPa to make a target for deposition. The target was ablated by a KrF excimer laser at a frequency of 2 Hz. The distance from the target to the substrate was fixed at 35 mm for all experiments. Crystal structures of the samples were characterized by x-ray diffraction (XRD) with the Cu Kα1 radiation and transmission electron microscopy. Thicknesses of the films were estimated from Laue oscillations of XRD intensity at a low angle reflection or around MgTiO₃ 0003 diffraction peak. The thicknesses of the films were typically 40 ± 5 nm, unless otherwise noted. The deposition rate of the thin films, which is one of the crucial parameters dominating crystalline orientation of MgTiO₃, was evaluated by dividing the thickness by the number of the supplied laser pulses. We regulated the laser intensity that was monitored at the outside of the PLD chamber to examine the deposition-rate dependences of the film growth. Since the precise evaluation of the energy density is difficult due to the restriction of the chamber configuration, the deposition rate is applied to a parameter for discussion of the film growth. Optical transmittance T_sample and reflectance R_sample of the samples against monochromatic visible light were measured using a spectrophotometer. The reflectance is a relative value that defines the reflectance of a reference aluminum mirror as 100%. Absorbance A_sample was estimated from T_sample and R_sample using the formula A_sample = −ln(T_sample + R_sample).

To optimize oxygen pressure, MgTiO₃ thin films were deposited under supplying oxygen gas with a deposition rate of roughly 13–15 nm/1000 pulses (nm/kp). Here, XRD spectra are magnified in Figs. 1(c)–1(e) presenting MgTiO₃ and ideally forbidden Al₂O₃ 0003 diffraction peaks around 20°, because, for MgTiO₃, the intensity of diffraction peak at 0003 is much stronger than that at 0006 (simulated by ICSD⁵⁸), in contrast to the well-investigated ATiO₃ with 3d transition elements, as shown in Fig. 1(b). Since the magnitude relationship of the 0003 and 0006 is reversed across the criteria of the ratio =1 in Fig. 1(b), the 0003 diffraction peak is suitable to examine the crystalline quality for MgTiO₃. As shown in Fig. 1(c), (111)-oriented Mg₂TiO₄ appears in a 2theta-omega scan of a sample fabricated with no supply of oxygen, where background pressure was around 10⁻⁶ Torr. Although MgTiO₃ diffraction also appears in Fig. 1(c), the peak intensity of MgTiO₃ 0003 is rather weak compared to that of Mg₂TiO₄ 111. With increasing P_O2 up to 10⁻⁵ Torr [Fig. 1(d)], only a diffraction peak of MgTiO₃ 0003 appears with suppression of that of Mg₂TiO₄ 111. Laue oscillations around the MgTiO₃ 0003 peak were clearly observed in Fig. 1(d), indicating the uniform thickness of the film. While the peak intensity of MgTiO₃ 0003 diffraction slightly becomes small as the pressure increases up to P_O2 = 10⁻² Torr [Fig. 1(e)], only MgTiO₃ 0003n peaks were detected for the samples deposited in between P_O2 = 10⁻⁵ and 10⁻² Torr. The insets in Figs. 1(c)–1(e) present the 2theta-omega scans around MgTiO₃ 0006 diffraction peak. As discussed in Fig. 1(b), the intensity of MgTiO₃ 0006 is rather small, but Mg₂TiO₄ 222 peak was observed only in the inset of Fig. 1(c). In view of the omega rocking curves shown as the red lines in Figs. 1(f) and 1(g) for the identical films in Figs. 1(d) and 1(e), respectively, the full-width-at-half-maximum (FWHM) for the film grown at P_O2 = 10⁻⁵ Torr is much narrower than that at 10⁻² Torr. Under the wide range of P_O2, (0001)-oriented single-phase thin films of IL-type MgTiO₃ were obtained.

To examine the crystalline orientation of the MgTiO₃ films, we measured azimuthal phi scans with in-plane rotation at MgTiO₃ and Al₂O₃ 104 diffraction peaks, as shown in Fig. 2(a). The diffraction of Al₂O₃ 104 (bottom) presents three-fold symmetry, which is consistent with the corundum structure. In contrast, the diffraction of MgTiO₃ 104 for the film deposited with a relatively high growth rate of 5.5 nm/kp (top) shows six peaks originating from the formation of twin domains. Here, we define A-domain (red-closed triangle): MgTiO₃[11$2̄$0](0001) || Al₂O₃[11$2̄$0](0001) and B-domain (red open triangle): MgTiO₃[$1̄1̄$20](0001) || Al₂O₃[11$2̄$0](0001). The ratio of peak intensities for B- and A-domains I_B/I_A directly corresponds to an abundance ratio of these domains. The seven times as high-intensity I_A as I_B in the top panel of Fig. 2(a) implies that the total volume of the A-domain is roughly seven times than that of the B-domain. Considering in-plane lattice mismatch of about 6% between MgTiO₃ and Al₂O₃, a parallel configuration such as A-domain is preferable. By reducing the deposition rate from 5.5 nm/kp (top) to 1.1 nm/kp [middle panel of Fig. 2(a)], I_B is dramatically suppressed. I_B/I_A of the film is estimated to be smaller than 0.4%, resulting in the three-fold symmetry that is consistent with the symmetry of IL-type single-crystal. Considering the surface atomic arrangement, twin-domain formation is usually observed in the (111)-oriented thin films for zinc-blende and rock-salt structures.^59–61 By contrast, the formation of the twin-domain is suppressed in corundum and IL-type structures.^{10,26,55,56,62,63} The arrangement of octahedral blocks may dominate the interface formation in contrast to the usual concept that oxygen arrangement at the terminated surface dominates the interface formation. In addition to the previous literature that P_O2 affects the in-plane orientation of the film,¹⁰ the deposition rate also plays a crucial role in the arrangement of crystalline orientation.

We summarize the XRD results as a growth phase diagram for IL-type MgTiO₃ films with P_O2 and deposition rate in Fig. 2(b). The area of the red circles reflects the intensity of the MgTiO₃ 0003 diffraction peak. We mainly examined two levels at P_O2 = 10⁻⁵ and 10⁻² Torr by applying different deposition rates. The circles at 10⁻⁵ Torr are clearly larger than those at 10⁻² Torr, implying that low P_O2 is enough to form IL-type MgTiO₃. As discussed in Fig. 1(c), however, the impurity phase of Mg₂TiO₄ segregates at the lowest P_O2 without an oxygen supply. Indeed, at both oxygen conditions, the size of the circles becomes larger as the deposition rate decreases, indicating improvement of the crystalline quality of the MgTiO₃ films. In fact, the FWHM for the film fabricated with the rate 1.1 nm/kp at P_O2 = 10⁻⁵ Torr is rather sharp with ∼0.09° [the black line in Fig. 1(f)]. In addition, I_B/I_A is overlaid in Fig. 2(b) with blue squares as a function of the deposition rate. Since I_B/I_A decreases with decreasing the deposition rate, the rate of roughly 1 nm/kp is enough low to suppress the B-domain volume below the detectable limit of the apparatus. Judging from these results, we concluded that P_O2 = 10⁻⁵ Torr and relatively low deposition rate ∼1 nm/kp are an optimum growth condition for IL-type MgTiO₃ thin films on Al₂O₃(0001) substrates.

XRD reciprocal space mapping was measured around the MgTiO₃ 0210 diffraction [Fig. 3(a)]. The measurement was performed on a 20-nm-thick film deposited with 1 nm/kp at 875 °C at P_O2 = 10⁻⁵ Torr. The different in-plane lattice constant between the MgTiO₃ film and the Al₂O₃ substrate indicates that the structure of the MgTiO₃ film relaxes from that of the Al₂O₃ substrate. The lattice constants of the film are estimated to be a = 5.06 Å and c = 13.88 Å, being both in-plane and out-of-plane lattice constants comparable as bulk values of a = 5.056 69(4) Å and c = 13.9034(2) Å (ICSD:⁵⁸ Collection Code 55285). All the films summarized in Fig. 2(b) except for the one with segregation of Mg₂TiO₄ show comparable c-axis lengths within the ±0.1% variation, which also indicates a small deviation of composition in the films. By applying transmission electron microscopy, the crystalline lattice at the interface was characterized along [11$2̄$0] [Fig. 3(b)]. IL-type lattice arrangement of the MgTiO₃ film is clearly observed on the Al₂O₃ substrate. The fast Fourier transformation of the image [Fig. 3(c)] shows the symmetry of the IL-type structure. The orientation relationship between MgTiO₃ and Al₂O₃ for the region shown in Figs. 3(b) and 3(c) is determined as A-domain: MgTiO₃[11$2̄$0](0001) || Al₂O₃[11$2̄$0](0001). It is noted that a small volume of the B-domain was partly observed in a different region of the identical sample though I_B/I_A for the sample was estimated to be smaller than 2%. To obtain a single-crystalline structure, the deposition rate should be carefully smaller than 1 nm/kp. The clean interface and high crystalline quality of MgTiO₃ on Al₂O₃ are evidenced by these structural characterizations.

Single-crystalline thin films are useful for the evaluation of bandgap by optical transmission spectroscopy. Figures 4(a) and 4(b) present the transmittance and reflectance of a 43-nm-thick MgTiO₃ film on an Al₂O₃ substrate and another bare Al₂O₃ substrate, and the absorbance spectrum calculated from the data for the MgTiO₃ film, respectively. The measured sample was deposited with 1.1 nm/kp at 875 °C and 10⁻⁵ Torr. As the reference of Al₂O₃ substrate, transmission T_sub (blue solid line) and reflection R_sub (blue dashed line) are almost constant for the overall wavelength region in Fig. 4(a), indicating no absorption of the substrate. The optical transmission of the MgTiO₃ film T_film (red solid line) in Fig. 4(a) displays reduction at a shorter wavelength region than 300 nm. In the visible region (roughly 400–800 nm), the T_film is slightly lower than that of the Al₂O₃ substrate T_sub. The difference originates from the higher reflection R_film (red dashed line) than R_sub. By using these spectra T_film and R_film, the absorbance of the MgTiO₃ film was calculated for estimation of the bandgap. Here, we plotted the absorbance and applied linear fitting [Fig. 4(b)] due to less information of the electronic band structure of MgTiO₃ whether direct or indirect transition. The linear fitting on the absorbance spectrum between 4.6 and 4.8 eV reaches to an intercept at around 4.4 eV, which is defined as the bandgap of the MgTiO₃ film in this study. This transition probably corresponds to the absorption from the O 2p to the unoccupied Ti 3d state, which is consistent with various Ti-oxide materials.^21,64–67 The estimated value of the bandgap of MgTiO₃ is roughly comparable to the values in the previous studies for films 4.11–4.19 eV (Ref. 45) with assuming indirect transition. Compared to other magnetic ATiO₃ (A = Mn, Fe, Co, Ni), no signature of absorption at the visible region and the larger bandgap of 4.4 eV than the other ilmenites of ∼3.0 eV (Ref. 21) suppose the less intervalence charge transfer in MgTiO₃. A relatively larger bandgap than rutile TiO₂ of about 3.37 eV (Ref. 68) implies that layered stacking of MgO₆ and TiO₆ enlarges the bandgap. High optical transmission at the visible region and uniform thickness of MgTiO₃ films fabricated by PLD is applicable to optical devices and to stacking with other IL-type oxides.

We explored growth conditions for MgTiO₃ thin films by pulsed-laser deposition. IL-type single-crystalline MgTiO₃ thin films were obtained at an optimum condition at P_O2 = 10⁻⁵ Torr and 875 °C with a low deposition rate of roughly 1 nm/kp. Deposition rate plays a crucial role in the suppression of twin-domain formation. An abrupt interface was observed, indicating honeycomb layered structure is well aligned as IL-type stacking. The bandgap is large enough to be a transparent window and an insulator without impurity states in the visible region. These results support that MgTiO₃ films are applicable to optical devices and heterostructures with magnetic oxides.

The authors thank K. Harata and S. Ito for their experimental support on spark plasma sintering and transmission electron microscope, respectively. The authors thank K. Miura for stimulating discussion. This work was partly supported by CREST (Grant No. JPMJCR18T2), JSPS KAKENHI (Grant No. 19H02423), and the Grant Fund for Research and Education of the Institute for Materials Research, Tohoku University. The spark plasma sintering process was performed under the GIMRT Program of the Institute for Materials Research, Tohoku University (Grant No. 202012-CRKEQ-0410).

The authors have no conflicts to disclose.

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