Stabilization of a honeycomb lattice of IrO6 octahedra by formation of ilmenite-type superlattices in MnTiO3

In quantum spin liquid research, thin films are an attractive arena that enables the control of magnetic interactions via epitaxial strain and two-dimensionality, which are absent in bulk crystals. Here, as a promising candidate for the development of quantum spin liquids in thin films, we propose a robust ilmenite-type oxide with a honeycomb lattice of edge-sharing IrO6 octahedra artificially stabilised by superlattice formation using the ilmenite-type antiferromagnetic oxide MnTiO3. Stabilised sub-unit-cell-thick Mn–Ir–O layers are isostructural to MnTiO3 and have an atomic arrangement corresponding to ilmenite-type MnIrO3. By performing spin Hall magnetoresistance measurements, we observe that antiferromagnetic ordering in the ilmenite Mn sublattice is suppressed by modified magnetic interactions in the MnO6 planes via the IrO6 planes. These findings contribute to the development of two-dimensional Kitaev candidate materials, accelerating the discovery of exotic physics and applications specific to quantum spin liquids. Atomically thin films are ideal candidate materials for realizing clean, long sought-after, Kitaev spin liquids. Here, a two-dimensional IrO6 honeycomb lattice is stabilized within a MnTiO3 ilmenite superlattice, inducing a suppression of antiferromagnetic order that suggests potential spin-liquid behavior.


Results
Structural characterisation using XRD. The films were grown on Al 2 O 3 (0001) substrates (lattice constants are a = 4.759 Å and c = 12.993 Å, JCPDS PDF 00-046-1212) by pulsed-laser deposition using a KrF excimer laser (see "Methods"). Our early attempt to grow ilmenite-type MnIrO 3 directly on Al 2 O 3 (0001) was unsuccessful and resulted in phase-separated films of Mn 3 O 4 and IrO 2 (see Supplementary Fig. 1 in Supplementary Information for the XRD pattern). We then employed a superlattice approach [16][17][18] based on ilmenite-type MnTiO 3 thin films 21,22 . In the superlattice structure, the common A-site MnO 6 plane is expected to promote the ilmenite-type stacking of MnO 6 and IrO 6 planes along the c-axis direction (Fig. 1a, b). The schematic structure of a typical superlattice is shown in Fig. 1c, where a MnTiO 3 buffer layer with a thickness ranging between 4.3 and 7.8 nm was initially grown on Al 2 O 3 (0001), followed by the alternate deposition of much thinner Mn−Ir−O and MnTiO 3 layers. A film thickness of 1.4 nm is approximately equal to the order of magnitude of typical c-axis parameters of ilmenite-type oxides. In a 0.5-nm-thick Mn−Ir−O (cap)/[4.3-nm-thick MnTiO 3 /0.5-nm-thick Mn−Ir−O] 15 /5.7-nmthick MnTiO 3 (buffer) film (superlattice cycle number n = 15 in Fig. 1c), the overall chemical composition of the film evaluated using energy-dispersive X-ray spectroscopy (EDX) was Mn:Ti:Ir = 50.4:45.1:4.5, which is in good agreement with the ideal composition of Mn:Ti:Ir = 50.0:44.9:5.1, as expected from the design thickness ratio. In the out-of-plane XRD pattern shown in Fig. 1d, four main diffraction peaks are observed from the film and are assigned to (0003m) (m: natural number) of the ilmenite-type structure. The satellite peaks near the (0006) and (00012)  In addition, the threefold symmetry of the ð10 14Þ diffraction peak was detected in the in-plane azimuthal ϕ scan measurements (Fig. 1f, g), indicating the epitaxial orientation relationship of the film ½10 10(0001)//Al 2 O 3 ½10 10(0001). These results indicate that the film grows as a c-axis-oriented, single-domain, ilmenite-type oxide with a periodically modulated internal structure.
We  15 superlattice. As shown in Fig. 2a, b, the films with d MIO = 0.5 nm (identical to the sample in Fig. 1) and 1.0 nm exhibit clear (0006) peaks, which are associated with satellite peaks (red arrows). In particular, for d MIO = 0.5 nm, thickness fringes appear near the (0006) peak, indicating that the total film thickness is uniform over the entire film. Considering the observed satellite peaks as superlattice reflections, the singleunit lengths of the superlattices were calculated to be 4.5 nm and 5.6 nm for d MIO = 0.5 nm and d MIO = 1.0 nm, respectively. This is consistent with the designed values of 4.8 nm (=4.3 + 0.5 nm) and 5.3 nm (=4.3 + 1.0 nm), respectively. However, in the thickest film with d MIO = 1.4 nm (Fig. 2c), superlattice reflections become indiscernible with the occurrence of a diffraction peak of segregated IrO 2 impurities; the designed superlattice structure is no longer formed for d MIO = 1.4 nm. The decrease in the (0006) diffraction intensity implies that the basal MnTiO 3 layers in the superlattice are disordered. Therefore, the upper bound of d MIO for stabilising the IrO 6 planes in the ilmenite lattice is~1.0 nm, which is smaller than typical c-axis parameters of ilmenite-type oxides (~1. Owing to the Z-contrast nature (Z: atomic number) 23 , the bright layers should contain a significant amount of Ir, which has the largest Z among the constituent elements (Mn, Ti, Ir and O). Tiny bright island-like regions are observed, which may be due to the presence of segregated Ir-rich impurities. As expected from the large lattice mismatch of~8% between MnTiO 3 and Al 2 O 3 and the observed relaxed growth (e.g., Fig. 1e), geometrical misfit dislocations were formed around the MnTiO 3 buffer/Al 2 O 3 substrate interface ( Supplementary Fig. 3). A thorough inspection of a single unit of the superlattice (Fig. 3b) revealed that a Mn−Ir −O layer between MnTiO 3 layers contains some bright atomic planes along the [0001] direction. Using the atomically resolved HAADF image in Fig. 3c (the area marked by yellow dashed lines in Fig. 3b), we compared each atomic site with a model structure (Fig. 3d). For MnTiO 3 , the MnO 6 (TiO 6 ) plane had a larger (smaller) atomic displacement along the [0001] direction between the in-plane Mn (Ti) ions 24 . By measuring the atomic displacement in the image, we determined that Ir atoms are located at the B-site, which is occupied by Ti atoms in MnTiO 3 , as schematically shown in Fig. 3d (also see in the left side of Fig. 3c). In addition, the bright Ir-containing planes are regularly sandwiched between the MnO 6 planes and have a similar dumbbell-like characteristic atomic arrangement. From these results, we concluded that a honeycomb lattice of edge-sharing IrO 6 octahedra crystallises in the Mn−Ir−O layer with ilmenite-type atomic ordering. Note that the Mn−Ir−O layer comprising two pairs of IrO 6 and MnO 6 planes corresponds to the 2/3-unit cell (u.c.) of the ilmenite lattice (Fig. 1a). We also performed electron energy loss spectroscopy (EELS) for the Mn L, Ti L and O K edges in the MnTiO 3 buffer and Mn−Ir−O layer regions ( Supplementary  Fig. 4). Although the Mn L-edge spectra of the two regions (i.e., Mn 2+ ) are in good agreement, a partial reduction of Ti 4+ to Ti 3+ occurred near the Mn−Ir−O layer 25,26 . Associated with this reduction, the presence of an oxygen deficiency is expected from the O K-edge spectra. We inferred that Mn 2+ and Ir 4+ are stable under the oxygen pressure of 10 mTorr used for the deposition, whereas Ti 4+ is reduced to Ti 3+ . Charge transfer between B-site TiO 6 and IrO 6 planes via A-site MnO 6 planes may be related to this point, although it is currently unclear.
Investigation of surface magnetic order using spin Hall magnetoresistance measurements. Based on these structural analyses, a honeycomb lattice of edge-sharing IrO 6 octahedra partly forms a two-dimensional network in a superlattice. In Ir 4+ -containing oxides, the delicate interplay of moderate electronic correlations and spin-orbit coupling determines whether the electronic structure of a material is metallic (gapless) [27][28][29][30][31][32] or insulating (gapped) 1,33,34 . Our superlattice samples were highly insulating (not shown), indicating the non-metallic electronic structure of Mn−Ir−O. Magnetism in ultrathin films, particularly for cases of antiferromagnetism and quantum spin liquids, is difficult to evaluate using bulk experimental methods 10,[13][14][15] . Thus, we studied the surface magnetic order using the spin Hall magnetoresistance (SMR) method, which enables electrical characterisation of ferromagnetic 35 and antiferromagnetic transitions [36][37][38] , as well as magnetic anisotropy. By measuring the SMR at the interface of Pt and MnTiO 3 ultrathin films, we recently demonstrated that bulk Néel temperature T N (~63 K) 39 and uniaxial magnetic anisotropy along the c-axis direction persist down to a film thickness of 2.9 nm~2 u.c. with six ("Methods"), we simultaneously measured the resistances of the two orthogonal Pt channels (i.e., R 1 for channel 1, and R 2 for channel 2) under an in-plane magnetic field H applied parallel to channel 1. Figure 4b-d show the temperature (T) dependence of the fieldinduced resistance variation, ΔR = (R 1 /R 2 ) 0.5 T − (R 1 /R 2 ) 0 T . Here, (R 1 /R 2 ) 0.5 T and (R 1 /R 2 ) 0 T are the resistance ratios measured at μ 0 H = 0.5 T and 0 T (where μ 0 is the vacuum permeability), respectively. If an antiferromagnetic transition occurs in the layer beneath the Pt layer, distinct SMR responses of R 1 and R 2 below and above T N result in an inflection in ΔR (refs. 22,36,37 ). In fact, the value of ΔR for both the thick-MnTiO 3 monolayer (Fig. 4b) and trilayer C (purple in Fig. 4d) exhibits a sudden increase near T = 60 K upon heating, wherein bulk MnTiO 3 undergoes an antiferromagnetic-to-paramagnetic transition 36     Extrinsic contributions, such as the surface roughness and the effect of size (refs. [40][41][42], should also be considered to understand the suppression of antiferromagnetic order in thin-MnTiO 3 (« 2 u.c.) and Mn−Ir−O top layers. Firstly, all measured samples had smooth surfaces with root-mean-square roughness values ranging from 0.2 to 1.0 nm, which cannot exclusively account for the distinct SMR responses. Furthermore, despite its island-like film morphology with increased surface roughness, the 2-u.c.thick thin-MnTiO 3 monolayer exhibits a bulk-like T N as in thicker and flatter samples (≥ 3 u.c.) 22 , supporting the negligible role of the surface roughness. Secondly, the total film thicknesses of the heterostructures (including MnTiO 3 and Mn−Ir−O layers) are significantly thicker than the 2 u.c.; therefore, a sufficiently large volume of the Mn sublattice is guaranteed for the entire heterostructure. This also helps minimise the potential influence of domain disconnection, which is generally pronounced in ultrathin films. In stark contrast, there is a systematic recovery of antiferromagnetic SMR responses with an increase in the b-d Field-induced variation (μ 0 H = 0.5 T) in the resistance ratio of two orthogonal Pt channels, R 1 /R 2 , for b, the thick MnTiO 3 , c the bilayer and d trilayers A, B and C. In the thick-MnTiO 3 monolayer (Fig. 4b) and trilayer C (Fig. 4d), SMR responses indicating antiferromagnetic transition were detected. The inset in b shows the measurement setup, where R 1 and R 2 are measured simultaneously under an in-plane magnetic field H applied parallel to channel 1 with R 1 (perpendicular to channel 2 with R 2 ). E 1  thickness of the MnTiO 3 top layer from trilayer A (2/3 u.c.), B (4/ 3 u.c.) and C (3 u.c.), which indicates the intrinsic origin that is responsible for spin interactions in a characteristic length. Notably, the clear difference observed between the antiferromagnetic 2-u.c.-thick thin-MnTiO 3 monolayer 22 and the non-antiferromagnetic trilayer B (4/3-u.c.-thick MnTiO 3 and 2/3u.c.-thick Mn−Ir−O (2 u.c. in total) on the MnTiO 3 buffer) indicates that the insertion of honeycomb-lattice IrO 6 planes affects the spin interactions between nearby Mn sites. Such a spin-disordered (or spin-frustrated) state may be associated with quantum spin liquids, although the experimental identification of the featureless magnetic ground state in ultrathin films presents a great challenge [43][44][45] . The next step is to understand the spin structures (including the Ir sites) and the excited-state properties in Mn−Ir−O using recently advanced diagnostics such as Raman spectroscopy 1,5,46 and in-plane spin transport measurements [47][48][49][50] . In addition, the structural quality of a sample is important for the formation of pure Kitaev materials. In particular, the interface roughness (Fig. 3b), oxygen vacancies ( Supplementary Fig. 4) and possible inter-diffusion can be the source of local structural distortions that destroy Kitaev-type interactions via the non-ideal splitting of Ir 5d orbitals 20 . Nevertheless, the suppression of the strong antiferromagnetic order in the Mn sublattice is promising for demonstrating the feasibility of controlling spin interactions by artificially engineered ilmenite-type oxides.
In summary, we have materialised a honeycomb lattice of edgesharing IrO 6 octahedra by using a superlattice formation with ilmenite-type MnTiO 3 . Systematic SMR measurements indicated the absence of antiferromagnetic order in the surface Mn−Ir−O layer grown on antiferromagnetic MnTiO 3 . Although A-site MnO 6 planes are common to these oxides, the spin interactions between the Mn sites are different. Spin fluctuations induced by strong spin-orbit interactions in the IrO 6 planes can disrupt the antiferromagnetic ordering in the Mn−Ir−O layer and the neighbouring regions of the MnTiO 3 layers. The stabilisation of a two-dimensional IrO 6 honeycomb lattice using the superlattice technique, as well as the potential control of the magnetism via dimensionality and the proximity effect, is expected to trigger the development of ilmenite-based Kitaev materials that produce exotic physical phenomena.

Methods
Thin-film growth. The Mn−Ir−O target was prepared from MnO 2 and IrO 2 powders by spark-plasma-sintering at 50 MPa and 900°C. The target composition, as measured by EDX, was Mn:Ir = 0.90:1 (atomic ratio). The MnTiO 3 buffer layer was grown at a substrate temperature of 850°C and an oxygen pressure of 10 mTorr using the Mn−Ti−O target 22 . Subsequently, Mn−Ir−O and MnTiO 3 layers were alternately deposited for n cycles at 800°C and 10 mTorr for the formation of a [Mn−Ir−O/MnTiO 3 ] n superlattice. The crystal structure and composition of the films were characterised by XRD using Cu K α radiation and EDX, respectively. SMR measurements. A Pt film with a thickness of~2 nm was deposited on the film surface by radio-frequency magnetron sputtering at 150°C. The heterostructure film was then patterned into an L-shaped multiterminal device structure using photolithography and Ar ion milling, followed by the electron-beam evaporation of Au/Ti electrodes. The resistance was measured by the four-probe method using a semiconductor parameter analyser (Agilent 4155C) and a nanovolt metre (Keithley 2182A) in a physical property measurement system (Quantum Design, Inc.) equipped with a one-axis sample rotator. The direction (polarity) of the current I was varied during the measurements, and the averaged resistance values, R j ¼ R j þI ð ÞþR j ÀI ð Þ 2 ðj ¼ 1 and 2Þ, were used for the analysis. Details of the measurement scheme and analysis are reported in ref. 22 .
Electron microscopy. To obtain an electron-transparent thin specimen, the thin film with the substrate was mechanically polished, and Ar ion beam milling was performed at 0.5 kV in the final stage. For the atomic and electronic structure analyses, an aberration-corrected STEM system (JEOL ARM300CF) equipped with a DELTA corrector, cold-field emission gun and EELS spectrometer (Quantum, Gatan, Inc.) operated at 300 kV was used. The probe-forming aperture was 30 mrad, and the collection semi-angle for HAADF was 85-200 mrad.

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