Fabrication of (111)-oriented Ca0.5Sr0.5IrO3/SrTiO3 superlattices; a designed playground for honeycomb physics

We report the fabrication of (111)-oriented superlattice structures with alternating 2m-layers (m = 1, 2, and 3) of Ca0.5Sr0.5IrO3 perovskite and two layers of SrTiO3 perovskite on SrTiO3(111) substrates. In the case of m = 1 bilayer films, the Ir sub-lattice is a buckled honeycomb, where a topological state may be anticipated. The successful growth of superlattice structures on an atomic level along the [111] direction was clearly demonstrated by superlattice reflections in x-ray diffraction patterns and by atomically-resolved transmission electron microscope images. The ground states of the superlattice films were found to be magnetic insulators, which may suggest the importance of electron correlations in Ir perovskites in addition to the much discussed topological effects.


Introduction
The technical advances in the fabrication of oxide heterostructures on an atomic level have opened up a new avenue for exploring new materials by design, where the presence of interfaces may give rise to a rich variety of distinct electronic phases [1][2][3][4]. It was recently proposed that, by providing spatial constraint using heterostructure, a geometrical motif hidden in the underlying lattice can be isolated and generate topological phases [5][6][7][8]. If the two-layer units of perovskite-type transition-metal oxides (TMOs), ABO3 (B = transition metal), are isolated along the [111] crystallographic axis by forming a heterostructure with an insulating spacer, they can be viewed as a buckled honeycomb lattice (Fig. 1) of transition metal B ions [5]. This bilayer of perovskite TMOs has a natively inverted band structure due to the geometry effect of the honeycomb lattice, which is similar to the situation in graphene [9]. With sizable spin-orbit coupling, a non-trivial band topology should be realized.
(111)-oriented superlattices with bilayers of perovskite containing heavy 5d transition metals were therefore proposed to be promising candidates for topological matter.
Realization of (111)-oriented perovskite superlattices has been reported for only a few 3d systems because of the apparent technical difficulties [10][11][12][13][14][15]. Here, we report the successful fabrication of (111)-oriented superlattices with 2m (m = 1, 2, and 3) unit cells of Ca0.5Sr0.5IrO3 and 2 unit cells SrTiO3 [(CSIO2m, STO2)k], grown epitaxially on STO(111) substrates by pulsed laser deposition. A solid-solution Ca0.5Sr0.5IrO3 was used to stabilize (111)-oriented perovskite-type iridate films at the atomic level. Superlattice reflections observed in x-ray diffraction patterns and atomically-resolved transmission electron spectroscopy images indicate atomic ordering of Ir 4+ and Ti 4+ as designed. The (111)-oriented bilayer of perovskite SrIrO3 has been discussed as a candidate for a topological insulator [5]. Under our experimental conditions, however, (111)-oriented perovskite SrIrO3 was not stabilized on the STO (111) substrate. The θ-2θ XRD scans for SrIrO3 films deposited on STO(111) substrates ABO3 oxides have been known to exhibit a sequential transformation of crystal structure as a function of the size of the A-site cation or applied pressure.
The evolution of the crystal structure can be described by the change in the stacking pattern of the two types of layer with corner-sharing and face-sharing BO6 octahedra. In general, the ratio of number of corner (C) to face-sharing (F) layers in the stacking (C/F) increases as 2H(0) → 9R(1/2) → 6H(2/1) → perovskite(∞), with replacing smaller A-site cations or increasing pressure. This trend suggests that an isoelectronic compound CaIrO3, with the smaller Ca 2+ ion rather than Sr 2+ on the A-site, could have a better chance of adopting a perovskite structure on STO(111) in contrast to the case for SrIrO3.
The growth of perovskite CaIrO3 films was attempted on STO (111) substrate. Perovskite CaIrO3 was in fact stabilized, which is evidenced by the θ-2θ XRD patterns, as shown in Fig. 2 implying that the majority of the CaIrO3 film is coherently strained on the substrate. The anisotropic shape of the Bragg spot, however, suggests a partial strain relaxation at the film surface. The nominal lattice mismatch between perovskite CaIrO3 (apc ~ 3.855 Å; pseudo-cubic notation) and STO (a = 3.905 Å) is relatively large (1.3%), which may result in a partial strain relaxation. Since the epitaxial strain can be released by a large surface deformation, the surface of strained films becomes unstable. This is known as the Asaro-Tiller-Grinfeld instability [19][20][21]. Indeed, no thickness fringe was seen around the 111 reflection of the CaIrO3 film in the θ-2θ XRD scan [ Fig. 2(b)], which indicates that film thickness is not well-defined due to the disordered surface. Atomic force microscopy observations (not shown) support the hypothesis of significant surface roughness of the CaIrO3 films grown on STO(111).
In order to obtain the atomically flat surface required to realize a superlattice structure, we attempted to reduce the lattice mismatch between iridium perovskite and the STO substrate by making a solid solution of CaIrO3 and SrIrO3. A solid solution of CaIrO3 and SrIrO3 can be formed only in a limited composition range by solid-state reaction [22]. In the case of epitaxial thin-film growth, however, the solid solution was reported to exist for any composition [23].
With partial Sr substitution for Ca, the unit cell volume of perovskite Ca1-xSrxIrO3 was increased. The best match of the lattice constant of Ca1-xSrxIrO3 with that of STO was achieved at around x = 0.5. As shown in Fig. 2 The superlattice peaks indicate the periodicity of (2m + 2)d, where d is the (111) interlayer distance of 2.25 Ǻ, evidencing that the designed layering was achieved.
The ordering of Ir 4+ and Ti 4+ ions in the m = 1 (bilayer of Ir) superlattice is visualized by the [-110] zone axis HAADF-STEM image with strong atomic number (Z) contrast presented in Fig. 3(b). A stripe modulation of Ir 4+ (bright spots) and Ti 4+ (dark spots) represents an alternate stacking of CSIO and STO bilayers along the [111] direction. In the intensity scan in a column along the [001] direction, shown in Fig. 3(c), pairs of Ir 4+ (high intensity peaks) and Ti 4+ (low intensity peaks) ions emerge alternately as expected for a m = 1 superlattice.
From these structural data, we conclude that the (111) bilayer of Ir, forming a buckled honeycomb lattice, and its higher harmonics were obtained in the coherently grown superlattice. The results are summarized in Fig. 4(a), where metal-insulator transition as a function of number of CSIO bilayers can be seen. The CSIO only, m = ∞, (111) film showed a poorly metallic behavior of resistivity ρ(T), almost temperature independent and with a magnitude ~ 1 mΩcm. This behavior is essentially the same as observed in bulk and thin-film samples of perovskite CaIrO3 [23,24] and SrIrO3 [25,26], consistent with a semimetallic ground state due to the presence of symmetry-protected Dirac nodes [27,28]. perovskite is on the verge of an orbital selective topological Mott transition and that a reasonable Coulomb U ~ 2 eV switches a topological insulator to a trivial insulator with the help of magnetic ordering [30]. This may provide a reasonable explanation for our observations on CSIO superlattices. In this context, it is tempting to investigate the suppression of the magnetic ordering, for example, by physical pressure to switch the system back to the topological state.
In conclusion, by optimizing the size of the alkaline earth ions, superlattice structures with alternating 2m (m = 1, 2, and 3) layers of (111)-oriented Ca0.5Sr0.5IrO3 perovskite and bilayer of SrTiO3 were successfully stabilized on STO(111) substrates. The structural characterization by XRD and STEM indicated that the superlattice structures were well controlled on an atomic level. The ground states of the superlattice films were found to be magnetic insulators. This may imply the importance of electron correlations which were proposed to compete with the topological effect expected from the honeycomb-based lattice structure of Ir in (111) superlattices. This study has established experimentally that (111) perovskite superlattice structures can be a realistic playground for exploring the interplay of electron correlations and topological effects.