The co-existence of strong spin–orbit coupling (SOC) and electron–electron correlations has triggered numerous novel quantum states in 5d transition metal iridates.^[1–7] Among perovskite iridates, the n = ∞ end-member SrIrO₃ of the Ruddlesden–Popper series Sr_n+1Ir_nO_3n+1, has attracted growing attention due to its possible exotic topological phases.^[8–14] In situ angle-resolved photoemission spectroscopy (ARPES) measurements have unveiled a Dirac cone in the bulk state of epitaxially grown (001)-SrIrO₃ films,^[8,9] which is reminiscent of a symmetry-protected nodal line predicted by theoretical calculations.^[11,12] Furthermore, a recent theoretical work based on the tight-binding model predicted that orthorhombic perovskite SrIrO₃ is a potential topological crystalline metal with zero-energy surface states protected by the mirror-reflection symmetry in the (110) plane.^[15] However, although the synthesis of (001)-SrIrO₃ using both pulsed laser deposition (PLD)^[16] and reactive molecular beam epitaxy (MBE) techniques^[8,9] have been widely reported, the epitaxial growth of (110)-SrIrO₃ thin films still suffers from severe technical barriers, and films of both high crystalline and surface quality are sill lacking, which is actually the prerequisite for further verifying its topological non-triviality.

In this work, we comprehensively investigate the synthesis process of (110)-SrIrO₃ thin films by reactive MBE and overcome the difficulty in the preparation of such polar multicomponent oxides. In situ reflection high-energy electron diffraction (RHEED) monitoring and ex situ x-ray diffraction (XRD) test both demonstrate the desired [110] orientation of the SrIrO₃ films as well as the high-quality crystallization. In addition, by comparing with the [001]-oriented counterpart using electrical transport measurements, (110)-SrIrO₃ exhibits more pronounced semimetallic behavior at temperatures from 2 K to 300 K.

Reactive MBE is considered to be the most significant growth method for perovskite oxides to date, providing ultra-high orientation and pure phase as well as atomic-level flatness. (110)-SrIrO₃ were epitaxially grown on (110)-SrTiO₃ substrates by co-deposition mode, which follows the thermodynamics mechanism^[17] and can automatically control the composition with all elements’ shutters being open during the growth process. In general, commercial substrates tend to undergo some pretreatments before film growth. Here, the mixed cut-off (110)-SrTiO₃ substrate was ultrasonically cleaned with acetone and dehydrated ethanol for 10 minutes each in the atmosphere, then transfered into a loadlock chamber with 200 °C heating for more than 2 hours to obtain a clean surface. In the growth chamber, under a distilled O₃ background pressure of 2.5 × 10⁻⁶ Torr (1 Torr = 1.33322 × 10² Pa), the substrate temperature was set to 650 °C. It is worth noting that once the substrate temperature is higher than 300 °C, a continuous oxidizing agent is needed to compensate for the oxygen loss of the substrate. Strontium and iridium atoms were respectively evaporated from a low-temperature effusion cell and electron beam evaporator with a precise flux ratio of 1:1 measured by a quartz crystal microbalance.

Perovskite SrIrO₃ (Pbnm) can be seen in the [001] direction as alternating layers of electrically neutral (SrO)⁰ and (IrO₂)⁰ with the layer spacing of a/2 = 1.98 Å, making (001)-SrIrO₃ a non-polar oxide. On the contrary, (110)-SrIrO₃ is identified as a polar oxide due to the stacks of oppositely charged layers (SrIrO)⁴⁺ and (O₂)⁴⁻ [Fig. 1(c)], which are seperated by Å. If a non-polar substrate is utilized to grow a polar film, the intrinsic polarity discontinuity of these two materials will induce polarity catastrophes,^[18,19] directly affecting the nucleation in the initial growth period and leading to complicated unexpected reconstruction. As a result, the interface and surface of the film become quite rough and multiphase co-existence occurs. The lattice constants of SrTiO₃ and SrIrO₃ along [ ] are 5.52 Å and 5.60 Å, respectively. Therefore, under the premise of minimizing lattice mismatch (1.4%), we consider the polar (110)-SrTiO₃ as the substrate to eliminate polarity catastrophes. In order to improve the interface flatness, homo-epitaxial 20-unit-cell (20 u.c.) SrTiO₃ buffer layers were deposited first [Fig. 1(b)].

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (color online) (a) Crystal structure of perovskite SrIrO3. (b) Schematic view of the as-synthesized heterostructure. (c) Schematic diagram of interface polarity between the SrTiO3 buffer layer and SrIrO3 along the [110] direction.

Figures 2(a)–2(c) show in-situ RHEED monitoring of the growth process. With the deposition of buffer layers, the intensity of (01) diffraction spot of (110)-SrTiO₃ substrate gradually increases, at the same time, another diffraction spot (02) appears (marked with a blue square), which indicates that the possible grooves on the surface of the substrate have already been filled, laying a strong foundation for subsequent high-quality film growth. It is worth noting that once we begin to grow SrIrO₃ on the as-prepared buffer layers, a remarkable streak [marked with red square in Fig. 2(c)] could be clearly observed between (00) and (01) diffraction spots. The intensity of this extra secondary diffraction increases as more layers of SrIrO₃ were deposited. Meanwhile, we rotated the sample manipulator by 90°, no other diffraction features were seen in the [001] direction, uncovering a (1 × 2) reconstruction on the surface of the as-grown SrIrO₃ films.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (color online) RHEED pattern before deposition (a), after depositing buffer layers (b), and after depositing 20 u.c. SrIrO3 films (c). (d) Diffraction intensity oscillation of buffer layers. (e) Comparison of integration curves over different periods. (f) Ideal surface of (110)-SrIrO3. (g) (1 × 2) reconstruction on the surface of the as-grown SrIrO3 films.

Due to the polarity of the (110) perovskite materials, an infinite dipole moment exists perpendicular to the surface, causing the pristine surface to be unstable and outermost atoms rearranged spontaneously.^[20] For (110)-SrTiO₃, slight annealing can lead to various types of reconstruction,^[21] and the (1 × 2) reconstruction results from desorption of a large amount of atoms on the Ti-terminated surface.^[22] Despite the non-negligible instability in (110)-SrTiO₃, the periodic oscillation curve [Fig. 2(d)] sufficiently declared the high crystalline quality of the buffer layers without any unfavorable reconstruction. However, there was no noticeable oscillation during the deposition of (110)-SrIrO₃ films, which is likely to be the cause of the surface reconfiguration with the distance between adjacent Ir atoms expanding from 5.52 Å to 11.04 Å [Figs. 2(f)–2(g))]. Periodic loss of considerable atoms leads to the appearance of vertical microfacets (100)/(010) IrO₂ on the surface. Hence, compared to (110)-SrTiO₃, (110)-SrIrO₃ seems more difficult to obtain an ideal surface. What is more, the atomic force microscopy (AFM) image indicates the mean roughness of (110)-SrIrO₃ films over an arbitrarily selected 2 μm segment is about 0.5 nm [Figs. 3(a)–3(b)]. Such a value is nearly equivalent to the thickness of 2-u.c. atoms, exactly in line with the (1 × 2) reconstruction result.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) (a) AFM image of SrIrO3 film. (b) The mean roughness over 2 μm is about 0.5 nm. (c) XRD pattern of SrIrO3 film. (d) T-dependent ρ curves for (001)-SrIrO3 and (110)-SrIrO3.

In order to investigate the lattice structure of the samples, we next conducted an ex situ XRD test. As seen in Fig. 3(c), apart from the diffraction peaks (green circles) of the substrate corresponding to (110) and (220) planes, there are very distinct diffraction peaks (green triangles) of the epitaxial films on the left side without any other miscellaneous peaks, which implied excellent single crystal orientation. Using the Bragg diffraction formula, we can calculate the lattice constant of 5.52 Å in the [ ] direction of SrIrO₃ films. Moreover, the rocking curve measurement for the (220) peak is shown in the inset of Fig. 3(c). The full width at half maximum obtained after fitting by Gaussian is about 0.06°, which is comparable with 0.05° of the substrate, indicating the good crystal quality of our epitaxially grown films.

A comparison between (001)-SrIrO₃ and (110)-SrIrO₃ on electrical transport is shown in Fig. 3(d), from which the slope for resistivity versus temperature of the (110) oriented film experienced a transition from negative to positive, reaching a minimum value around 170 K. This upturned resistivity feature can be frequently observed not only in SrIrO₃ thin films^[23–27] but also in SrIrO₃ crystals,^[28] possibly derived from weak localization effects^[29] or pseudogap formation.^[28] The ratio of ρ_min and ρ_max in (110)-SrIrO₃ is approximately equal to 0.9, such a narrow resistance range implies semimetallic properties.^[26,30] When the temperature drops to 2 K, the resistivity of (110)-SrIrO₃ is only 55% of the (001)-oriented counterpart and hits 70% at 300 K. Therefore, (110)-SrIrO₃ indeed conducts more pronounced semimetallic behavior than (001)-SrIrO₃ throughout the temperature range, which may result from the unique surface states protected by lattice symmetry of the (110)-oriented films.

The In situ RHEED patterns combined with ex situ XRD tests complementarily reveal the high-quality crystallization of hetero-epitaxy (110)-SrIrO₃ films. However, the evolution of RHEED fringes also reveals surface reconstruction in the [ ] direction. Even if we tried to anneal these samples at 650 °C or higher temperatures for 30 minutes under an ozone pressure of 2.5 × 10⁻⁶ Torr, or reduced the deposition rate by adjusting the temperature of each source synchronously, this phenomenon still exists. In addition, the (110)- and (001)-oriented epitaxial films suffer from different stress from the substrate, which possibly affects the interaction between SOC and electron-correlation, may lead to new physical phenomena.

In summary, we have successfully synthesized high-quality (110)-SrIrO₃ thin films with effective polar buffer layers utilizing MBE for the first time. The precise orientation and excellent crystallinity have been confirmed by both RHEED monitoring and XRD test. A (1 × 2) reconstitution was observed simultaneously, which is originating from an increased lattice constant in the [ ] direction. Compared with (001)-SrIrO₃, (110)-SrIrO₃ displays more obvious semimetallic behavior below 300 K, in good agreement with the theoretical prediction. Our research provides a valuable reference for the quality enhancement of polar oxide films. More importantly, it lays a solid foundation for the research on mutual competition between SOC and strong electron-correlation in perovskite iridates. More electronic structure investigations such as ARPES studies are expected to confirm the hypothetical non-trivial topological states.