Solid-phase epitaxy of a CuAlO₂ template on c-Al₂O₃ for delafossite growth

Delafossite oxides with the chemical formula ABO₂ are materials with fascinating physical properties due to the alternate layer stackings of A-cation layers with BO₂ layers.^1–5 Depending on the A- and B-cations, physical properties can range from ultraclean metals (i.e., A = Pd and Pt) or transparent semiconductors (i.e., A = Ag and Cu and B = Al, Cr, Fe, Ga, Y, and In) to complex magnetic ground states.^2–5 To explore their distinctive properties, honing the material architecture as thin films will have great potential in designing heterostructures by combining different ABO₂ materials, which currently are in uncharted territory. One major drawback for ABO₂ thin film research is the lack of commercially available lattice-matched substrates; currently no known delafossite substrates are available. Current substrates that have been used for ABO₂ growth include triangular substrates, including c-Al₂O₃, β-Ga₂O₃, (111) SrTiO₃, (111) MgO, and (001) 4H–SiC.^4–14 The structural and chemical dissimilarity and huge lattice mismatch between the available substrates and delafossite thin films limit the successful epitaxy and film quality. Therefore, advances are required to enable novel delafossite-based materials and their applications for oxide electronics. We recently reported the vital role of Cu-based delafossite buffer layers in growing Pd-based delafossites on commercially available single-crystal substrates.^9,10 Such progress offers a step toward advanced materials, enabling the deliberate fabrication of artificial heterostructures, with enhanced electronic and magnetic properties.

Template engineering or buffer layers have been commonly used in various oxide heterostructure syntheses (i.e., perovskites) to achieve epitaxial film growth.^15–17 Even in the thin film growth of delafossite, a templated CuCrO₂ layer significantly reduced the formation of impurity phases.¹⁰ Compared to the CuCrO₂ template, CuAlO₂ is an ideal system due to its nonmagnetic and transparent properties. In this study, we focus on fabricating the CuAlO₂ structure on the surface of the commercially available c-Al₂O₃ substrate through solid-phase epitaxy.

In literature, Cu–Al–O phase diagrams were reported by mixing starting reagents of Cu₂O or CuO with Al₂O₃, followed by thermal treatments.^18–21 Earlier reports also showed that CuAlO₂ thin film layers could be formed at the interface when a layer of Cu₂O is chemically reacted with Al₂O₃.^12–14,22 There have been reports of a two-step process to form CuAlO₂: one used Cu evaporated on an Al₂O₃ substrate (CuO is the starting precursor),¹² and another sputtered a Cu₂O layer on a sapphire substrate and covered with an Al₂O₃ layer (Al₂O₃|Cu₂O|sapphire).^13,14 By using Cu|sapphire, it was observed that the formation of CuAlO₂ also included unreacted CuO, and in the sandwich method (Al₂O₃|Cu₂O|sapphire), the top reaction with the Al₂O₃ film would not lead to epitaxy, indicating that Cu₂O reacts with the substrate interface to produce the epitaxial structure.^12–14 Here, we investigate how one can create such an epitaxial template of CuAlO₂ on Al₂O₃ substrates with minor defects to be explicitly used for ABO₂ film growth thereon. We start with the following chemical reaction, as shown in Eq. (1):

We used a “bake-method” technique to fabricate the CuAlO₂ surface layer.^12–14,23 By characterizing the films with x-ray diffraction (XRD) and high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) measurements, we confirmed the growth of the CuAlO₂ structure on c-Al₂O₃ substrates. Moreover, we demonstrate the epitaxial growth of PdCrO₂ on a templated CuAlO₂ on a c-Al₂O₃ substrate. PdCrO₂ films were reported in earlier studies,^10,11 where PLD-grown films had a rate-limiting effect, causing thicker films (>33 nm) to form impurity phases. Utilizing the templated CuAlO₂ structure on c-Al₂O₃ has negated the thickness limitation. It displays enhanced structure (no secondary phases observed) and similar physical properties as measured by temperature-dependent resistivity in contrast to other PdCrO₂ films.

A polycrystalline Cu₂O target was sintered by solid-state reaction and used for PLD. c-Al₂O₃ substrates were cleaned by rinsing in acetone, isopropanol, and deionized water for 10 min each, then were annealed at 1000 °C for 1 h in a box furnace. Substrates were then ready to be used for deposition. The Cu₂O thin films were grown on the prepared substrates by PLD using a KrF excimer laser (λ = 248 nm). Substrate growth temperature and oxygen partial pressure during the polycrystalline Cu₂O growth were 700 °C and 10 mTorr at a laser repetition rate of 5 Hz. Thin films were cooled to room temperature under the growth pressure. For post-growth treatment, the Cu₂O films were placed in an alumina boat by capping with a bare Al₂O₃ substrate to protect the surface and suppress the evaporation of Cu₂O.¹³ The furnace temperature, T, and dwell times, t, were varied in this study to find the best thermal treatment conditions. The samples were annealed with different reaction temperatures ranging from 1000 to 1300 °C, and dwell times ranged from 30 min to 5 h. Atomic force microscopy (AFM) was used to scan the surface morphology in each step. XRD was used to measure wide-angle diffraction patterns during each cycle process. HAADF-STEM experiments were performed using a Nion UltraSTEM200 operated at 200 kV. The microscope is equipped with a cold-field emission gun and a corrector of third- and fifth-order aberration for sub-Angstrom resolution. The convergence half-angle of 30 mrad was used, and the inner angle of the HAADF-STEM was ∼65 mrad. Resistivity was measured as a function of temperature with a Physical Properties Measurement System (PPMS) for a PdCrO₂ film on a CuAlO₂/c-Al₂O₃ substrate.

Figure 1(a) displays our approach to forming epitaxial CuAlO₂ through solid-phase epitaxy. First, we discuss the two-step process in Fig. 1 by (1) depositing Cu₂O by PLD and then (2) annealing at a high temperature (>1000 °C) to activate the solid-state interface reaction between Cu₂O (film) and Al₂O₃ (substrate), depicted by the colors in blue and yellow, respectively. CuAlO₂ begins to form during the solid-state reaction, as depicted by the green color. As the simple cartoon picture demonstrates, when undergoing a dwell time at a high temperature, all of the Cu₂O (blue) eventually reacted with Al₂O₃ (yellow), forming CuAlO₂ (green). AFM micrographs of a bare Al₂O₃, a PLD deposited Cu₂O polycrystalline film (8 nm in thickness), and a CuAlO₂ film after the solid-state reaction at 1100 °C for 1 h are also shown in Fig. 1(b). Remarkably, the surface of a CuAlO₂ thin film is atomically flat with a well-defined step-and-terrace structure. The average step height is ∼5.26 Å, around one-third of the lattice parameter of the CuAlO₂ unit cell stacked with three monolayers (c = 16.958 Å). This finding suggests that the surface roughness consists of a CuAlO₂ monolayer. We further noted that the surface of CuAlO₂ thin film becomes rough when a thicker Cu₂O is used, which will be discussed later.

Figure 2(a) shows an as-grown film of 800 nm thick Cu₂O with XRD peaks revealing the polycrystalline phase used to make the sample measured in Figs. 2(b)–2(e). Figure 2(b) displays the 0003L peaks confirming the CuAlO₂ phase without the CuO₂ phase prepared using the reaction temperature at 1100 °C with a dwell time of 1 h. The full width at half maximum (FWHM) of the rocking curve of the 0006 film peak in Fig. 2(c) represents the high crystallinity. The other reaction temperatures, shown in Fig. S1 in the supplementary material, show that at ∼1000 °C Cu₂O films transformed to the CuAl₂O₄ phase, and above 1200 °C, the deterioration was confirmed by vanishing the film peaks. To add, Fig. S2 shows XRD scans of the phase evolution performed to find the optimal condition as a function of the dwell time at the annealing temperature of 1100 °C ranging from 30 min to 5 h. Dwell time greater than 3 h showed other impurity peaks with overall film quality deteriorated, confirmed by vanishing the film peaks at dwell times greater than 4 h. Rocking curves of the 0006 film peak were measured as shown in Fig. 2(c), and the corresponding FWHM was analyzed and shown in Fig. 2(d) for the samples annealed at 1100 °C with varying dwell times.

Under the optimum condition, CuAlO₂ films are oriented along the c axis, indicating that the films are stabilized by exposing a (0001)-oriented surface or forming a (0001)-oriented interface. Figure 2(e) shows a HAADF-STEM image of a CuAlO₂ thin film formed at the optimum annealing condition (T = 1100 °C and t = 1 h). Notably, a sharp interface is observed between the CuAlO₂ film and c-Al₂O₃ substrate with atomic-scale reconstructions at the interface. Specifically, an extra AlO_x [denoted by the yellow arrow in Fig. 2(e)] layer, different from the Al₂O₃ layers, was formed at the substrate surface. Then, AlO₂ and Cu sublayers are stacked alternatively to constitute the delafossite phase; the first AlO₂ sublayer of the film is denoted by the green arrow in Fig. 2(e). Such a characteristic atomic structure must lower the interfacial energy between the thin film and substrate, which may enable the formation of atomically flat CuAlO₂ thin films oriented along the [0001] direction.

A phase diagram is constructed in Fig. 3 for the solid-state reaction of 800 nm Cu₂O polycrystalline films on c-Al₂O₃ substrates used under different temperatures and dwell times. The data points are from samples analyzed by XRD. Cu₂O and CuO are shown in the phase diagram, as reported elsewhere²², demonstrated by in situ high-temperature XRD scans of a Cu₂O film on Al₂O₃, which oxidizes into CuO at 300 °C. Based on the rocking curve FWHM of the 0006 peak of CuAlO₂ thin films and the phase diagram, the optimal growth regime resides at 1000 < T < 1200 °C and 0.5 < t < 3 h.

Although XRD patterns of the thin films grown under the optimal conditions show single-phase CuAlO₂, we find that its surface morphology highly depends on the thickness of the pre-deposited chemical reagent Cu₂O. To understand how the thickness of a Cu₂O film affects the interfacial surface chemistry, Fig. S3 displays AFM micrographs when varying the thickness of Cu₂O films (note that due to the rough and grainy surface, it is hard to determine the film thickness precisely. Thus, we approximately estimate the nominal thickness to be around 8, 80, 130, and 270 nm). The AFM taken after the reaction from Fig. S3(a) was shown in Fig. 1(b), where the surface of CuAlO₂ maintains flat step terraces. Due to the thickness limitation, we find that the thinner layer of Cu₂O used resulted in undetectable film peaks in our XRD measurements. However, the surface morphology characterized by AFM showed flat step terraces. The difference in the number of pulses deposited by PLD plays a role in CuAlO₂ film growth. As the Cu₂O film thickness is increased to ∼80 nm, the steps become wavy, as shown in Fig. S4. We also note that shiny specks on the surface, which are not detected in XRD, are often observed in thick CuAlO₂ thin films. We further note that a large portion of Cu₂O films was re-evaporated without participating in the reaction on the surface; so, the final CuAlO₂ thickness is much smaller. For about 8 nm Cu₂O, we have observed the formation of only a 1–2 unit cell thick CuAlO₂ film. The loss of Cu layers was also observed previously. Thus, this requires further study to better understand the underlying mechanism for the disappearance of Cu.¹⁰ 

The obtained CuAlO₂ thin films with an atomically flat surface can be utilized as a template for the growth of metallic delafossite. A PdCrO₂ thin film was grown by PLD on a templated CuAlO₂ under the optimal growth condition for PdCrO₂ (T = 650 °C, P_O = 200 mTorr).¹⁰ CuAlO₂ provides a small lattice mismatch of −2.4% and offers the structural similarity between PdCrO₂ and CuAlO₂. These conditions play a crucial role in stabilizing a good-quality PdCrO₂ thin film. More importantly, the templated CuAlO₂ layer enables the growth of very thick films (≈70 nm). We noted that such thick PdCrO₂ films could not be grown previously because many impurities started to form above a certain thickness (∼30 nm). Thick films of delafossite would allow opportunities for various experiments, such as neutron scattering, magnetization, and specific heat, that can provide ways to deepen our understanding of delafossite thin films and heterostructures.

Figure 4(a) shows an XRD pattern for a 70 nm thick PdCrO₂ thin film where only delafossite 0003L peaks are resolved. By comparing the FWHM of the rocking curve with our previous PdCrO₂ grown on CuCrO₂,¹⁰ we found that the quality of our film was poorer (Δω ∼1.4°). We attribute this to the intrinsic nature of the solid-phase epitaxy and the unavoidable exposure of our sample to air for thermal annealing in a furnace. To add, we observed thicker PdCrO₂ grown on CuAlO₂ resulted in an increase in FWHM. Temperature-dependent resistivity of the 70 nm thick PdCrO₂ film showed a metallic behavior with a residual resistivity ratio, RRR ≈ 1.4, as shown in Fig. 4(b). The RRR value is comparable to PdCrO₂ films grown on c-Al₂O₃ (18 nm),¹¹ β-Ga₂O₃ (19 nm),¹¹ and CuCrO₂ buffer grown on c-Al₂O₃ (33 nm),¹⁰ while we admit that there is room to improve as compared to the bulk value.²⁴ By taking the derivative of the resistivity curve as a function of the temperature and applying a Gaussian fit to the peak, we find the transition temperature T_N ≈ 35 K, which corresponds to the antiferromagnetic transition in PdCrO₂.^25,26 Figure 4(c) shows a HAADF-STEM image of a PdCrO₂ thin film grown on an ultrathin CuAlO₂ template. Multiple AlO_x layers (denoted by the green and yellow arrow) were observed below the CrO₂ sublayer, consistent with a pristine CuAlO₂ template film grown on c-Al₂O₃ [see Fig. 2(e)]. In HAADF-STEM, the scattering intensity is approximately proportional to the square of the atomic number. Therefore, Cu layers should be observed in HAADF-STEM images at the template layers, but there is no bright layer between the AlO_x and CrO₂ sublayers. In other words, there is no clear sign of Cu atomic columns at the template layers. We noted that the absence of Cu in the templated layer was also observed in previous results when using CuCrO₂ as the buffer layer.¹⁰ This phenomenon is not yet fully understood, but we think that the Cu atoms most likely diffused into the film’s interior. Nevertheless, these findings suggest that the templated CuAlO₂ plays a significant role in alleviating the lattice mismatch and structural dissimilarity, which may enable high-quality delafossite film growth.

We have used the solid-phase epitaxy approach to form single-phase CuAlO₂ as a template layer for delafossite growth on a sapphire substrate. Since large-scale delafossite substrates are not available, the demonstrated formation of the CuAlO₂ delafossite structure is of considerable topical interest for epitaxial growth of other metallic or magnetic delafossites. We have shown the optimal reaction temperature and dwell time to obtain single-phase CuAlO₂ from our structural characterizations. Furthermore, we have successfully grown a 70 nm thick PdCrO₂ thin film on this templated substrate, establishing a viable example for successfully synthesizing high-quality PdCrO₂ films, which are otherwise difficult to grow epitaxially. We hope that this process will be a stepping-stone to drive further research in delafossite thin films and heterostructures.

See the supplementary material for additional XRD and AFM data of the samples studied.

This work was supported by the U.S. Department of Energy, Basic Energy Sciences, Materials Sciences and Engineering Division. J.M.O. acknowledges support in part for the data analysis from BrainLink program funded by the Ministry of Science and ICT through the National Research Foundation of Korea (Grant No. 2022H1D3A3A01077468) and National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (Grant No. 2022R1F1A1074425).