Growth of high-quality ruthenium films on sapphire Open Access

The growth of crystalline films of elemental metals on insulating oxide substrates is challenging due to low film–substrate binding energies.^1,2 This difficulty unfortunately also applies to the epitaxial growth of Ru films on sapphire. Such Ru films are of keen interest as ultrathin conductors^3–6 and as templates for the epitaxial growth of other metals^7–11 and graphene.^3,12,13 Sapphire is a preferred substrate for many applications because it is available in high quality and large sizes at comparably low cost.¹⁴ 

Indeed, the c-plane (0001) surface of sapphire also offers favorable properties for the epitaxial growth of Ru films. Both Ru and the oxygen sublattice of sapphire have hexagonal close-packed (hcp) crystal lattices, which lattice match by rotational alignment with a mismatch of only ≈1.3% [a(Ru) = 2.706 Å, d_O–O = 2.747 Å].¹⁵ Moreover, the chemical inertness of sapphire is a favorable property, as it avoids chemical reactions between Ru and the substrate during film growth or annealing steps.

Several efforts have been undertaken to grow Ru films on sapphire. Owing to the low vapor pressure of Ru, sputtering has been the preferred growth method. In these studies, Ru films of varying crystalline quality were obtained, including pronounced mosaic structures, growth islands, and significant intermixing with the substrate.^10,16,17

Given the capability to prepare sapphire surfaces of improved quality^18,19 and to access higher substrate temperatures T_s by means of substrate heating with CO₂ laser irradiation,²⁰ we systematically explored the growth of Ru films in this extended parameter range by depositing Ru films via thermal laser evaporation. Thermal laser evaporation has the advantages of low-pressure thermal deposition, potentially boosting film purity and growth rates.²⁰ 

We show that high-quality single-crystalline Ru films can indeed be obtained on sapphire substrates, however not by growth at high temperatures, but rather by growth at low T_s followed by high-temperature annealing. The Ru films fabricated in this manner feature superior crystallinity and an atomically sharp interface with the substrate. The surfaces of films as thick as 60 nm follow the step-and-terrace structure determined by the vicinal cut of the sapphire. Microstructural analysis by scanning transmission electron microscopy (STEM) reveals the microscopic origin of this growth behavior. Deposition at low T_s minimizes the formation of pronounced growth islands during the early stages of film growth. For films grown at high T_s, these islands evolve dynamically during further film growth and deteriorate the microstructural quality of the resulting Ru film.

Ru films were grown by thermal laser evaporation using a thermal laser epitaxy (TLE) system equipped with CO₂ laser heating^20,21 and a base pressure of about 6 × 10⁻¹¹ mbar on the (0001) surfaces of 5 × 5 mm², single-side polished Al₂O₃ substrates from CrysTec. The substrates were terminated by annealing in situ for 200 s at 1700 °C in a background pressure <2 × 10⁻¹⁰ mbar to show the Al-rich √31 × √31R ± 9° reconstruction as described in Ref. 19. This low pressure was maintained during deposition and possible subsequent anneals. Substrate temperatures were measured with an optical pyrometer on the backside of the substrate. Owing to its low vapor pressure, the Ru was evaporated by heating it with a continuous-wave beam of a fiber laser.²² The 1.06-μm laser was operated at 170 W output power. It irradiated ≈1 mm² areas of cylindrical Ru rods (4.0 N) of 3-mm diameter, yielding growth rates of ≈0.25 Å/s. Film thicknesses were varied by adjusting deposition times.

The substrates were heated by thermal radiation of the Ru sources while these were hot and additionally, if desired, by CO₂ laser radiation. The thermal radiation of the hot sources alone heated the substrates up to ≈280 °C within ≈7 min after the start of the deposition. Possible postgrowth annealing steps performed at various temperatures lasted at least 180 s. Before and after film growth, the substrate and film surfaces were monitored by in situ reflection high-energy electron diffraction (RHEED) with an electron energy of 15 kV. To avoid potentially modifying the film microstructure by electron irradiation,^23,24 the electron beam of the RHEED was turned off when substrates were hot.

Aiming to grow high-quality Ru films, we first followed the obvious approach and attempted to deposit films epitaxially by growing them at high T_s on the prepared substrate surfaces. To identify the best T_s for epitaxial growth, we fabricated a first series of samples in which we varied T_s from 280 to 1300 °C. This procedure consists of terminating the substrates in situ and growing the film at elevated temperatures. We will refer to this procedure as “two-step process.”

To reveal the influence of T_s on the microstructure of the Ru films, we start by investigating the structural properties of Ru grown at the lowest T_s ≈ 280 °C.

Figure 1(a) shows the RHEED pattern and atomic force microscopy (AFM) images of such a Ru film. The prominent spots of the RHEED image are noticeably blurred and large, indicating a three-dimensional surface structure with epitaxial alignment, which is consistent with the AFM images, which show many small surface features.

Now proceeding to growth at higher temperatures, Fig. 1(b) shows the corresponding images for growth at T_s = 400 °C. The RHEED images now show rather broad lines, indicating a smooth, yet also disordered surface. The corresponding granular crystal structure with grains that are larger than the ones obtained at T_s = 280 °C is well visible in the AFM image.

For growth at 700 °C [Fig. 1(c)], the RHEED image shows for the first time Kikuchi lines in addition to the diffracted spots and lines known from the growth at lower T_s, suggesting an ordered microstructure. However, the surface of this sample is not closed, as shown in the corresponding AFM image. The Ru grains have assembled into a meandering, percolative arrangement. Gaps exist where the grains approach each other.

The surfaces of Ru films grown at 1000 °C [Fig. 1(d)] display a rather different morphology. Also for these films, the RHEED image shows very narrow lines with strongly pronounced Kikuchi lines, again indicating a smooth, slightly disordered surface of an epitaxially grown film. The corresponding AFM image shows many large islands. Whereas different islands have different heights, the surfaces of the islands are atomically flat. The edges of the islands tend to be aligned to the hexagonal crystal lattice of the sapphire. The shape of the islands is reminiscent of a Wulff shape.^25,26 These are the films with the best microstructure we could obtain by growing at elevated temperatures. Nevertheless, we have to note that, although the films are epitaxial, they are not continuous. Figure 2(a) shows an STEM cross-sectional image of such a Ru film. As expected from Fig. 1(d), the cross-sectional view also shows pronounced growth islands. Moreover, it reveals that the grooves between the islands extend all the way down to the substrate. Furthermore, EELS reveals undesired intermixing of cations at the interface between these islands and the underlying sapphire [Fig. 2(b)]. Considering the high solubility of Al in Ru, this intermixing is likely a result of Al-diffusion into the Ru film.²⁷ This free Al must be generated by chemical reactions between Ru and the underlying sapphire substrate, the results of which are directly observed at even higher T_S.

Increasing T_S further does not improve the film quality. Indeed, as highlighted by Fig. 1(e), which shows the data obtained after deposition at 1300 °C, the processes occurring during film growth change. The RHEED patterns of such samples taken after the growth procedure [Fig. 1(e)] represent the √31 × √31R ± 9 reconstruction of the Al₂O₃ (0001) substrate, suggesting that the Ru atoms deposited at this temperature do not remain on the hot substrate surface. The AFM images show numerous triangular holes, suggesting that some of the Ru atoms form a volatile oxide by reducing the Al₂O₃ and, thus, etching the triangular holes into the sapphire. These holes likely decorate dislocation cores.^28,29

The systematic variation of T_s during film deposition, therefore, showed that, by increasing T_s to the highest point at which growth could still be observed (T_s = 1000 °C), the growth islands become larger and develop flat surfaces. However, even for T_s = 1000 °C, at which the film in the islands have the best microstructural quality, grooves remain that reach through the entire film thickness between the islands. These films are, therefore, not continuous.

These observations suggest that the formation of extended islands driven by the increased adatom diffusion, ripening and dewetting at high T_s must be suppressed to obtain continuous films. As these effects are weaker at lower T_s, and because good film crystallinity requires high-temperature diffusion, we explored a three-step process for the growth of Ru films that is reminiscent of solid-phase epitaxy: (solid-phase epitaxy) the films are deposited on the prepared substrates at low T_s and are subsequently annealed at high temperatures. In the process we applied, the deposited Ru films are 3D epitaxially aligned prior to the annealing step. This is subtly different from solid-state epitaxy, which is defined such that amorphous films are deposited and subsequently crystallized.³⁰ 

We deposited films at T_s = 280 °C and annealed them for at least 180 s at 1000 °C. This is the temperature at which the two-step process described above yielded the best film quality.

Figure 3 shows the RHEED diffraction pattern and the AFM surface image of a 14-nm-thick Ru film layer grown by this three-step process. The surface morphology of this film shown in the AFM image differs from those of films fabricated with the conventional two-step process described above. First, the surface reveals a continuous film. Second, the surface shows a complex surface structure on a length scale <50 nm and a height <1 nm, which resembles small flames or the sun’s surface imaged in the ultraviolet range.³¹ The variations of the sample height at these structures are significantly smaller than the lattice parameter. Thus, it is unlikely that they result from dislocations that are close to the surface. However, it seems possible that this surface structure is shaped by strain fields of dislocations located well below the film surface, for example, at the substrate–film interface, as suggested by the protruding lines of constant width, which consist of straight or arced segments of constant curvature. Third, a series of lines crosses the surface from the lower left to the upper right. These reflect the terrace steps of the sapphire substrate, indicating a uniform thickness of the Ru layer. The remarkably low RMS surface roughness of this film is 0.18 nm. As for films deposited at 1000 °C in the two-step process [Fig. 1(d)], the RHEED image exhibits a two-dimensional surface structure and a highly ordered microstructure. The elongation of the spots indicates a variance in diffraction conditions, which can be attributed to the complex surface structure observed in the AFM image and/or to atomic-level surface disorder.

The cross section of this sample is shown in the STEM image, see Figs. 4(a) and 4(b). The Ru film is continuous and features a sharp, apparently defect-free interface to the substrate. Film–substrate interdiffusion is very limited, as shown by EELS investigations [Fig. 4(c)]. According to the STEM images, the films have a highly ordered microstructure. Taking a closer look, we see that the STEM cross section [Fig. 4(b)] also reveals a subtle, wavelike shift of the atomic layers with a period of ≈2 nm. These shifts may contribute to the surface morphology as shown in Fig. 3.

Further insight into the effects of the postgrowth anneal on the microstructure of the Ru films is provided by x-ray diffraction studies. Figure 5 shows θ/2θ diffraction measurements of two characteristic samples. These are a first sample fabricated with the two-step process by deposition at 1000 °C without being subjected to a subsequent anneal, and a second sample fabricated with the three-step process by deposition at 280 °C with a subsequent anneal at 1000 °C for at least 180 s. The two scans show exclusively the sapphire 0006 and 00012 and the Ru 0002 and 0004 peaks. Compared with the diffraction pattern of the two-step sample, that of the three-step sample shows a stronger and narrower Ru 0002 peak as well as more pronounced Laue oscillations [Fig. 5(b)], indicating a higher crystallinity of the film obtained by the three-step process and an excellent film–substrate interface, consistent with the STEM cross-sectional results. The high degree of crystallinity of films produced by the three-step process is evidenced directly by the rocking curves of the Ru 0002 peaks [Fig. 5(c)]. The full-width at half-maximum (FWHM) of the curve for the sample shown is only 0.063°, which is about one order of magnitude smaller than the best values published in the literature.^6,15

As the film–substrate interdiffusion is greater for samples grown with the two-step than three-step process, we speculate that the interdiffusion in the two-step samples is fostered by stress fields associated with the pronounced Ru growth islands in these samples. During growth at high temperatures, the substrate surface is covered only under the growth island by Ru. At the islands’ edges, stress field gradients may enhance Al–Ru interdiffusion such that the Ru islands may appear to be slightly sunk into the sapphire [Fig. 2(a)]; a process referred to as endotaxy.³² During growth at low temperatures, the island density is much higher, disallowing relatively long-range stress fields to develop. Obviously, interdiffusion is also suppressed. Heating a surface fully covered by Ru yields no lateral variation in diffusion rates, and, thus, substrate–film interdiffusion is suppressed. Another explanation for the lack of interdiffusion is that the closed Ru layer kinetically limits the formation of the volatile ruthenium oxide.

As a final point, we note that, to yield high-quality Ru films, the optimal three-step growth process requires the deposited Ru film to be thicker than ≈10 nm. Thinner films do not appear to be sufficiently continuous to stabilize the Ru layer during the postdeposition anneal or dewet as a result of thermodynamic instability.³³ The anneal then yields Ru layers that again are marred by islands and trenches. This dewetting threshold temperature of 1000 °C for ≈10-nm-thick epitaxial Ru films on Al₂O₃ (0001) is much higher than reported for polycrystalline Ru on SiO₂.³⁴ 

In summary, we have provided a systematic study of the growth behavior of Ru films on c-plane sapphire substrates, a case study for the epitaxial growth of films of elemental metals on oxide insulators. The films were deposited by laser evaporation on thermally prepared (0001) sapphire surfaces with an Al-termination. Without further processing, films deposited at any substrate temperature were 3D, although epitaxial. Growth at 1000 °C yielded films of the highest crystallinity, yet these films were still discontinuous. A three-step process was found to ameliorate the underlying detrimental effects of island formation on film quality. Following in situ substrate termination, Ru films were deposited at low temperatures and then annealed in a high-temperature step. With this process, epitaxial Ru films of unprecedented crystalline quality have been obtained on sapphire (0001).

The authors acknowledge Ute Salzberger for assistance with TEM sample preparation.

The authors have no conflicts to disclose.

Lena N. Majer: Conceptualization (equal); Data curation (lead); Formal analysis (equal); Investigation (lead); Methodology (equal); Validation (equal); Visualization (lead); Writing – original draft (equal). Sander Smink: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Wolfgang Braun: Conceptualization (equal); Data curation (supporting); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Hangguang Wang: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Peter A. van Aken: Supervision (equal); Validation (equal); Writing – review & editing (equal). Jochen Mannhart: Conceptualization (equal); Investigation (equal); Supervision (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal). Felix V. E. Hensling: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Validation (equal); Visualization (supporting); Writing – original draft (equal); Writing – review & editing (lead).

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