In situ x-ray studies of growth of complex oxides on graphene by molecular beam epitaxy

Remote epitaxy is a novel synthesis technique that allows for the fabrication of thin, freestanding single crystals and nanomembranes.^1–5 It relies on a sacrificial graphene layer between a thin film and a single crystalline substrate: during film deposition, the electronic interactions across the graphene are strong enough to enable epitaxial growth but weak enough to allow mechanical release of the film. Others have demonstrated methods for the fabrication of freestanding structures,^6–8 but the procedures are often materials-specific in terms of the interlayer permitting epitaxial growth. Use of a more universal interlayer can facilitate the development of a field we term “stacktronics,” where a variety of single crystalline materials can be laid atop each other at low to moderate temperatures, leading to the synthesis of complex van der Waals heterostructures.⁹ 

The science underlying remote epitaxy continues to be investigated. Based on density functional theory calculations for GaAs, Kim et al.² suggested that remote epitaxy can take place when the substrate–epilayer gap is within 9 Å. This was followed by experimental verification, where homoepitaxial growth of GaAs, InP, and GaP was demonstrated through a monolayer of graphene. Kong et al.³ found that while remote epitaxy is active for compounds with a polar lattice arrangement, such as GaAs and GaN, it is not active for the growth of non-polar materials, such as Si and Ge. The long-range interactions stemming from the ionicity of the former materials appear to be critical for promoting epitaxy across the graphene interlayer.

Although the complex oxides exhibit ionic bonds, the growth of these materials requires deposition under oxidizing conditions that would normally etch carbon-based compounds. In addition, the behaviors of many complex oxides are highly sensitive to both the oxygen stoichiometry and film thickness,^10,11 and strict control over defects, such as oxygen vacancies, can be paramount to the development of properties, such as superconductivity and the 2D electron gas.^12–17 Kum et al.¹ ameliorated the problem by first growing a buffer layer, i.e., a thin oxide (5–10 nm) in low oxygen pressure prior to continuing growth in a more oxidizing environment such that oxygen ions could subsequently diffuse into the reduced buffer layer. Furthermore, a slightly thicker interlayer (e.g., bilayer graphene) can be employed to mitigate etching in the growth environment and high temperature reaction with the oxide substrate.¹⁶ 

Advances in stacktronics require improved understanding of remote epitaxy and insight into the impact of interlayer thickness and deposition conditions on the nucleation and growth of thin films. This necessitates in situ studies sensitive to the atomic-level structure conducted in the growth environment. In situ measurements are particularly important for the synthesis of complex oxides, where small changes to the degree of oxygen incorporation can impact the properties of the film/interface and degrade the graphene interlayer.¹⁸ Here, we conduct in situ studies of thin film synthesis on polycrystalline graphene, using synchrotron x-ray scattering to investigate the deposition of SrTiO₃ (STO) and LaNiO₃ (LNO) by molecular beam epitaxy (MBE) onto graphene-coated STO (001) substrates (Fig. 1). We find that while high-quality STO can be grown on graphene in a small oxygen background, the thicknesses of the graphene interlayer and film are important factors with regard to the final structure. We also observe the etching of graphene in an ozone environment even with the presence of an STO buffer layer.

The STO (001) substrates were prepared by selective etching in buffered hydrofluoric acid and an anneal in flowing O₂.¹⁹ This results in surfaces that exhibit a step-terrace morphology and maintain the TiO₂ double layer, as reported previously.²⁰ Graphene layers grown by chemical vapor deposition (CVD) were then transferred onto the STO surfaces (“Methods”). After the transfer of one-, two-, or four-monolayer-thick graphene (1ML, 2ML, and 4ML G), the presence and thickness of graphene were measured by Raman microscopy and spectroscopy (Fig. S1). The Raman images confirm that the transferred graphene is uniform over a 100 × 100 μm² region. The graphene-coated substrates (G/STO) were then placed in the oxide MBE system described in Ref. 21. The chamber is mounted on a six-circle diffractometer at the Advanced Photon Source, allowing in situ x-ray and reflection high-energy electron diffraction (RHEED) studies during and after MBE growth.

Features in reciprocal space originating from the Fourier transform of a planar surface, i.e., crystal truncation rods (CTRs) as illustrated in Fig. 1, are extremely sensitive to the surface structure, and studies performed during MBE provide snapshots of the growth process with sub-Ångstrom resolution.^22,23 Figures 2(a), 2(c), and 2(e) show scans run along the specular rod (00L) at different temperatures for 1, 2, and 4ML G/STO, respectively. In general, a peak appears at L ∼ 1.2 rlu,²⁴ intensifying as the temperature reaches 550 °C, with another weak peak appearing at L = 2.4 rlu at 760 °C. Coherent Bragg rod analysis (COBRA) was used to analyze the 00L spectra (“Methods”). The resulting electron density profiles are shown in Figs. 2(b), 2(d), and 2(f). The position of the TiO₂-terminated STO substrate surface is placed at z = 0. Above this lies the TiO_2−x adlayer and the atomic planes due to graphene (G1, G2, etc.); the adlayer does not change position with the addition of graphene. The electron densities at G1, G2, etc., increase with heating, and additional peaks are observed to develop at the higher temperatures. For example, a small G2 peak appears for the 1ML G/STO sample at 390 °C and above [Fig. 2(b)]. This may be due to a variety of reasons, such as the removal of adsorbed water from the transfer process, but it is important to note that the x-ray results reflect an average over a relatively large lateral area (several mm²) and that while the interplanar spacings of the peaks (i.e., the Z-positions) are very accurate and are here consistent with multiple layers of graphene, there are relatively large error bars associated with the quantitative size of the peaks (i.e., the electron density).

The process of remote epitaxy is highly sensitive to the distance between the top of the substrate (i.e., the crystalline template) and the initial monolayer of the film. From the electron density profiles shown in Fig. 2, the interplanar spacings were determined for all temperatures studied during heating (Fig. S2). The layer spacings for the topmost planes of 1ML, 2ML, and 4ML G/STO are shown in Fig. 3(a) as a function of temperature. For 1ML G/STO [bottom of Fig. 3(a)], the distance between the topmost TiO_2−x plane and G1 (T-G1) remains ∼3.7 Å for all temperatures. This value is also observed for the 2ML and 4ML G samples, especially after heating to 760 °C; at lower temperatures, residual liquid from the transfer process (which could necessitate multiple graphene transfers) may remain between the layers. From these results, it appears that the average distance between the top of the STO and the top of the graphene ranges from 3.7 Å for single layer graphene to 7.5 Å for double layer graphene and to 14.0 Å for four layers of graphene at elevated temperatures [Fig. 3(b)].

A comparison of the CTRs at 760 °C in UHV is presented in Fig. 4(a). The specular rod for bare STO is shown at the bottom. A minimum of intensity appears near L = 1.05 rlu, which is a characteristic of the TiO₂ double layer.²⁰ After the addition of a single graphene layer (pink profile), a peak appears at L ∼ 1.2 rlu, and oscillations appear along the 00L below the 001 Bragg reflection. With additional graphene layers, the peak sharpens, its position characteristic of the lattice spacing of multilayer graphene as was shown in Fig. 3(a). Although oxygen can diffuse from the substrate into an overlayer,¹⁶ it appears that graphene survives at high temperatures on the surface of STO. However, the growth of stoichiometric oxides by MBE requires an oxygen partial pressure, pO₂. We measured the intensity of the graphene peak at L = 1.2 rlu at 760 °C for 2ML G/STO while increasing the pO₂. As depicted in Fig. 4(b), the intensity drops with increasing pO₂, but the peak does not substantially decrease until reaching ∼6 × 10⁻⁵ Torr. Once the pO₂ reaches 2 × 10⁻⁴ Torr, the intensity minimum at L = 1.05 rlu reappears, indicating the removal of most of the graphene by high temperature oxidation. Figure S3(b) and Table S1 show the results of calculated spectra for samples with different areal coverages of multilayer graphene, simulating the effect of oxygen partial pressure.

Interestingly, the Debye–Scherrer ring from the polycrystalline graphene remains detectable throughout the entire pressure range, indicating that not all of the graphene is removed across the footprint of the x-ray beam.²⁵ Figure 4(c) shows a ring from the 100 reflection. By collecting similar images as a function of pO₂ and applying the modified Scherrer equation,²⁶ we find that the average in-plane graphene domain size stays in the range of 350–400 nm, as shown in Fig. 4(d). Scans along the ring (ϕ) (Fig. S4) show that although the graphene is mostly polycrystalline, the in-plane orientation is not random when grown by CVD on copper foil but exhibits a modest texture with the hexagonal symmetry.^27,28

Twelve-unit-cell-thick (12UC) films of reduced STO were subsequently grown at 760 °C on 1ML G/STO; to minimize oxidation of the graphene, growth was conducted at a chamber pressure of 3 × 10⁻⁸ Torr. This provides a sufficient amount of oxygen to grow thin STO without introducing oxygen from a separate source (“Methods”).^13,21 The RHEED patterns shown in Fig. 5(a) demonstrate that the surfaces prior to and after growth on graphene remain smooth. Furthermore, according to the in-plane grazing incidence diffraction (GID) data in Fig. S5(a), the STO film is epitaxial with the substrate. We also examined 12-unit-cell-thick STO films deposited on 2ML G/STO and 4ML G/STO. The intensity measured at 001/2 rlu during deposition shows oscillations during growth of the first three unit cells on 4ML G, demonstrating the epitaxial layer-by-layer growth mode [Fig. S5(b)] despite the 14 Å gap [Fig. 3(b)].²⁹ Ultimately, however, the 12UC STO film exhibits multiple diffraction peaks, indicating that the film becomes polycrystalline [Fig. S5(a)]. We then find that remote epitaxy depends on the thickness of film and the graphene interlayer: the quality of thicker oxide films grown by MBE benefit from a single layer of graphene. Although epitaxy can occur through two layers of graphene,¹ the observation of additional electron density at G3, which lies ∼11 Å from the STO surface [Fig. 2(d)], suggests that the epitaxial growth process may not take place uniformly over the entire surface. This distance is in rough agreement with 9 Å limit described by Kim et al. for GaAs.² Spatial heterogeneity in the graphene thickness can, therefore, lead to defects in the epitaxial structure and eventual grain boundaries.

In Fig. 5(b), we show the Raman spectra for the 1ML G/STO and the 12UC STO film on 1ML G/STO, demonstrating that a smooth STO thin film can be epitaxially grown at 760 °C without the removal of the graphene. Note that the G band from graphene at ∼1594 cm⁻¹ merges with the high-frequency scattering feature from the 2LO₃ (longitudinal optical branch) overtone of the STO substrate at ∼1600 cm⁻¹.³⁰ After subtracting the STO signal, the Raman spectrum for the G/STO shows the characteristic G and 2D (∼2682 cm⁻¹) bands with I_2D/I_G ∼ 2. This, combined with the absence of the D band, confirms the high quality of the monolayer graphene. After the deposition of 12UC STO film, we observed a broader G band. This is likely due to overlap with the D′ band and a D band appearing at ∼1350 cm⁻¹. The D band comes from inter-valley phonon and defect scattering, the D′ stems from an intra-valley phonon process, and both appear in disordered graphene. In addition, the 2D intensity weakens and a weak D + G band emerges, both of which can be attributed to breakdown of the hexagonal crystal lattice.³¹ Compared to pulsed laser deposition (PLD), growth by MBE exhibits lower kinetic energies and slower growth rates. Even so, the deposition of a 12 unit-cell-thick STO film appears to produce defects in the transferred graphene. We note that the 2D band from the STO film is broader as compared to that for 1ML G/STO and is blue-shifted by ∼10 cm⁻¹. This arises from the symmetry lowering that takes place when increasing the number of layers of graphene in the sample,³² consistent with the COBRA results.

As mentioned above, the homoepitaxial STO film can serve as a buffer layer for the deposition of oxides that require higher pO₂ such as the nickelates. We attempted the heteroepitaxial growth of LaNiO₃ on the 12UC STO film on 1ML G. After heating the sample to 590 °C in an ozone background of 2.8 × 10⁻⁶ Torr, we grew 12 unit cells of LNO. Measurements were conducted along the specular rod immediately after the growth of each unit cell, as shown in Fig. 6(a). The Laue fringes between the substrate Bragg peaks can be easily discerned, providing evidence of smooth interfaces. The same thickness LNO was grown directly on STO for comparison, as shown in Fig. 6(b). Upon close examination of the rods, we find that although the thickness fringes from the film grown on 12UC STO/1ML G are slightly less distinct than those for LNO grown directly on the substrate, the rods are very similar, especially for films thicker than four unit cells. This is presumably due to full oxidation of the reduced STO film in the ozone background. The average out-of-plane lattice parameters 12UC LNO, as determined from the positions of the 003 peaks, are identical at ∼3.80 Å, which also agrees with a previous report on ultrathin LNO/STO.³³ However, we find no evidence of graphene post deposition. The Raman spectrum in Fig. S6 shows no peaks from graphene over a 100 × 100 μm² area. Although a monolayer of graphene can be easily etched by direct exposure to ozone,^34–36 this also appears to occur through an intermediate, a 12-unit-cell-thick STO buffer layer.

A series of in situ synchrotron x-ray experiments were conducted on graphene-coated STO under various film synthesis conditions. At the high temperatures necessary for epitaxial growth, the spacing between the STO surface and the adjacent layer of graphene is ∼3.7 Å – similar to the interplanar spacing between multiple layers of graphene. We also find that upon heating additional electron density appears ∼3.7 Å above the topmost layer of graphene transferred at room temperature.

With the intensity of the peak at L = 1.2 rlu serving as a metric for the amount of graphene, we observed the effect of oxygen partial pressure on graphene-coated STO, finding that most of the graphene is removed at pO₂ = 2 × 10⁻⁴ Torr at 760 °C. Lower partial pressures can be used for the growth of SrTiO₃ on graphene, which was confirmed by the deposition of 12UC STO on 1ML G. The crystallinity of the film, however, depends on both its thickness and the thickness of the interlayer: for example, 12UC STO is epitaxial when grown on 1ML G but polycrystalline when grown on 2ML G; 3UC STO is epitaxial on 4ML G but becomes more polycrystalline with increasing film thickness. The crystal quality appears to depend on the additional electron density lying above the topmost graphene layer (>9 Å), as nuclei formed on these regions may have poor interaction with the underlying crystal substrate. We also find that 1ML graphene does not survive the growth of LNO in ozone despite the presence of a 12UC STO buffer layer. Thicker oxide buffer layers and the use of single crystalline graphene may be necessary for the synthesis of nickelate heterostructures on graphene.

Commercial graphene flakes grown by chemical vapor deposition³⁷ were immersed in deionized water for a minimum of 2 h. The floating graphene layers were then transferred to etched and annealed SrTiO₃ substrates, dried at room temperature for 30 min, and dried at 100 °C for an additional 20 min. To remove the 500-nm-thick PMMA coating from the graphene, the samples were immersed in acetone (with the acetone preheated to 50 °C) for 30 min. After rinsing, the samples were dried at 200 °C under flowing argon for 10 min.

After background subtraction and the application of geometrical and polarization correction factors to the measured intensities from the 00L CTR, we used the COBRA phase retrieval method to convert the intensities into 1D electron density profiles. Details on the COBRA method are provided in Ref. 38. The x-ray energy was set to 15.5 keV and the data collected with a Pilatus 100 K area detector.

The films were grown in an oxide MBE chamber mounted on a six-circle diffractometer located at Sector 33 of the Advanced Photon Source.²¹ The deposition rates for each source were calibrated with a quartz crystal microbalance (QCM), and the substrates temperatures were determined with an infrared optical pyrometer. Either oxygen or ozone can be used as the oxidant.³⁹ The RHEED was performed at 10 keV.

See the supplementary material for additional results from Raman microscopy and x-ray scattering.

This research used resources of the Advanced Photon Source and Center for Nanoscale Materials, both U.S. Department of Energy (DOE) Office of Science User Facilities, and is based on work supported by Laboratory Directed Research and Development (LDRD) funding from Argonne National Laboratory, provided by the Director, Office of Science, of the U.S. DOE under Contract No. DE-AC02-06CH11357.

The authors have no conflicts to disclose.

Xi Yan: Data curation (lead); Formal analysis (equal); Writing – original draft (equal). Hui Cao: Data curation (equal); Formal analysis (equal). Yan Li: Data curation (equal); Formal analysis (equal). Hawoong Hong: Data curation (equal). David J. Gosztola: Data curation (equal). Nathan P. Guisinger: Data curation (equal). Hua Zhou: Data curation (equal); Formal analysis (equal); Writing – review & editing (equal). Dillon D. Fong: Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.