Introduction

The holy grail of semiconductor heterostructures is the ability to produce high-quality semiconductor films with tunable bandgaps that are also lattice-matched to commercially available substrates. One common strategy to achieve this is to alloy two or more semiconductors and judiciously choose the alloy composition for the desired bandgap and lattice parameter. In principle, this approach can be applied to the alkaline-earth stannates, a system of three materials (namely CaSnO3, SrSnO3, and BaSnO3) that has recently fascinated researchers for potential applications in next-generation power electronics and ultraviolet optoelectronics. The alloys of these three materials, whose crystal structures are summarized in Fig. 1b, span the purple shaded region of bandgaps (Eg) and pseudocubic lattice parameters (a) in Fig. 1a. However, the scarcity of commercially available substrates makes the range of accessible bandgaps in lattice-matched systems rather limited. The use of GdScO3 as the substrate facilitates the largest bandgap range for lattice-matched alloys. However, the span is only 0.1 eV, from Eg = 4.2 eV (Ba0.13Ca0.87SnO3) to Eg = 4.3 eV (Sr0.26Ca0.74SnO3). Such a limited bandgap range provides few opportunities for modulation doping using lattice-matched oxide heterostructures.

Fig. 1: Summary of alkaline-earth stannates and alkaline-earth germanates.
figure 1

a Bandgap (Eg) vs pseudocubic lattice parameter (a) of alkaline-earth stannates and alkaline-earth germanates, SrGeO3. The purple shaded region represents values of Eg and a that are available to alloys of the three materials. The yellow region represents those additional values that are available when including perovskite SrGeO3 in the alloy system. The lattice parameters of commercially available substrates are shown as vertical lines. b Crystal structures of alkaline-earth stannates. c Crystal structures of alkaline-earth germanates. All stannate perovskite structures are from ref. 37. The CaGeO3 structure is from ref. 10. The SrGeO3 structure is from ref. 26. †Denotes metastable crystal structures that were quenched to ambient conditions. *Denotes a rhobohedral distortion at room temperature.

Full size image

Adding a second tuning parameter to this material system would expand the gamut of accessible properties. Replacement of Sn with Ge offers one such tuning parameter. Oxides containing Ge4+ typically have conduction bands derived from Ge 4s orbitals, which, analogous to Sn 5s orbitals, produce dispersive conduction bands with low electron-effective masses. However, the different sizes and energies of the atomic orbitals lead to nuanced differences in the physical and electronic structures. This variation has been exploited in rutile oxides, where the substitution of Sn with Ge yields semiconductors with bandgaps ranging from 3.6 to 4.7 eV1 with predicted ambipolar doping, offering encouraging prospects as ultrawide bandgap (UWBG) semiconductors for power electronics2,3,4,5,6,7.

The crystal structures of alkaline-earth germanates are summarized in Fig. 1c. While perovskite BaGeO3—a chemical analog to BaSnO3—has not yet been synthesized, it is predicted to be metastable in the cubic structure8. The cubic perovskite SrGeO3 has been successfully synthesized by a high-pressure, high-temperature method and quenched to ambient conditions; it displays an indirect bandgap Eg = 2.7 eV, but its wider direct bandgap (3.5 eV) makes it transparent to visible light9. SrGeO3 has also been doped with La, yielding room-temperature mobility of 12 cm2 V−1 s−1 (see ref. 9). However, DFT calculations suggest that both SrGeO3 and BaGeO3 have the potential to achieve phonon-limited mobilities superior to those of BaSnO38. Finally, CaGeO3 has a metastable orthorhombic perovskite structure quenchable to ambient conditions. Although this material has not been optically or electrically characterized, it is almost certainly a transparent semiconductor with Eg > 2.7 eV10,11,12,13.

The yellow region in Fig. 1a shows the additional range of Eg and a made available by adding SrGeO3 to the stannate alloy system. We note that the region encompasses nearly all commercially available perovskite oxide substrates, and alloys lattice-matched to DyScO3 substrates have bandgaps that span 1.5 eV, from Eg = 2.9 eV (Ba0.49Sr0.51Sn0.49Ge0.51O3) to Eg = 4.4 eV (Sr0.07Ca0.93SnO3), providing ample opportunity for lattice-matched heterostructures in modulation-doping field-effect transistors (MODFET) and, potentially, even optoelectronic devices.

Molecular beam epitaxy (MBE) is considered a gold-standard technique to grow high-quality heterostructures. However, since MBE relies on the codeposition or shuttered growth of individual elements, achieving a composition with a target Eg and a while simultaneously maintaining the A:B-site cation stoichiometry presents a formidable challenge in flux calibration. Therefore, it is desirable to use adsorption-controlled growth, which exploits thermodynamics to automatically regulate the A:B-site cation stoichiometry. Hybrid MBE is a technique that draws on the high vapor pressure of metal–organic precursors to achieve adsorption-controlled growth. It has been successfully applied to the adsorption-controlled growth of titanates14, vanadates15, and stannates16. However, it has not yet been applied to the growth of Ge-containing oxides. Although ozone-assisted MBE growth using germanium-suboxide as a source of germanium has recently been demonstrated5, hybrid MBE may offer additional advantages: (1) there is no need to break growth chamber vacuum to replace materials; (2) oxygen-containing precursor provides an additional source of oxygen, and therefore, eliminates the need of ozone; (3) since the chemical precursor is not directly kept inside the UHV chamber, it eliminates the possibility of source oxidation (and hence the flux instability).

Here, we report on a hybrid MBE approach for the growth of Ge-based oxides using germanium tetraisopropoxide (GTIP) as a metal–organic precursor for Ge. Figure 2 shows that GTIP has a much higher vapor pressure than elemental Ge. GTIP only requires temperatures as low as 50–100 °C whereas elemental Ge requires significantly higher temperatures (>1000 °C) to achieve sufficient flux. Furthermore, GTIP vapor pressure is also comparable to established hybrid MBE metal–organic precursors. We chose Sn1-xGexO2 and SrSn1-xGexO3 as model systems to demonstrate the use of GTIP in the growth of binary and ternary oxides, successfully synthesizing epitaxial rutile Sn1-xGexO2 and coherent perovskite SrSn1-xGexO3 films. Phase-pure rutile GeO2 and perovskite SrGeO3 were not chosen for this study as they may require epitaxial stabilization and therefore, will be a subject of future study.

Fig. 2: Vapor pressures and structures of metal–organic precursors.
figure 2

a Vapor pressures of common elements in Ge-based oxides compared to precursors for hybrid MBE. b–g The chemical structures of the precursors including hexamethylditin (HMDT, b)16,38, hexamethyldigermanium (HMDG, c), titanium tetraisopropoxide (TTIP, d)39,40, germanium tetraisopropoxide (GTIP, e), vanadium oxytriisopropoxide (VTIP, f)41, and zirconium tert-butoxide (ZTB, g)42,43. All metal vapor pressures use the equations from Alcock and coworkers44. Ge uses a fit to data from ref. 45. The vapor pressures for GTIP, TTIP, and VTIP use Antoine parameters from ref. 46 who themselves took this data from ref. 47. HMDT data use the enthalpy of vaporization from ref. 48 and the standard entropy of vaporization fit to boiling temperatures from chemical suppliers. HMDG data use Trouton’s rule along with a boiling temperature provided from Sigma-Aldrich.

Full size image

Results and discussion

AFM and XRD results of Ge-based oxides

Figure 3a shows AFM of rutile Sn1-xGexO2/TiO2(001) with different Ge concentrations. Increasing the germanium fraction x from 0 to 0.54 decreased the root mean square (RMS) roughness from 1373 to 461 pm. Supplementary Note 1 discusses the delicate interplay among film composition, film thickness, and surface roughness.

Fig. 3: Surface morphology and structural quality of Ge-based oxides.
figure 3

a Atomic force microscopy (AFM) of Sn1-xGexO2/TiO2 (001) films showing smooth film surfaces. b Room-temperature high-resolution X-ray diffraction (HRXRD) 2θ-ω coupled scans and rocking curves around the (002) film peak of GexSn1-xO2/TiO2(001) films. c AFM images of SrGexSn1-xO3/GdScO3(110) showing film surfaces that get smoother with Ge incorporation. d Room-temperature HRXRD 2θ-ω coupled scans and rocking curves around the (002)pc film peak of SrGexSn1-xO3/GdScO3(110). The insets of (b, d) show the film structures.

Full size image

Figure 3b shows the rutile HRXRD 2θ-ω coupled scans and corresponding rocking curves around the (002) film peaks. The 2θ-ω coupled scans show that the film lattice parameters decrease with Ge incorporation, consistent with Ge’s smaller ionic radius. The full width at half maximum (FWHM) of each film decreases from 0.93° to 0.086° as Ge incorporation increases from 0 to 0.54. This goes against the intuitive expectation that adding Ge would increase the structural disorder by disrupting translational symmetry through random substitution. It is also noteworthy that these rocking curves appear to be a linear combination of two Gaussians (a narrow and a broad component). In the discussion of the RSMs below, we explain both observations in terms of strain relaxation.

Figure 3c shows AFM images of SrSn1-xGexO3/GdScO3(110) with different Ge concentrations. These micrographs show surface roughnesses that decrease with Ge incorporation from 503 to 171 pm. Unlike the rutile samples, however, this trend cannot be explained by film thickness. Instead, the perovskite film thickness is not affected by Ge incorporation, suggesting the Sn-species desorption is not affected by Ge incorporation in this material system. This is likely a result of the higher thermodynamic stability of the perovskite system compared to the rutile system. The decreased surface roughness with Ge incorporation might instead be explained by other factors, such as modified adatom mobility or a decreased driving force for adatom agglomeration.

Figure 3d shows the HRXRD 2θ-ω coupled scans and corresponding rocking curves around the perovskite (002)pc film peak. The 2θ-ω coupled scans demonstrate that the replacement of Sn with Ge decreases the film lattice parameters, consistent with smaller Ge4+ at the Sn4+ site. The Kiessig fringes and rocking curve FWHM of 0.07–0.08° demonstrate uniform films with high structural quality.

To investigate how Ge incorporation influences the strain relaxation of these films, we measured reciprocal space maps of both the rutile and perovskite samples. Figure 4a–c shows the RSMs around the (202) reflection of the rutile films. Contours and guidelines have been added to show the expected peak position depending on composition and strain, following a similar procedure used for (Al,Ga)N heterostructures by Enslin and coworkers17. Each contour represents all possible strains at a single composition, and the two guidelines represent all possible compositions for fully coherent and fully relaxed films. For the SnO2 film in Fig. 4a, the film peak is centered over the x = 0 contour, close to where it intersects the relaxed guideline suggesting a nearly complete film relaxation.

Fig. 4: Reciprocal space maps of Ge-based oxides.
figure 4

a–c Reciprocal space maps (RSMs) of rutile GexSn1-xO2/TiO2 (001) films in the (202) region. d–f RSMs of perovskite SrGexSn1-xO3/GdScO3 (110) films in the (103)pc region. All RSMs include composition contours and relaxed/coherent guidelines to show the expected peak positions based on composition and strain.

Full size image

For the film in Fig. 4b, the film peak resides slightly north of the x = 0.25 contour, consistent with the x = 0.28 Ge fraction determined from XPS. Furthermore, the peak lies between the relaxed and coherent guidelines, indicating the film is compressively strained and has undergone a small degree of relaxation toward its bulk lattice parameter. For the x = 0.54 film in Fig. 4c, the film peak resides directly over the coherent guideline, indicating a film that is completely tensile strained to the substrate. The reader may notice that the film is expected to lie north of the x = 0.50 contour but is in fact south of it. This small discrepancy is mostly likely due to small deviations from Vegard’s law, which was used to calculate the positions of these contours. This discrepancy may also be caused by an error in the composition determined from XPS.

The progression from a nearly fully relaxed film at x = 0 to a fully strained film at x = 0.54 can be explained by considering two facts. First, the growth rates decrease with Ge incorporation, so higher values of x correspond to thinner films that have less built-up elastic strain energy. Second, films with higher Ge fractions have smaller lattice parameters, better matching the TiO2 substrate, also resulting in less elastic strain energy. The resulting trend in strain relaxation fully explains why Ge incorporation improves the rocking curves in Fig. 3b. The two-Gaussian-shaped rocking curve is a well-understood phenomenon commonly observed during the strain relaxation of epitaxial films18.

The RSMs of three representative perovskite films are shown in Fig. 4d–f. Unlike the rutile films, each perovskite film is fully strained to the GdScO3 substrate. Again, one may notice small deviations between the film peak positions and their expected position based on the composition contours. For example, the x = 0 film peak in Fig. 4d is slightly north of the x = 0 contour. These discrepancies are probably due to a lack of accurate experimental Poisson ratios used to calculate the contours.

Valence state and occupation site of Ge in Ge-based oxides

One major challenge associated with the growth of high-quality oxides containing late transition metals (like Ru, Ni, and Cu) or main group metals (like Bi, Ge, and Sn) is achieving full oxidation of these high-electronegativity metals. To investigate the oxidation of Ge, we performed XPS on the rutile and perovskite films. Figure 5a shows Ge 3d core-level spectra of rutile films compared to that of a Ge reference wafer with native oxide; we mark the Ge 3d binding energy of different valence states for comparison, as shown at the top of Fig. 5a, b using assignments from Molle and coworkers19. We can clearly see that film peak position matches the Ge4+ position in the reference wafer, suggesting that Ge stays in the 4+ state in which it is delivered via GTIP.

Fig. 5: Valence state and site occupation of Ge in Ge-based oxides.
figure 5

a, b Ge 3d core-level hard X-ray photoelectron spectra (HAXPES) of Sn1-xGexO2/TiO2 (001) films (a) and SrGexSn1-xO3/GdScO3(110) films (b). The top of (a, b) shows a GeOx/n-Ge reference wafer to assist in oxidation state determination. The suboxide spectrum was vertically offset and expanded ×2. c Drift-corrected high-angle annular dark-field scanning transmission electron micrographs (STEM-HAADF) of the SrGexSn1-xO3/GdScO3 interfaces. d Energy dispersive X-ray spectroscopy (STEM-EDS) of the Sn L peak and electron energy-loss spectroscopy (STEM-EELS) of the Sr L and Ge L edges. Composite maps and integrated line profiles of the SrSn0.84Ge0.16O3 film show clear alignment of Sn and Ge signals.

Full size image

In the XPS of the perovskite samples shown in Fig. 5b, however, the film peak position better matches the Ge3+ position in the reference wafer, which suggests that Ge is in the 3+ state. However, this is unlikely because Ge is generally not stable in the 3+ state. The Ge in the reference wafer was oxidized by exposure to air where oxidation is limited by solid-state diffusion. Hence, the reference wafer can achieve the otherwise unattainable oxidation states of 1+ and 3+, which still only constitute a very small fraction of the analyzed volume. We posit that the unusual coordination environment for Ge in the perovskite structure results in a different binding energy for B-site Ge4+ than is observed in amorphous GeO2 due to differences in total electrostatic potential.

To determine the coordination environment of the Ge in the SrSnO3 host lattice, cross-sectional STEM was performed. Figure 5c shows STEM-HAADF images of two perovskite samples, in which image intensity scales with atomic number (Z~1.7). The images show high-quality interfaces free of dislocations, consistent with the conclusion of fully coherent films determined from RSMs in Fig. 4e, f. The x = 0.08 sample shows a subtle low-Z band at the interface. This feature might be a result of slight A:B-site nonstoichiometry due to effusion cell or substrate temperature transients caused by opening the shutter at the beginning of growth. Figure 5d shows atomic-resolution STEM-EDS and STEM-EELS elemental maps and line profiles of Ge, Sr, and Sn. The line profiles show clear alignment of Ge and Sn peaks occurring in the valleys of the Sr signal, demonstrating direct substitution of Sn with Ge on the B-site. Therefore, our experimental data indicate that Ge resides in an octahedral coordination environment in the perovskite lattice.

To further examine the site preference of Ge, we have carried out DFT calculations to determine the formation enthalpy of Sr1-xGexSnO3 (referred to as A-site alloys) and SrSn1-xGexO3 (referred to as B-site alloys) as a function of Sn, Ge, and Sr chemical potentials. We can define the preference of Ge occupying Sr- vs. Sn site by taking the difference in the formation enthalpies of A-site and B-site alloys as a function of Sn and O chemical potentials, such as:

$${\varDelta H}^{{{{{{\rm{f}}}}}}}\left[{{{{{{\rm{Sr}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{\rm{Sn}}}}}}{{{{{{\rm{O}}}}}}}_{3}\right] =\; {E}_{{{{{{\rm{t}}}}}}}\left[{{{{{{\rm{Sr}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{\rm{Sn}}}}}}{{{{{{\rm{O}}}}}}}_{3}\right]-{E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{SrSn}}}{{{\mbox{O}}}}_{3}\right)\\ +\left[{\left(1-x\right)E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{Sr}}}\right)+{\mu }_{{{\mbox{Sr}}}}\right]-x\left[{E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{Ge}}}\right)+{\mu }_{{{\mbox{Ge}}}}\right]$$
(1)

and

$${\varDelta H}^{{{{{{\rm{f}}}}}}}\left[{{{{{\rm{Sr}}}}}}{{{{{{\rm{Sn}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{{\rm{O}}}}}}}_{3}\right]=\; {E}_{{{{{{\rm{t}}}}}}}\left[{{{{{\rm{Sr}}}}}}{{{{{{\rm{Sn}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{{\rm{O}}}}}}}_{3}\right]-{E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{SrSn}}}{{{\mbox{O}}}}_{3}\right)\\ +\left[{\left(1-x\right)E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{Sn}}}\right)+{\mu }_{{{\mbox{Sn}}}}\right]-x\left[{E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{Ge}}}\right)+{\mu }_{{{\mbox{Ge}}}}\right]$$
(2)

where \({E}_{{{{{{\rm{t}}}}}}}\left[{{{{{{\rm{Sr}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{\rm{Sn}}}}}}{{{{{{\rm{O}}}}}}}_{3}\right]\) and \({E}_{{{{{{\rm{t}}}}}}}\left[{{{{{\rm{Sr}}}}}}{{{{{{\rm{Sn}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{{\rm{O}}}}}}}_{3}\right]\) are total energies of the alloys, \({E}_{{{{{{\rm{t}}}}}}}\left({{\mbox{SrSn}}}{{{\mbox{O}}}}_{3}\right)\) is the total energy of the host material, and Et (Sr), Et (Sn), and Et (Ge), and are total energy per atom of the Sr, Sn, and Ge bulk phases, to which the chemical potentials μSr, μSn, and μGe are referenced (μSr, μSn, μGe ≤ 0). These chemical potentials are not independent, but must satisfy the stability condition of the parent material SrSnO3, i.e.,

$${\mu }_{{{\mbox{Sr}}}}+{\mu }_{{{\mbox{Sn}}}}+3{\mu }_{{{\mbox{O}}}}={\varDelta H}^{{{{{{\rm{f}}}}}}}\left({{{{{\rm{SrSn}}}}}}{{{{{{\rm{O}}}}}}}_{3}\right)$$
(3)

avoiding the formation of secondary phases SrO, SnO2, and GeO2, i.e.,

$${\mu }_{{{\mbox{Sr}}}}+{\mu }_{{{\mbox{O}}}} \, < \, {\varDelta H}^{{{{{{\rm{f}}}}}}}\left({{{{{\rm{SrO}}}}}}\right)$$
(4)
$${\mu }_{{{\mbox{Sn}}}}+2{\mu }_{{{\mbox{O}}}} \, < \, {\varDelta H}^{{{{{{\rm{f}}}}}}}\left({{{{{\rm{Sn}}}}}}{{{{{{\rm{O}}}}}}}_{2}\right)$$
(5)

and

$${\mu }_{{{\mbox{Ge}}}}+{2\mu }_{{{\mbox{O}}}} \, < \, {\varDelta H}^{{{{{{\rm{f}}}}}}}\left({{{{{\rm{Ge}}}}}}{{{{{{\rm{O}}}}}}}_{2}\right)$$
(6)

Combining Eqs. (3) and (4), we obtain μSn + 2μO > \({\varDelta H}^{{{{{{\rm{f}}}}}}}\left({{{{{\rm{SrSn}}}}}}{{{{{{\rm{O}}}}}}}_{3}\right)-{\varDelta H}^{{{{{{\rm{f}}}}}}}\left({{{{{\rm{SrO}}}}}}\right)\), which together with Eqs. (5) and (6) limit the region in the μSn vs. μO diagram where SrSnO3 is stable, as shown in Fig. 6. The lines separating the region where the B-site alloys \({{{{{\rm{Sr}}}}}}{{{{{{\rm{Sn}}}}}}}_{1-x}{{{{{{\rm{Ge}}}}}}}_{x}{{{{{{\rm{O}}}}}}}_{3}\)have lower formation enthalpies than the A-site alloys Sr1−xGexSnO3, i.e., ΔHf[SrSn1−xGexO3] ≤ ΔHf[Sr1−xGexSnO3], are located in the upper right corner of the μSn vs μO diagram.

Fig. 6: Thermodynamic site preference of Ge in SrSnO3.
figure 6

Calculated region in the tin chemical potential (μSn) vs oxygen chemical potential (μO) plane showing where B-site alloys have lower formation enthalpy than the A-site alloys. The lines in the upper right corner (x = 0.0625, x = 0.125, and x = 0.25) separate the regions below which B-site alloys are preferred. This result indicates that for all allowed values of μSn and μO for which SrSnO3 is stable (orange region at the center), Ge will prefer to occupy the Sn site. The stability of SrSnO3 is limited by the formation of SrO on the left (dark-green line, corresponding to) and the formation of SnO2 on the right (light-green line), i.e., μSn +2μO \( > {\varDelta H}^{f}\left({{{{{\rm{SrSn}}}}}}{{{{{{\rm{O}}}}}}}_{3}\right)-{\varDelta H}^{f}\left({{{{{\rm{SrO}}}}}}\right)\), and obtain \({\mu }_{{{\mbox{Sn}}}}+2{\mu }_{{{\mbox{O}}}} < {\varDelta H}^{f}\left({{{{{\rm{Sn}}}}}}{{{{{{\rm{O}}}}}}}_{2}\right)\). The formation of GeO2 poses a lower limit to the oxygen potential, as indicated in the bottom region, i.e., \({\mu }_{{{\mbox{Ge}}}}+{2\mu }_{{{\mbox{O}}}} < {\varDelta H}^{f}\left({{{{{\rm{Ge}}}}}}{{{{{{\rm{O}}}}}}}_{2}\right)\). The formation enthalpy of A-site alloy will be lower than that of B-site alloy only in the upper right corner of the μSn vs. μO diagram, a region where SrSnO3 itself is unstable and SnO2 is favorable to form.

Full size image

We can see that SrSnO3 and SrSn1−xGexO3 alloys are only stable in the orange region at the center of Fig. 6, limited by the formation of SrO (left), GeO2 (below), and SnO2 (right). The conditions for which the formation enthalpies of the A-site alloys Sr1−xGexSnO3 would be lower than that of B-site alloys SrSn1−xGexO3 occur in a region of μSn and μO chemical potentials where the secondary phase SnO2 is most favorable to form, i.e., where SrSnO3 itself is unstable. This result clearly indicates that Ge prefers the Sn octahedral site over the Sr site.

We also note that in the A-site alloys, the Ge atom displays a large offsite displacement toward 3 of the original 12 nearest-neighbor O atoms (along the [110] pseudocubic direction), with Ge–O distances of ~2.06 Å. This further indicates that Ge2+ on the Sr site is unstable. In contrast, Ge sits on the Sn octahedral sites in SrSn1−xGexO3, with Ge–O distances of ~1.93 Å (equatorial) and of ~1.97 Å (apical), reflecting the tetragonal structure, and shows no offsite displacement. These results, again, reflect the fact that Ge strongly prefers the octahedral Sn site over the Sr site in SrSnO3, consistent with our experimental observations.

Conclusion

In summary, we have demonstrated the growth of epitaxial Sn1-xGexO2 and SrSn1-xGexO3 films via hybrid MBE. AFM, HRXRD, XPS, and STEM characterization shows that the GTIP precursor can be used as an effective source of Ge for the growth of both rutile and perovskite oxides while allowing excellent surface morphology and structural quality. DFT calculations indicate that Ge strongly prefers the Sn site in SrSnO3 as opposed to the Sr site. This work opens another synthetic route to achieving Ge-containing oxides. Future studies should build upon this work by exploring process parameters to achieve the pure germanate end-members, demonstration of a growth window, and applying hybrid MBE to other Ge-based oxides.

Methods

Film growth

Sn1-xGexO2 (x = 0, 0.28, 0.54) and SrSn1-xGexO3 (x = 0, 0.05, 0.08, 0.16) films were successfully grown using hybrid MBE. This approach employs a conventional solid source for Sr where necessary, hexamethylditin (HMDT) as a metal–organic precursor for Sn, germanium tetraisopropoxide (GTIP) as a metal–organic precursor for Ge, and an inductively coupled RF plasma for O. Rutile films were grown on TiO2(001) substrates at 600 °C, and perovskite films were grown on GdScO3(110) substrates at 950 °C. When Sr was used, its beam equivalent pressure (BEP)—measured by a retractable beam flux monitor—was fixed at 2.3 × 10−8 Torr. For HMDT, the liquid precursor crucible and injector temperature were held at ~60 °C to achieve adequate vapor pressure, whereas the delivery lines were maintained at a slightly higher temperature of ~75 °C to prevent precursor condensation in delivery lines. Likewise, for GTIP, the liquid precursor was maintained at ~50 °C, the lines were maintained at ~65 °C, and the injector was maintained at ~70 °C. The BEPs of HMDT and GTIP were varied to control the Sn:Ge ratio in the films, although the final Ge fraction was not necessarily proportional to the BEP ratio. The oxygen flow was set to 0.7 sccm to achieve an oxygen background pressure of 5 × 10−6 Torr while applying 250 W RF power to the plasma coil. Each rutile film was grown for 60 min, and each perovskite film was grown for 30 min.

Film characterization

Surface topography was measured with a Bruker Nanoscope V Multimode 8 atomic force microscope (AFM). A Rigaku SmartLab XE was used for X-ray scattering. High-resolution X-ray diffraction (HRXRD) 2θ−ω coupled scans were used to determine out-of-plane lattice parameters. Film thickness was determined from HRXRD finite thickness fringes when possible, or otherwise extracted from X-ray reflectivity (XRR). Reciprocal space maps (RSMs) were used to measure in-plane lattice parameters and determine strain relaxation. On each RSM, Vegard’s law was used to calculate relaxed lattice parameters and Poisson ratios were used to calculate how the lattice parameters would change under biaxial stress. For the rutile system, lattice parameters were taken from powder neutron diffraction20,21, and Poisson ratios were interpolated from elastic tensor data of the end-members22,23. For the perovskite system, substrate lattice parameters from Liferovich and coworkers24 were used, whereas the SrSnO3 tetragonal lattice parameters from Glerup and coworkers25 and the parameters of ambient temperature (quenched) SrGeO3 from Nakatsuka26 were used for the film. Due to a lack of elastic tensor data for SrGeO3, the DFT-predicted Poisson ratio for SrSnO3 of 0.19227 was used for the entire alloy series.

X-ray photoelectron spectroscopy (XPS) was used to determine the Ge fraction and valence. To determine composition, survey scans were measured using a Physical Electronics 5000 VersaProbe III photoelectron spectrometer with monochromatic Al Kα X-rays at the University of Minnesota. Empirical sensitivity factors from Wagner and coworkers were used28. To determine Ge valence, XPS was also performed at PNNL using an Omicron/Scienta R3000 analyzer with monochromatic Al Kα X-rays, a 100 eV analyzer pass energy, a 0.8 mm slit width, and a normal emission geometry. The resulting energy resolution was ~ 400 meV as judged by fitting the Fermi edge for a clean, polycrystalline Ag foil to the Fermi–Dirac function. The insulating nature of these films required the use of a low-energy electron flood gun to compensate the positive photoemission charge that builds up on the surface. The flood gun makes it possible to measure accurate core-level (CL) line shapes on insulating samples. However, the measured binding energies are artificially low due to overcompensation. At the same time, we sought to use a highly conductive n-Ge(001) crystal with its thin native oxide as an internal binding energy standard for Ge0, Ge2+, Ge3+ and Ge4+, as assigned by Molle and coworkers19. In order to compare Ge 3d binding energies from the insulating MBE-grown films to those from the GeOx/Ge standard, all samples were affixed to an insulating MgO(001) wafer in order to uniformly isolate them from the ground. The flood-gun beam energy was set to ~1 eV. The charging-induced binding energy shifts were close, but not identical, for the GeOx/n-Ge sample and the epitaxial films, as judged by the aliphatic C 1 s peak binding energy from the surface contamination. After correcting for differences in charging as judged by the C 1 s binding energies, all spectra were shifted by a constant amount such that the Ge 3d5/2 lattice peak in the GeO2/n-Ge spectrum fell at 29.4 eV, the value measured when GeO2/n-Ge(001) is mounted directly on the grounded sample holder without an MgO wafer for electrical isolation.

Cross-sectional scanning transmission electron microscopy (STEM) samples were prepared using a FEI Helios NanoLab DualBeam Ga+ Focused Ion Beam (FIB) microscope with a standard lift-out procedure. STEM high-angle annular dark-field (STEM-HAADF) images were collected on a probe-corrected JEOL GrandARM-300F microscope operating at 300 kV, with convergence semi-angle of 29.7 mrad and a collection angle range of 75–515 mrad. For the STEM-HAADF images shown in Fig. 5c, a series of ten frames was collected at 512 × 512 px sampling, with a dwell time of 2 µs px−1. The images were rigid-aligned using the SmartAlign program to minimize drift and then averaged to improve signal-to-noise29. Simultaneous STEM energy dispersive X-ray spectroscopy (STEM-EDS) and electron energy-loss spectroscopy (STEM-EELS) elemental mapping were used to determine site occupation, using the Sn L peak and Sr L and Ge L edges, respectively. For this configuration, a STEM-EELS acceptance angle range of 113–273 mrad was used. Mapping was performed using a dual JEOL Centurio detector setup for STEM-EDS and a 1 eV ch−1 dispersion, with a 4× energy binning in the dispersive direction for STEM-EELS. No denoising was applied, but the composite map shown in Fig. 5d was filtered using a smoothing kernel in Gatan Microscopy Suite 3.4.3.

First-principles calculations

First-principles calculations were performed to investigate whether Ge prefers the octahedral Sn site with oxidation state 4+, or the Sr site with oxidation state 2+. The calculations are based on density functional theory30,31 within the generalized gradient approximation revised for solids (PBEsol32), with projector augmented wave potentials33,34 as implemented in the VASP code35,36. We considered both SrSn1-xGexO3 and Sr1-xGexSnO3 using supercells containing 80, 40, and 20 atoms representing concentrations x = 0.0625, 0.125, and 0.25, respectively. We used an energy cutoff of 500 eV for the plane wave expansions and meshes of k-points that are equivalent to 6 × 6 × 4 for the 20-atom cell of the tetragonal SrSnO3. All the atom positions in the cell were allowed to relax. To simulate the epitaxial growth of the alloy on GdScO3(110) substrates, we fixed the in-plane lattice parameters to that of GdScO3, allowing the out-of-plane lattice parameter to relax together with all the atomic positions, minimizing the stress tensor and the total energy.