Growth of α-Ga₂O₃ on α-Al₂O₃ by conventional molecular-beam epitaxy and metal–oxide-catalyzed epitaxy

Over the past decade, Ga₂O₃ has gained much attention as a wide-band gap semiconductor. Monoclinic β-Ga₂O₃ possesses an ultra-wide bang gap of ∼4.7 eV, ¹⁾ and it has been the most studied phase owing to its thermal stability and the availability of large-area, native, semi-insulating, and conductive substrates. ^2,3) To further increase its band gap β-Ga₂O₃ can be alloyed with Al to form β-(Al,Ga)₂O₃, but achieving a high Al content has remained challenging due to the tendency to have phase segregation. ⁴⁾ In contrast, α-(Al,Ga)₂O₃ becomes more stable as the Al is increased because the crystal is isostructural with the α-Al₂O₃ substrate and the lattice mismatch is reduced as the Al concentration is increased. ⁵⁾ This has enabled the entire compositional range of α-(Al,Ga)₂O₃ to be readily achieved, ^5,6) and it has enabled band gaps exceeding those of AlN, BN, or diamond. ^7,8)

With the recent advances enabling α-Ga₂O₃ to remain stable during high-temperature anneals, ⁹⁾ the next challenge is to achieve electrical conductivity. To date, conductivity in α-Ga₂O₃ has been achieved by chemical vapor deposition (CVD), ^10,11) but has remained elusive for films grown by molecular-beam epitaxy (MBE). Additionally, conductive β-Ga₂O₃ films grown by MBE on α-Al₂O₃ have yet to be achieved. ⁷⁾ While the exact reasons these films remain insulating are unknown, the thermodynamics during MBE growth and the low formation energy of defects may cause these Ga₂O₃ films on Al₂O₃ to be insulating.

Multiple compensating point defects (e.g. cation vacancies, oxygen vacancies, donor impurities ^12,13)) and extended crystallographic defects (e.g. rotational domains and threading dislocations ¹⁴⁾) occur within the Ga₂O₃ films grown on sapphire. Reasons for the emergence of these defects include the lattice mismatch between the film and the substrate, ¹⁴⁾ and the growth regime in which the films are grown. ^12,13,15) For example, Ga vacancies (V_Ga) may act as compensating acceptors for introduced n-type dopants in grown Ga₂O₃ thin films. ¹⁵⁾ O-rich growth environments are likely to generate a significant amount of V_Ga (due to their low formation energy) whereas Ga-rich growth environments are found to significantly suppress the formation of V_Ga (due to their high formation energy). ¹³⁾ Thus, the growth of Ga₂O₃ in the highly Ga-rich regime—accessed by new epitaxial growth concepts ^16,17)—may improve the transport properties of Ga₂O₃ thin films since the Ga-rich growth regimes lead to higher V_Ga formation energies, resulting in lower V_Ga densities within the Ga₂O₃ layers.

One approach to address these issues is through the use of metal–oxide-catalyzed epitaxy (MOCATAXY). ¹⁸⁾ This is a growth process where a catalytic element (e.g. In) is introduced to the growth system and results in metal-exchange catalysis. ¹⁹⁾ This growth mode has been observed for β-(Al,Ga)₂O₃ on different substrates and surface orientations, as well as for different epitaxial growth techniques. ^20–24)

Many benefits arise from using MOCATAXY during the growth of Ga₂O₃. For example: (i) it can improve the surface morphologies of β-Ga₂O₃-based films. ²¹⁾ (ii) The synthesis of Ga₂O₃ can occur in previously inaccessible kinetic and thermodynamic growth regimes (e.g. in highly metal-rich regimes) which can be beneficial for the suppression of undesired point (such as V_Ga) defects in Ga₂O₃. ^12,13,19) (iii) The formation of thermodynamically unstable Ga₂O₃ phases becomes energetically favorable, ^16,19,24) e.g. the formation of the /κ-phase of Ga₂O₃, which has enabled novel /κ-Ga₂O₃-based heterostructures. ²³⁾ (iv) The growth rate (Γ), possible growth temperatures (T_G), and crystalline quality of β-(Al,Ga)₂O₃-based thin films can be vastly enhanced. ¹⁸⁾

In this work, we introduce the growth of α-Ga₂O₃ by MOCATAXY, resulting in an expansion of the α-Ga₂O₃ growth window combined with an increased Γ and an improvement in its out-of-plane mosaic spread. It is the first demonstration of a catalytic growth process during the growth of α-Ga₂O₃.

Samples were grown in a Veeco GEN930 plasma MBE system with standard Ga and In effusion cells. For all samples, the substrates were cleaned in acetone and isopropanol for 10 min and the α-Ga₂O₃ samples were grown for 60 min. The growth temperature (T_G) was measured by a thermocouple located within the substrate heater. The Ga flux (ϕ_Ga) and In flux (ϕ_In) were monitored by beam equivalent pressure (BEP) chamber readings. For conventional MBE and MOCATAXY, the O₂ flux (ϕ_O) was measured in standard cubic centimeters per min (SCCM) and a radio-frequency plasma power of 250 W was employed during all growths. To convert the measured values of ϕ_Ga (BEP), ϕ_In (BEP), and ϕ_O (SCCM) into units of nm⁻² s⁻¹ conversion factors are taken from Ref. 25. Note, when using In-mediated catalysis, the available ϕ_O for Ga to Ga₂O₃ oxidation is about 2.8 times larger than for Ga oxidation in the absence of In. ^16,19)

For samples grown by conventional MBE and MOCATAXY, the impact of ϕ_Ga and T_G is studied. In the case of MOCATAXY growth, the impact of ϕ_In is also investigated. All the growth parameters used in this work are collected in Table I. For scanning transmission electron microscopy (STEM), samples were prepared using a Thermo Fisher Helios G4 UX Focused Ion Beam with a final milling step of 5 keV to minimize damage. Carbon and Au–Pd layers were sputtered to reduce charging during sample preparation. Carbon and platinum protective layers were also deposited to minimize ion-beam damage. STEM measurements were taken with an aberration-corrected Thermo Fisher Spectra 300 CFEG operated at 300 keV.

Figure 1 shows the growth-rate-diagram of α-Ga₂O₃($10\bar{1}0$) grown on α-Al₂O₃($10\bar{1}0$) by conventional MBE (the gray shaded area) and MOCATAXY (the purple shaded area). For conventionally grown samples two distinct growth regimes are observed: (i) the O-rich regime where O adsorbates are in excess over Ga adsorbates (i.e. the Ga flux limited regime), and (ii) the Γ-plateau regime (i.e. the Ga₂O desorption limited regime). The O-rich regime is characterized by an increasing Γ with increasing ϕ_Ga, whereas the plateau regime is characterized by a constant Γ, being independent of ϕ_Ga. Within this regime, however, Γ may decrease with increasing T_G (see inset in Fig. 1) as the desorption of the volatile suboxide Ga₂O becomes thermally more active. ²⁶⁾ The data in the inset of Fig. 1 plots Γ as a function of T_G for (i) α-Ga₂O₃ grown the O-rich regime and (ii) α-Ga₂O₃ grown in the Γ-plateau regime.

Zoom In Zoom Out Reset image size

To expand the accessible growth window of α-Ga₂O₃ to higher ϕ_Ga and to higher T_G, combined with increased Γ and improved crystalline quality, In-mediated catalysis was employed in the formation of α-Ga₂O₃. ¹⁹⁾ The red stars in Fig. 1 show the resulting Γ as a function of ϕ_Ga at constant T_G. The gray shaded and purple shaded areas in Fig. 1 depict model-based descriptions of Γ for α-Ga₂O₃ grown by conventional MBE and MOCATAXY, respectively. The maximum Γ obtained for each growth technique is ${\rm{\Gamma }}\approx 1.5\,{\rm{nm}}\,{{\rm{\min }}}^{-1}$ and ${\rm{\Gamma }}\approx 3.3\,{\rm{nm}}\,{{\rm{\min }}}^{-1}$, respectively. Using MOCATAXY, a more than 2-times increase in Γ for α-Ga₂O₃ at given growth conditions, as well as a shift far into the adsorption-controlled regime (i.e. far into the Ga-rich flux regime) is observed. This direct comparison between the two growth types clearly shows the expanded growth window made possible with MOCATAXY, for example, enabling ${\rm{\Gamma }}\approx 1.8\,{\rm{nm}}\,{{\rm{\min }}}^{-1}$ for α-Ga₂O₃ at ϕ_Ga = 5.5 nm⁻² s⁻¹. In contrast, at these growth conditions, no growth of α-Ga₂O₃ is obtained by conventional MBE. The catalytic effect on Γ of α-Ga₂O₃ is modeled as a function of ϕ_O within the supplemental section. ²⁷⁾ We note that the depicted models use arbitrary kinetic parameters, based on kinetic parameters extracted for the growth ofβ-Ga₂O₃. ²⁸⁾

To describe the growth of α-Ga₂O₃ by MOCATAXY, ϕ_O is scaled by a factor of 2.8 compared with the growth of α-Ga₂O₃ by conventional MBE. This additional O comes from the catalytic nature of In forming a catalytic adlayer (A) with O adsorbates, e.g. A= In–O, which provides more active O for the Ga to α-Ga₂O₃ oxidation. In other words, A increases the reaction probability of Ga with O on the respective growth surface, facilitating the formation of the final Ga₂O₃ compound at much higher ϕ_Ga and T_G, which enables excellent crystal quality. ^16,19) We further note that the same factor of 2.8 was needed for modeling the MOCATAXY growth of β-Ga₂O₃ on different substrates and different surface orientations. ^16,19) We note, however, that for a quantitative extraction of all kinetic growth parameters more Γ-studies of α-Ga₂O₃ are needed and are beyond the scope of this work. Nevertheless, the models help validate the Γ-data and provide insight into the growth regimes and growth mechanisms of α-Ga₂O₃. For example, once ϕ_Ga exceeds the active O flux, i.e. ϕ_Ga > ϕ_O, the growth will enter the Ga-rich regime and Γ will start to decrease, as shown by the gray shaded area in Fig. 1. Thus, this is the first direct indication that the growth of α-Ga₂O₃ is limited by the formation and subsequent desorption of Ga₂O, like what is observed for β-Ga₂O₃ grown by conventional MBE. ²⁸⁾

Figure 2 directly compares the impact of both MBE growth techniques on the structural quality of the epitaxially grown films. In Fig. 2 (a), 2θ-ω XRD scans of two selected α-Ga₂O₃ films are shown, one grown by conventional MBE (depicted as the blue trace) and one grown by MOCATAXY (depicted as the red trace). The reflections of the films coincide with the α-Ga₂O₃ $30\bar{3}0$ peak. This, along with the absence of other diffraction peaks, indicates phase-pure α-Ga₂O₃(10 $\bar{1}$ 0) with In incorporation of <1% in the grown α-Ga₂O₃ layers, similar to what is observed for β-(Al,Ga)₂O₃ grown by MOCATAXY. ¹⁸⁾ Figures 2(b) and 2(c) plot transverse scans (rocking curves) for the conventional MBE and MOCATXY grown α-Ga₂O₃ samples as plotted in Fig. 2(a). The rocking curves are measured across the symmetric $30\bar{3}0$ peak. The full width at half maxima (FWHM) of ω quantifies the out-of-plane mosaic spread of the α-Ga₂O₃ film. For conventionally grown films the out-of-plane crystal distribution is Δω ≈ 0.55° and for MOCATAXY grown films it is Δω ≈ 0.45°. The film thicknesses d of the conventionally and MOCATAXY grown films are d = 73 nm and d = 127 nm, respectively. Jinno et al. reported that α-Ga₂O₃ films are fully relaxed for d > 60 nm. ⁵⁾ Since lattice mismatch and relaxation are not impacted by MOCATAXY, it is noteworthy that despite the MOCATAXY film being thicker, Δω is substantially smaller compared to what is obtained by conventional growth. The same MOCATAXY grown sample shown here is studied by STEM in Fig. 4.

Zoom In Zoom Out Reset image size

Surface morphologies and root mean square roughnesses (R_q) are measured by AFM and depicted in Figs. 2(d) and 2(e). The best surface roughness for conventionally grown α-Ga₂O₃ with d = 66 nm is R_q = 0.64 nm, while the smoothest one for MOCATAXY grown samples with d ∼ 270 nm has an R_q = 0.94 nm. The larger surface roughness for the MOCATAXY grown sample is likely due to facetting on the top surface of the α-Ga₂O₃ thin film [see Fig. 4(a)]. We speculate that In does not only act as a catalyst but also acts as a surface active agent (surfactant) for the growth α-Ga₂O₃ thin films. It is widely understood that In can act as a surfactant for the epitaxial growth of GaN-based films, ²⁹⁾ and has also been observed during the growth of β-Ga₂O₃ ²¹⁾ and β-(Al,Ga)₂O₃. ¹⁸⁾ Depending on the growth conditions and growth surface, which can affect the surface diffusion kinetics, surface chemical potentials, and the assessed growth mode, the suppression of facetting may be accomplished through the use of optimized conditions, while using In as a surfactant, enabling a modification in the surface free energies of the growing α-Ga₂O₃ thin film and a change in its growth mode. ^18,21,30,31) However, surfactant-induced morphological phase-transitions from two-dimensional (2D) layer growth to three-dimensional (3D) island growth have also been observed during MBE growth. ³²⁾ We believe that a similar effect occurs for the α-Ga₂O₃ surfaces studied here when In may act as an (anti)surfactant during the growth of these films. Note, we have not fully explored all growth regimes made accessible through MOCATAXY in this study. Further studies may lead to additional improvements in the crystalline quality and surface morphologies of the α-Ga₂O₃ thin films.

In Figs. 3(a) and 3(b), the impact of ϕ_Ga and T_G, respectively, on Δω for samples grown by conventional MBE in the O-rich regime (blue squares) and in the Γ-plateau regime (green circles), as well as for samples grown by MOCATAXY (red stars), are shown. XRD data and Δω are obtained by the same methods as described above for Fig. 2. Within the O-rich regime at T_G = 640 ^◦C, a large Δω is observed, Fig. 3(a). At higher growth temperatures (i.e. T_G ≥ 660 ^◦C), Δω are similar (or slightly improving) with increasing temperature, regardless of growth regime. We speculate that the reason the crystal quality improves with T_G, is that there is an increase in the kinetic energy and a subsequent increase in the diffusion length of the adsorbates, allowing the Ga and O to reach the proper lattice site. However, if T_G is increased too much, a decrease in the surface lifetime of Ga adsorbates may occur, resulting in a reduction in the crystalline quality of the growing thin films. Using MOCATAXY in the Ga-rich regime with a fixed T_G, excess Ga may now reduce the needed surface diffusion length, improving the crystalline quality in the obtained α-Ga₂O₃ layers. More studies to separate the effects of ϕ_Ga and T_G on Δω need to be performed, but to the best of our knowledge, the obtained Δω values are the lowest reported in the literature for α-Ga₂O₃ grown on α-Al₂O₃.

Zoom In Zoom Out Reset image size

Finally, to directly quantify and identify how MOCATAXY affects the crystal structure of α-Ga₂O₃ thin films, high-angle annular dark-field STEM (HAADF-STEM) was performed along the $\langle 0\bar{1}10\rangle$ zone axis, and is plotted in Fig. 4. The sample shown here is the same as the one shown in Fig. 2(c). In Fig. 4(a), a clear contrast differentiates the sapphire substrate, the epitaxial film (α-Ga₂O₃), and the protective Au–Pd sputtered coating. The bright contrast observed at the substrate/film interface (see Fig. 4(b) and Ref. 27) is due to additional scattering of the electron beam and indicates the presence of misfit dislocations. These dislocations arise due to the film relaxation caused by strain. A subset of the observed misfit dislocations propagate and lead to threading dislocations. From the contrast variation observed within the film [see Fig. 4(a)], an average frequency of one threading dislocation every 30 nm laterally along the film/substrate interface is observed. While more investigation is needed to determine the cause of the faceting and verify the above hypothesis (e.g. due to the changed growth mode when using In-mediated catalysis), it is observed that the threading dislocations can merge and then continue to propagate toward the film surface. These dislocations terminate at the bottom of intersecting surface planes, where faceting along the ($10\bar{1}1$) plane is observed. The complimentary facet is unidentified since the facet is not perpendicular to the beam and tilts out-of-plane. This tilting is detected in Fig. 4(a) by the fading of contrast along the surface, in contrast to the sharp change in contrast on the ($10\bar{1}1$) plane.

Zoom In Zoom Out Reset image size

Figure 4(b) shows an enlarged image of the film/substrate interface. A pair of edge dislocations is observed and is highlighted with their Burgers circuits. This edge dislocation pair is observed along the film/substrate interface, and its dislocation density is estimated to be 5 × 10⁵ cm⁻¹ (or ∼10¹¹ cm⁻²), i.e. occurring every 20 nm. This is similar to what is reported by conventional MBE. ⁵⁾ To quantify Al/Ga inter-diffusion at the interface, a line scan (see S-Fig. 2 Ref. 27) was performed to quantify the contrast change. An interface width of σ ≈ 0.9 nm was measured from an error function fitted to the Al intensity line scan profile (see S-Fig. 2 Ref. 27).

A fast Fourier transform (FFT), of the interface region shown in Fig. 4(b), is displayed in Fig. 4(c). A thin film completely strained to the substrate will show a singular diffraction peak. However, when the film relaxes its interplanar spacing d_hkl changes, resulting in an additional peak, shifted from the substrate peak. However, shifted peaks in the in-plane direction are not visible because the α-Ga₂O₃ ($000\bar{6}$) reflection peak is approximately 10x weaker than in α-Al₂O₃. The strain relaxation is observed in the $20\bar{2}\bar{4}$ and $10\bar{1}10$ diffraction peaks of α-Ga₂O₃. The strain relaxation is accomplished by the formation of edge dislocations at the interface, where the $20\bar{2}\bar{4}$ peak is correlated to the yellow Burgers circuit and the $10\bar{1}10$ peak to the cyan Burgers circuit. In addition, no phase separation or secondary phases were observed by STEM within the α-Ga₂O₃ film grown by MOCATAXY. However, a bi-layer structure from overlapping α-Ga₂O₃ grains when viewed in projection is observed with a slip along the $[10\bar{2}\bar{2}]$ direction (see S-Fig. 3 Ref. 27). The presence of this bi-layer structure indicates that the film is not single-crystalline. The bi-layer structure was confirmed using an ab initio TEM (abTEM) simulation ³³⁾ which produced a matching HAADF image from the crystallographic information framework.

This TEM investigation of MOCATAXY grown α-Ga₂O₃ shows comparable crystal quality to what is measured for conventional MBE ⁵⁾ with regards to edge dislocation density and phase purity. We note that the difference in projection direction may have prevented imaging of the bi-layer structure in this previous report. No faceting of α-Ga₂O₃ was observed by conventional MBE when grown on m-plane α-Al₂O₃. ^5,9)

Phase-pure α-Ga₂O₃($10\bar{1}0$) on α-Al₂O₃($10\bar{1}0$) was grown using conventional MBE and MOCATAXY with thickness up to d = 262 nm. We mapped out the Γ-dependence on ϕ_Ga and T_G and its impact on the crystalline quality and surface morphologies. We identified and explored previously inaccessible growth regimes by MOCATAXY, and showed that it vastly extends the growth regime and improves the out-of-plane mosaic spread of the grown α-Ga₂O₃ films. Using In-mediated catalysis, we observe facetting on top of the α-Ga₂O₃($10\bar{1}0$) layers. This study confirms that this new MBE growth mode can be applied to the growth of α-Ga₂O₃–and is not limited to the growth of the β-Ga₂O₃ and β-(Al,Ga)₂O₃ polymorphs. We emphasize more studies are needed to determine the kinetic parameters that form α-Ga₂O₃ during conventional MBE and MOCATAXY growth, as well as to further improve the quality of the grown α-Ga₂O₃/α-Al₂O₃ heterostructures, and to understand the mechanisms leading to the surface faceting of α-Ga₂O₃.

This research is supported by the Air Force Research Laboratory-Cornell Center for Epitaxial Solutions (ACCESS), monitored by Dr. Ali Sayir (FA9550-18-1-0529). J. P. M. acknowledges the support of a National Science Foundation Graduate Research Fellowship under Grant No. DGE2139899. M. A.-O. acknowledges financial support from the Central Research Development Fund (CRDF) of the University of Bremen. This work makes use of PARADIM under Cooperative Agreement No. DMR-2039380. This work uses the CCMR and CESI Shared Facilities partly sponsored by the NSF MRSEC program (DMR-1719875) and MRI DMR-1338010, and the Kavli Institute at Cornell (KIC).