Systematic investigation of the growth rate of β-Ga₂O₃(010) by plasma-assisted molecular beam epitaxy

For high-power and high-temperature vertical electronic devices, β-Ga₂O₃ is one of the promising materials owing to its wide band gap (4.9 eV)^1,2) and predicted high breakdown field (8 MV/cm).³⁾ Bulk β-Ga₂O₃ can be grown by highly scalable, low-cost, melt-based methods,^4–7) in comparison to other power-device materials, such as SiC, GaN, and diamond. Bulk β-Ga₂O₃ enables the growth of high-quality β-Ga₂O₃ epitaxial films. Recently, metal–oxide–semiconductor field-effect transistors (MOSFETs) of β-Ga₂O₃ homo-epitaxially grown by molecular-beam epitaxy (MBE) were demonstrated to have excellent properties.⁸⁾ However, there are few reports of detailed investigations of β-Ga₂O₃ epitaxial growth, including those concerning its growth rate, growth mode, and crystalline quality. The clarification of the growth kinetics of β-Ga₂O₃ films will allow further improvements in β-Ga₂O₃ devices.

A β-Ga₂O₃ single crystal has a thermodynamically stable monoclinic structure (b-axis unique) belonging to the space group C2/m.^9,10) The crystal structure of β-Ga₂O₃ has two crystallographically non-equivalent Ga positions (tetrahedrally and octahedrally coordinated) and two cleavage planes [(100) and (001)].^10,11) The cleavage planes have a high decomposition rate in vacuum or in environments with low oxygen partial pressure, such as those realized in MBE systems, so that the epitaxial growth on the cleavage planes yields a low growth rate. High epitaxial-growth rates are desirable for the development of electron devices. The dependence of β-Ga₂O₃ growth rates on the growth planes was investigated in an MBE system using an $\text{ozone}:\text{oxygen}$ ($1:20$) gas mixture.¹²⁾ The β-Ga₂O₃(100) cleavage plane was grown at a rate of 0.2 nm/min, while the β-Ga₂O₃(010) non-cleavage plane was grown at a substantially higher rate of >2 nm/min. The (010) non-cleavage plane is practical for β-Ga₂O₃ epitaxial growth.

For the β-Ga₂O₃(010) homo-epitaxial growth, an ozone MBE system was used. The ozone MBE system has the advantage of higher growth rates of β-Ga₂O₃(010) compared with a reported maximum rate of 10 nm/min.³⁾ However, the precise control of the intentional doping densities in β-Ga₂O₃(010) films grown by an ozone MBE is difficult because of the high background pressure (>10⁻⁴ Torr). In our study, we used a plasma-assisted (PA) MBE system (PAMBE) with a comparatively lower background pressure of ∼10⁻⁵ Torr. In β-Ga₂O₃ growth using PAMBE, the Ga/O ratio influences the growth rate. In β-Ga₂O₃(100) growth, the growth rate decreases under metal-rich conditions as the stable Ga₂O₃ surface oxide is converted to a sub-oxide (Ga₂O) with a high vapor pressure: Ga₂O₃ + 4Ga → 3Ga₂O.^13–15) Under Ga-rich conditions, the overall growth rate of Ga₂O₃ is limited by the competition between the Ga₂O₃ deposition and the Ga₂O desorption. In this study, we systematically investigate the growth rate and growth mode of β-Ga₂O₃(010) films grown by PAMBE.

Sn-doped n-type β-Ga₂O₃(010) substrates (Tamura) were used for the growth studies presented here. For all substrates, the (010) surface was treated by chemical–mechanical polishing (CMP) and had an rms roughness of 0.1 nm over a 1 × 1 µm² area. After solvent cleaning with acetone and isopropanol, the β-Ga₂O₃ substrates were indium bonded to a silicon backing wafer and subsequently baked at 220 °C for 2 h in ultra-high vacuum. On the β-Ga₂O₃(010) substrates, Ga₂O₃ was homoepitaxially grown using a Varian/Veeco 620 MBE system, equipped with an effusion cell to evaporate liquid Ga (99.99999%) and a Veeco Uni-Bulb RF plasma cell to produce active oxygen. Liquid Sn (99.9999%) was used as an n-type doping source of β-Ga₂O₃(010). Prior to the film growth, the elemental beam equivalent pressures (BEPs) of Ga and Sn were measured using a nude ion gauge located in the substrate position. For the oxygen plasma source, research-grade O₂ gas (99.998%) was supplied through an in-line SAES pure gas dryer by an MKS pressure control module. Before the growth, an oxygen plasma treatment was carried out at a BEP of 1.2 × 10⁻⁵ Torr at 900 °C for 30 min to remove adsorbates from the substrate surface. We consider the internal thermocouple temperature near a heater behind the Si backing wafer as the growth temperature. Metallic Al (melting temperature 660 °C) on the Ga₂O₃ substrate melted at the thermocouple temperature of 645 °C. The surface morphologies and structures of the Ga₂O₃ layers were evaluated using atomic force microscopy (AFM; Veeco/Digital Instruments Dimension 3000) and reflection high-energy electron diffraction (RHEED), respectively. The RHEED measurements were conducted with an electron-beam energy of 7 keV.

For the determination of the substrate vicinity (intentional or unintentional miscut), the crystal orientation was determined using high-resolution X-ray diffraction (XRD; X'pert PRO) by locating the (110) and (022) peak positions in a skew-symmetric geometry [inclined by 14.3 and 37.7° from (010), respectively]. This measurement revealed that the (100) plane could be cleaved most easily. The miscut components were determined by aligning the substrate surface parallel to the incident beam (measurements at 2θ = 0°), setting ω = 0°, and then determining the offset from 1/2(2θ)–ω to satisfy Bragg's law. The substrate was subsequently rotated by 90° about its surface normal and the procedure was repeated. The substrate miscut was 2.2° in a counterclockwise direction of 17° from the [001] direction (Fig. 1).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The growth rate of the β-Ga₂O₃(010) epitaxial films was determined using thickness fringes from XRD ω–2θ scans of the (020) diffraction plane. In wide-angular range 2θ–ω scans (2θ = 10–70°), the (010) and (020) diffraction peaks were detected, indicating that β-Ga₂O₃(010) single-crystal films were grown. The (020) ω–2θ scan diffraction peak of a 135-nm-thick β-Ga₂O₃(010) epitaxial film is shown in Fig. 2(a). Despite the homoepitaxial growth, up to 7th order satellite thickness fringes were clearly observed in high-resolution ω–2θ scans. The interface offsets are a result of a slight rigid-body displacement of the epitaxial film from the underlying substrate due to non-stoichiometry or impurities at the regrowth interface.¹⁶⁾

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The growth rate of the β-Ga₂O₃(010) films strongly depended on the O₂ gas flux at a plasma power of 200 W [Fig. 2(b)]. At an oxygen flux of 1.2 × 10⁻⁵ Torr (BEP), the growth rate reached a peak of 2.2 nm/min. To establish which species were responsible for the growth, the emission of the oxygen plasma was measured by optical emission spectroscopy (OEM) from a viewport at the back of the oxygen plasma source. The light emission of the oxygen plasma displayed sharp peaks at wavelengths of 616, 777, and 845 nm, which are attributed to transitions of the excited oxygen.¹⁷⁾ The flux of active atomic oxygen, or the oxygen plasma emission intensities, should have a strong influence on the growth rate of β-Ga₂O₃ by PAMBE. However, at a plasma power of 200 W, the oxygen plasma emission intensities continuously increased with an increasing O₂ gas flux from 0.4 to 2.9 × 10⁻⁵ Torr (BEP), i.e., there was no peak in the plasma emission intensities for the O₂ gas flux. The plasma emission intensity was measured not from a substrate surface, but from the back side of the oxygen plasma source. Further investigation is required to clarify the specific active atomic oxygen fluxes at the substrate surface. Attempts to grow β-Ga₂O₃ with the same oxygen flow but with the plasma source turned off were unsuccessful, i.e., the growth of β-Ga₂O₃(010) required active atomic oxygen from the plasma source. For the remaining growth processes, the rf plasma power and oxygen flux were maintained at 200 W and 1.5 × 10⁻⁵ Torr, respectively, corresponding to 1 nm/min of O^* flux. 1 nm/min is equivalent to 0.11 monolayer (ML)/s, where 1 ML of Ga₂O₃(010) corresponds to b/2 = 0.15 nm.

The RHEED patterns of the β-Ga₂O₃(010) substrate surfaces are shown in Fig. 3. With a 104° substrate rotation, the RHEED patterns were observed along the [100] and [001] azimuths. Because of the small miscut (<0.5°) parallel to the [100] direction, the RHEED pattern exhibited vertical rods for the [001] azimuth [Fig. 3(a)]. The as-received substrate had vague streaky 1 × 2 RHEED patterns [Fig. 3(a)], while after the oxygen plasma treatment the surface had the streaky 1 × 2 RHEED patterns associated with smooth surfaces [Fig. 3(b)]. During growth under temperatures lower than 450 °C, the 1 × 2 RHEED patterns became spotty, indicating that growth temperatures above 450 °C are required for a smooth surface. The other surface reconstruction was not observed during or after growth. The RHEED intensity oscillation was not observed during the growth, indicating a continuous step-flow growth mode due to a large miscut angle. The realization of a layer-by-layer growth mode requires on-axis substrates which have sufficiently large terrace widths to enable island formation. After opening the Ga and O^* plasma shutter at growth temperatures below 600 °C under Ga-rich conditions, the RHEED intensity sharply decreased with decreasing growth temperature [Fig. 3(c)]. We propose that excess Ga adlayers on the β-Ga₂O₃(010) surface exist under Ga-rich conditions, causing an increase in diffuse scattering. For growth temperatures above 600 °C, the RHEED intensity was constant, which was attributed to desorption of excess Ga from the surface.

Zoom In Zoom Out Reset image size

To calibrate the Ga flux (Φ_Ga) and the O^* flux ($\Phi _{\text{O}^{*}}$) in Ga₂O₃ growth-rate units (nm/min), we used thickness fringes from XRD ω–2θ scans of the (020) diffraction plane for Ga₂O₃ grown under Ga- and O-limited conditions, respectively.¹⁸⁾ The dependence of the Ga₂O₃ growth rate on an impinging Ga flux at T = 500 and 700 °C is shown in Fig. 4(a). Under Ga-limited conditions ($\Phi _{\text{Ga}} - \Phi _{\text{O}^{*}} < 0$), the growth rate at T = 500 and 700 °C proportionally increased with the Ga flux, indicating that the impinging Ga flux was incorporated into the Ga₂O₃ films. At T = 700 °C, the growth rate reached a peak at 1 nm/min under stoichiometric conditions ($\Phi _{\text{Ga}} - \Phi _{\text{O}^{*}} = 0$) and remained at 1 nm/min for Ga-rich conditions ($\Phi _{\text{Ga}} - \Phi _{\text{O}^{*}} > 0$). At T = 700 °C under metal rich conditions [e.g., up to Φ_Ga (= 20 nm/min)], no Ga droplets were observed on the sample surface. This was consistent with a constant RHEED intensity during growth above 600 °C for metal-rich growth. We attribute the constant streaky RHEED pattern for metal-rich growth at high temperatures to a continuous step-flow or step-flow-like growth without the formation of Ga metal droplets on the surface. Presumably, excess Ga adatoms desorbed from the β-Ga₂O₃(100) surface for T > 600 °C. Surprisingly, the growth rate at T = 500 °C decreased with increasing Ga flux. Low temperature β-Ga₂O₃ growth shows similar tendencies as β-Ga₂O₃(100) growth,¹⁵⁾ in which excess Ga adatoms form the volatile Ga₂O sub-oxide and reduce the growth rate under Ga-rich conditions. Although speculative, we believe that the residency time for excess Ga on the (010) surfaces at high temperatures (>600 °C) was short enough for the direct desorption of Ga, whereas at lower temperatures the reaction of excess Ga with activated oxygen favored the formation of the volatile Ga₂O sub-oxide.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The growth rate dependence on the growth temperature under slightly Ga-rich conditions ($\Phi _{\text{Ga}} - \Phi _{\text{O}^{*}} = 0.3\,\text{nm}/\text{min}$) is shown in Fig. 4(b). For a growth temperature between 650 and 750 °C, the growth rate was constant at 1 nm/min. For T > 750 °C, the growth rate decreased with increasing growth temperature and was independent of the Ga/O ratio (>1). We assume that the reduced growth rate at a high growth temperature resulted from the decomposition of the β-Ga₂O₃(010) films. For T < 650 °C, the growth rate was reduced with decreasing growth temperature, yielding a low growth rate at 500 °C under slightly Ga-rich conditions [Fig. 4(a)]. As mentioned above, the RHEED intensity continuously decreased during growth under Ga-rich conditions for T < 600 °C. We consider that the excess Ga adlayer for T $\lesssim$ 600 °C prevents oxygen atoms from incorporating into the Ga₂O₃ films, resulting in the reduction of the growth rate. Another possibility is the formation of volatile gallium sub-oxides (Ga₂O). For T < 600 °C, the excess Ga reacts with Ga₂O₃ surface oxides, resulting in the formation and desorption of Ga₂O. However, for T $\gtrsim$ 600 °C, the excess Ga can be desorbed from the surface without the formation of Ga₂O. To verify the mechanisms responsible for the reduced growth rate at low and high growth temperatures, further studies are planned using line-of-sight quadrupole mass spectroscopy (QMS).¹⁹⁾

The surface morphologies of ∼60-nm-thick β-Ga₂O₃(010) films are shown in Fig. 5. For T < 900 °C, the Ga₂O₃ layer had a smooth surface. The rms roughness in Figs. 5(a), 5(b), 5(c), and 5(d) was 0.1, 0.3, 0.6, and 0.1 nm, respectively, over 1 × 1 µm². The smooth surface of the β-Ga₂O₃(010) films may be attributed to the step-flow growth mode due to the large miscut angle (>2°) of the substrate. The surface morphology and roughness were insensitive to the Ga/O ratio, indicating a large growth window for a smooth surface. From these results, we conclude that slightly Ga-rich conditions between 650 and 750 °C are suitable for obtaining smooth surfaces with a high growth rate.

Zoom In Zoom Out Reset image size

Hall measurements were performed at room temperature on 300-nm-thick Sn-doped β-Ga₂O₃(010) films with various Sn BEPs (7 × 10⁻¹⁰–7 × 10⁻⁹ Torr) using 150 × 150 µm² van der Pauw (vdP) patterns. In PAMBE, the Sn flux could be stably controlled by the cell temperature. The Sn-doped β-Ga₂O₃(010) homoepitaxial films were electrically isolated by 50-nm-thick undoped β-Ga₂O₃(010) films with high resistivity and had no SnO₂ peaks in XRD 2θ–ω scans. An Ohmic contact was achieved by BCl₃ reactive ion etching prior to the fabrication of Ti (20 nm)/Au (200 nm) metal stacks annealed at 500 °C for 1 min in N₂.³⁾ The relation of the Hall carrier concentration and Hall mobility to the Sn BEP is shown in Fig. 6. The carrier concentration and mobility of the Sn-doped β-Ga₂O₃(010) epitaxial films were around 1 × 10¹⁷ cm² and 80 cm² V⁻¹ s⁻¹, respectively. This is comparatively close to the values of Sn-doped β-Ga₂O₃(010) films grown by ozone MBE.¹²⁾

Zoom In Zoom Out Reset image size

In conclusion, we grew homoepitaxial β-Ga₂O₃(010) layers using PAMBE. A high growth rate of 2.2 nm/min was achieved for β-Ga₂O₃(010) by optimizing the O₂ flux, Ga/O ratio, and growth temperature. Under Ga-rich conditions for T < 650 °C, the growth rate decreased with decreasing growth temperature, while for T > 750 °C, the growth rate decreased with increasing growth temperature. Slightly Ga-rich conditions between 650 and 750 °C are optimal for a smooth surface with an rms surface roughness of 0.1 nm at a high growth rate.

Acknowledgments