Prospects for phase engineering of semi-stable Ga₂O₃ semiconductor thin films using mist chemical vapor deposition

Gallium oxide (Ga $2$O $3$⁠) semiconductor materials and devices are doubtless one of the hottest topics today in device-oriented applied-physics research because Ga $2$O $3$ is a new material whose fundamental properties have yet to be well elucidated, and Ga $2$O $3$-based devices are attractive for power electronics and deep ultraviolet optics that require a material, such as $Ga 2 O 3$⁠, with an ultrawide bandgap (⁠ $≈5$ eV).^1,2 Research on Ga $2$O $3$ was triggered by the development of highly crystalline bulk crystals grown by conventional melt-growth methods,^3–5 which is similar to the history of GaAs- or InP-based devices and unlike that of SiC- and GaN-based devices. This has become motive force for the development of Ga $2$O $3$-based devices by homoepitaxial growth.

Ga $2$O $3$ has at least five polymorphous crystal phases [ $α$⁠, $β$⁠, $γ$⁠, $δ$⁠, and $ε$ (or $κ$⁠)], and the monoclinic $β$-gallia structure (⁠ $β$-Ga $2$O $3$⁠) is the most thermodynamically stable phase.^6,7 Therefore, Ga $2$O $3$ bulk crystals are essentially composed of the $β$ phase. At the outset, $β$-Ga $2$O $3$ bulk crystals had low-resistance n-type conductivity and optical transparency in the visible, attracting attention for use as substrates for GaN-based light-emitting diodes.^8–11 

In contrast, the semiconducting properties of Ga $2$O $3$ became evident through the demonstration of transparent field-effect transistors (FETs) with tin-doped polycrystalline Ga $2$O $3$ channels fabricated on sapphire.¹² Later, single-crystalline $β$-Ga $2$O $3$ thin films were grown by homoepitaxy on $β$-Ga $2$O $3$ substrates.^13,14 Although the bandgap of $β$-Ga $2$O $3$ is far larger than ever reported for semiconductor-device materials, a series of publications followed that focused on its semiconducting properties and device applications followed. The earliest achievements include the formation of semi-insulating layers on n-type $β$-Ga $2$O $3$⁠,^15,16 the formation of Schottky and Ohmic contacts on $β$-Ga $2$O $3$⁠,^15,16 highly sensitive above-bandgap photoconductivity,¹⁶ and bandgap tuning with $β$ (Al,Ga) $2$O $3$ alloys.¹⁷ In 2012 and 2013, the promise of $β$-Ga $2$O $3$ as the basis of new power devices was evinced by molecular-beam homoepitaxial growth of high-quality $β$-Ga $2$O $3$⁠,¹⁸ doping-controlled conductivity,¹⁸ metal–semiconductor FETs,¹⁹ conductivity control by ion implantation,²⁰ and metal-oxide-semiconductor FETs (MOSFETs).²¹ In the following years, world-wide research into $β$-Ga $2$O $3$ materials expanded significantly.^22–30 

In addition to $β$-Ga $2$O $3$⁠, other semi-stable phases of $α$⁠, $γ$⁠, $δ$⁠, and $ε$ (or $κ$⁠) are available. We expect $β$-Ga $2$O $3$ layers homoepitaxially grown on $β$-Ga $2$O $3$ to be of the highest quality and to be the best candidates for device applications because bulk substrates of other phases do not exist. However, other phases of Ga $2$O $3$ are expected to have a variety of unique properties, but they must be heteroepitaxially grown on appropriate substrates. Nevertheless, heteroepitaxy is not always a fatal problem for device applications (consider, e.g., GaAs on Si, GaN on sapphire, GaN on SiC, and GaN on Si). Buffer layers and/or epitaxial layer overgrowth (ELO) contributed to overcoming various problems caused by heteroepitaxy, especially dislocation defects. Therefore, exploring Ga $2$O $3$ for semi-stable phases is a worthwhile challenge. Our goal is to fabricate single-crystal layers of a semi-stable phase, overcoming the natural tendency of Ga $2$O $3$ to prefer the $β$ phase. Success in this endeavor should open the way to new science and offer new applications for the attractive material, that is, Ga $2$O $3$⁠.

In 2008, we reported the growth on sapphire of corundum-structured $α$-Ga $2$O $3$⁠, which has the same corundum structure without severe inclusions of other phases.³¹ This was blossomed in 2016 as the demonstration of a low-on-resistance Schottky barrier diode (SBD),³² which is now commercially available. $α$-Ga $2$O $3$ is advantageous because of its low cost, a wide bandgap compared to $β$-Ga $2$O $3$⁠, flexibility of heterojunctions with other materials, and a wide range of bandgap tuning.³³ Later, other semi-stable phases of $γ$³⁴ and $ε$ (⁠ $κ$⁠)^35–37 were demonstrated. The evolution of a variety of semi-stable phases has widened the range of possible materials for a target application and proposes new and novel functions; here, we should explore how to control the phases and determine why a semi-stable phase predominates over the most thermodynamically stable phase.

Bosi et al.³⁸ and Tak et al.³⁹ published review papers on polymorphs of Ga $2$O $3$⁠, emphasizing how they can be tailored in terms of growth methods and growth conditions. Our long-term research on mist chemical vapor deposition (mist CVD; details of this technology are discussed later) shows that mist CVD offers a wide growth window of semi-stable phases of Ga $2$O $3$ by permitting an appropriate choice of the substrate or underlying buffer layers, especially for corundum-structured $α$-Ga $2$O $3$ on sapphire.

In this paper, we briefly review the achievements of research on phase control of Ga $2$O $3$ and emphasize how mist CVD is suitable for this application because the Ga $2$O $3$ layers grown are effectively phase locked with the underlying crystal structure. Focusing on $α$-Ga $2$O $3$ on sapphire, we discuss the growth mechanism and basic properties, giving, as appropriate, perspectives of device applications of semi-stable $α$-Ga $2$O $3$⁠.

Table I summarizes the basic parameters of Ga $2$O $3$ polymorphs, which are schematically illustrated in Fig. 1. Polymorphs other than the $β$ phase have been detected in nanocrystals that rarely form under extreme environmental conditions (i.e., high temperature and high pressure), although recent research has demonstrated the possibility of growing single-phase thin films.

Most investigations have studied the corundum $α$ phase since well-aligned $α$-Ga $2$O $3$ films were grown epitaxially on sapphire substrates, as evinced by an x-ray 0006 diffraction $ω$-scan rocking curve full-width at half maximum as small as 60 arc sec.³¹ The growth of the semi-stable corundum $α$ phase is attributed to the matching of the crystal structure to the sapphire substrate, which has the corundum structure (⁠ $α$-Al $2$O $3$⁠). Previous work shows that the matching of the crystal structure between a substrate and an epilayer is one of the basic requirements for the epitaxial growth of a semi-stable material, as in the case of cubic GaN on GaAs,^43–46 cubic ZnCdS on GaAs,⁴⁷ and wurzite MgZnO on ZnO.⁴⁸ In contrast, on a corundum-structured sapphire substrate, the growth of both $α$ and $β$-Ga $2$O $3$ has been reported, with more publications reporting the growth of $β$-Ga $2$O $3$^49–53 than $α$-Ga $2$O $3$⁠.^32,54–61 Table II lists publications reporting growth methods and conditions for $α$ or $β$ polymorphs for the Ga $2$O $3$ layer that preferably grows on sapphire substrates. $α$-Ga $2$O $3$ is grown at low temperature by mist CVD, halide vapor phase epitaxy (HVPE), and atomic layer deposition (ALD), but not by laser molecular-beam epitaxy (MBE). Actually, Ga $2$O $3$ grown on sapphire by MBE (Table II lists plasma-assisted MBE) or metal organic CVD (MOCVD) tends to form the $β$ phase^49–53 or mixed $α$ and $β$ phases.^62–64 Some papers report the growth of either the $β$ or $ε$ phase by MOCVD depending on the growth temperature or the growth pressure.^65,66 Thin layers of $α$-Ga $2$O $3$ are grown⁶³ under limited growth conditions^62,68 or for specific substrate orientations.⁶⁹ It is, at this stage, difficult to grow homogeneous thick $α$-Ga $2$O $3$ layer on sapphire by MBE and MOCVD.

The next question is how to grow $α$-Ga $2$O $3$ layers on sapphire or what determines the polymorph of Ga $2$O $3$ on sapphire. A hint appears in the papers reporting the variation of the crystal phases of Ga $2$O $3$ in terms of the growth methods, growth conditions, and substrate materials. Schewski et al.⁶³ showed that very thin (three atomic layers) $α$-Ga $2$O $3$ layers were grown pseudomorphically on sapphire, and then the phase was changed to $β$ during the growth processes of MOCVD, MBE, and PLD. Misfit dislocations were detected at the $β$-Ga $2$O $3$– $α$-Ga $2$O $3$ interface but not at the $α$-Ga $2$O $3$–sapphire interface. Akazawa⁷⁰ demonstrated that $α$-Ga $2$O $3$ grows on $a$-plane sapphire at 600  $°$C and undergoes solid-phase crystallization at 800  $°$C, which is higher than the transition temperature from $α$ to $β$⁠,⁶ suggesting stable-phase locking from the corundum-structured substrate. Sun et al.⁶² and Yao et al.⁶⁸ reported the formation of $α$-Ga $2$O $3$ rather than $β$-Ga $2$O $3$ when HCl is in the growth atmosphere. These results suggest that the essential feature for the growth of a semi-stable phase of $α$-Ga $2$O $3$ is not only the use of a corundum-structured substrate but also the growth mechanisms.

As evinced by Table II, the method of mist CVD offers a broad opportunity to form $α$-Ga $2$O $3$ on sapphire substrates. In the remainder of this paper, we discuss the growth mechanism at work in mist CVD suitable for the growth of semi-stable phases of Ga $2$O $3$ due to phase matching with the substrates, not only for $α$-Ga $2$O $3$ but also for other phases.

Mist CVD is a simple, safe, and cost-effective technology for depositing oxide thin films.⁷¹  Figure 2 shows schematically an example of the growth system. The basic idea is to use mist particles containing source materials and formed by ultrasonic atomization of water or alcohol solution instead of organometallic sources as in MOCVD. The mist particles serve as a precursor for CVD. To grow Ga $2$O $3$ by MOCVD, trimethylgallium (TMGa) or triethylgallium (TEGa) is used as a gallium source, together with O $2$ or N $2$O as a oxygen source. With mist CVD, gallium acetylacetonate [Ga(III) 2,4-pentanedionate, Ga(acac) $3$] or gallium chloride (GaCl $3$⁠) may serve as a starting source material. They are dissolved in water with a slight addition of hydrochloric acid (HCl), and the solution is atomized using an ultrasonic transducer. Mist particles formed by atomization are transferred by a carrier gas to a reactive area containing a heated substrate.

This technology constitutes a non-vacuum-based deposition method of ferroelectric thin films⁷² or transparent conductors;^73,74 for the latter, the technology was called “ultrasonic spray pyrolysis.” Applying this technology to epitaxial growth of device-quality single-crystalline semiconductors proved challenging, but the earlier work on ZnO achieved layer-by-layer growth⁷⁵ and reasonable residual electron concentrations (on the order of 10 $17$ cm $− 3$⁠),⁷⁶ which encouraged the development of Ga $2$O $3$ electronics based on this growth technology (which offers phase control).

$α$-Ga $2$O $3$ is a semi-stable phase of Ga $2$O $3$ crystals grown at high pressure (44 kbar) and high temperature (1000  $°$C) with a NaOH flux generated from $β$-Ga $2$O $3$ powders.⁷⁷ Note that $α$-Ga $2$O $3$ thin films are grown on sapphire (⁠ $α$-Al $2$O $3$⁠) substrates at low temperatures and atmospheric pressure by mist CVD, overcoming the large lattice mismatch (4.8%) and thermodynamics. To elucidate the growth mechanism and determine how $α$-Ga $2$O $3$ thin films are realized, we investigate in detail the structure of the $α$-Ga $2$O $3$– $α$-Al $2$O $3$ crystal interface.

Figures 3(a) and 3(b) show cross-sectional transmission electron microscopy (TEM) images of the $α$-Ga $2$O $3$– $α$-Al $2$O $3$ interface observed along the [11 $2 ¯$0] and [1 $1 ¯$00] zone axes, respectively.^33,78 A periodic structure with a period of 8.6 nm appears in Fig. 3(a) at the interface along the [1 $1 ¯$00] axis.

Since the unit crystal cell of $α$-Ga $2$O $3$ and $α$-Al $2$O $3$ in the [1 $1 ¯$00] direction is 0.43 and 0.41 nm long, respectively, the length of 20 crystal cells (8.60 nm) of $α$-Ga $2$O $3$ corresponds to the length of 21 crystal cells (8.61 nm) of $α$-Al $2$O $3$⁠.

Figure 3(b) confirms a similar periodic structure along the [11 $2 ¯$0] axis with a 5.0-nm-long unit crystal cell. The unit crystal cell along the [11 $2 ¯$0] direction of $α$-Ga $2$O $3$ and $α$-Al $2$O $3$ is 0.249 and 0.238 nm, respectively; therefore, 20 crystal cells (4.98 nm) of $α$-Ga $2$O $3$ correspond to 21 crystal cells (4.99 nm) of $α$-Al $2$O $3$⁠. These results suggest that the $α$-Ga $2$O $3$ crystal grows on an $α$-Al $2$O $3$ substrate via semi-coherent growth or domain-matching epitaxy⁷⁹ with 20 unit cells of $α$-Ga $2$O $3$ for 21 crystal cells of $α$-Al $2$O $3$⁠.

Such semi-coherent growth is well known for zinc blende- or diamond-structured crystal systems, such as GaAs/Si,⁸⁰ InAs/GaAs,⁸¹ and Ge/Si.⁸² Furthermore, for corundum-structured crystal systems, a similar periodic structure has been reported for an $α$-Fe $2$O $3$ crystal grown on an $α$-Al $2$O $3$ substrate,⁸³ where the lattice mismatch is about 5%. Therefore, the semi-coherent growth or the lattice-relaxation mechanism of domain-matching epitaxy may also be characteristic of the growth of corundum-structured oxides on sapphire substrates. This is also evinced by x-ray reciprocal-lattice diffraction, which shows that $α$-Ga $2$O $3$ films are almost completely relaxed on $α$-Al $2$O $3$ substrates.³³  Figure 4 shows schematically the structure at the interface.

When using mist CVD, the crystal structure of a thin film grown on a substrate tends to well follow the crystal structure of the substrate, even if the result is a semi-stable structure for the film. Stable-phase bixbyite indium oxide (⁠ $c$-In $2$O $3$⁠) was grown on yttria-stabilized zirconia (YSZ) substrates (cubic crystal).^85–87 Similarly, $c$-In $2$O $3$ was grown on YSZ by mist CVD.⁸⁴ Conversely, indium oxide adopts the semi-stable phase of corundum (⁠ $α$-In $2$O $3$⁠) when deposited on $α$-Fe $2$O $3$⁸⁸ or $α$-Ga $2$O $3$⁸⁹ buffer layers or directly on $α$-Al $2$O $3$ substrates.⁹⁰ In spite of being in a semi-stable phase, the thin films tend to be “phase locked” by the substrates when grown by mist CVD. In Secs. III C and III D, we discuss the mechanism for phase locking and show other semi-stable phases of Ga $2$O $3$ grown by mist CVD.

Figure 5 shows the transitions between crystal structures in the growth of Ga $2$O $3$ on $c$-plane sapphire. Figure 5(a) shows the result of mist CVD, where $α$-Ga $2$O $3$ is grown continuously in spite of being in the semi-stable phase. Figure 5(c) shows the results of Schewski et al. obtained by MOCVD, MBE, and PLD, where $β$-Ga $2$O $3$ is grown on an initially strained $α$-Ga $2$O $3$ layer with the epitaxial relationship $β$-Ga $2$O $3$(⁠ $2 ¯$201)// $α$-Ga $2$O $3$(0001).⁶³ This relationship stems from Ga atoms in the $β$-Ga $2$O $3$(⁠ $2 ¯$201) plane forming a nonplanar hexagonal atomic arrangement, as shown in Fig. 5(b). In this case, the $β$-Ga $2$O $3$ layer only partially references the atomic arrangement during its initial growth. Figure 5(d) shows the general case of $β$-Ga $2$O $3$ grown on $c$-plane sapphire, where the epitaxial relationship is $β$-Ga $2$O $3$(⁠ $2 ¯$01)//sapphire(0001), similar to Fig. 5(c).

To grow the stable phase, such as with MOCVD or other vacuum growth techniques, atoms that have become supersaturated on the substrate surface naturally arrange themselves so that the most stable phase is prioritized over the atomic arrangement of the substrate. Therefore, it is reasonable that the crystal structure changes as the growth proceeds, as shown in Figs. 5(c) and 5(d). In contrast, when using mist CVD, the growth proceeds in the metastable phase, as shown in Fig. 5(a). One reason for this result is that the strong phase-locked mechanism faithfully reproduces the atomic arrangement of the surface despite the relatively low internal and/or migration energy due to the comparatively low growth temperature in mist CVD (see Table II). However, low growth temperature tends to obstruct epitaxial growth and the growth of high-quality crystals. Continuous growth of high-quality semi-stable phases requires phase-locked epitaxial growth, which is based more strongly on the atomic arrangement of the growth surface.

The source molecule used to grow Ga $2$O $3$ via mist CVD is currently Ga(acac) $3$ or gallium halides, such as GaCl $3$⁠, gallium bromide (GaBr $3$⁠), and gallium iodide (GaI $3$⁠). Regardless of the source molecule used for source solutions, Ga ions are basically hydrated and some aqua ligands are replaced by hydroxide ions, acetylacetonate ligands, and/or halide ligands. Many investigations of Ga $2$O $3$ growth by mist CVD added HCl to the source solutions to enhance the solubility of source molecules. However, the HCl additive complicates not only the interpretation of the experimental results but also the discussion of the growth mechanism.

To investigate the mechanism exploited by mist CVD to produce phase-locked epitaxial growth, we prepared a Ga source solution with a molar concentration ratio of $[ Ga 3 +]:[ Cl −]=1:3$⁠.⁹² Such a solution is safely obtained by dissolving a Ga metal in HCl. Most Ga ions in the solution form hexahydrate complexes [Ga(H $2$O) $6 3 +3$Cl $−$].^93,94 Acetylacetonation of Ga is controlled by adding acetylacetone (Hacac). The acetylacetonate (acac) anions coordinate around Ga atoms in their enolate form. The possible ligand coordination structures are Ga $3 +$(H $2$O) $6$⁠, Ga(acac) $2 +$(H $2$O) $4$⁠, Ga(acac) $2 +$(H $2$O) $2$⁠, and Ga(acac) $3$⁠. Ga(acac) $3$ powder has very low solubility in water. Using this technique, acetylacetonated Ga aqueous solutions can be prepared with a definite concentration. Actually, to shorten the ligand-formation time, NH $4$(acac) should be used instead of H(acac).

Figure 6 shows the growth rates of $α$-Ga $2$O $3$ thin films as a function of the acetylacetone concentration.⁹² The growth conditions were $[ Ga 3 +]=0.02$ mol/L, and the growth temperature was 400, 450, or 500  $°$C. The substrates were $c$-sapphire, and $N 2$ served as a process gas. The growth rate is almost zero when using a source solution without acetylacetone and increases with increasing acetylacetone concentration,⁹² with a tendency to saturate at all temperatures. These results imply that hexahydrate Ga ions do not contribute to film formation, whereas acetylacetonated Ga ions promote film formation on $c$-sapphire substrates. The concentration of acetylacetonated Ga ions, such as Ga(acac) $2 +$⁠, Ga(acac) $2 +$⁠, and Ga(acac) $3$⁠, can be calculated using the equilibrium constants already reported: $K n=[ Ga ( acac ) n ( 3 − n ) +]/([ Ga 3 +] [ acac − ] n)$⁠, $log⁡ K 1=9.4$⁠, $log⁡ K 2=17.8$⁠, and $log⁡ K 3=23.7$⁠.⁹⁵ The total-concentration profile of acetylacetonated Ga ions follows a similar trend as the growth rates shown in Fig. 6. Additionally, the root mean square examined using the dynamic force mode of atomic force microscopy (AFM) clearly depends on the growth temperature. Higher growth temperature reduces not only the root mean square but also its dependence on [NH $4$acac].⁹² From the viewpoint of growth mechanisms, these results indicate that the growth reactions occur at the film surface.

The use of acetylacetonate complexes to enhance the formation of oxide films by molecularly designed dispersion has been reported previously.^96–98 A similar enhancement mechanism should occur when $α$-Ga $2$O $3$ is grown by mist CVD. Figure 7 shows a possible mechanism for the growth of $α$-Ga $2$O $3$ by mist CVD. According to Tamba et al., the surface of $α$-Ga $2$O $3$ is terminated by Ga atoms.⁹⁹ Under the conditions used for mist-CVD growth, Lewis acid-base interactions should lead to surface hydroxyl coverage by water molecules. Next, as shown by step 1 in Fig. 7, a gallium acetylacetonate complex approaches the surface and anchors to a surface hydroxyl by hydrogen bonding. Subsequently, in steps 2–4, a Ga–O bond forms by a ligand-exchange mechanism and the ligand leaves as acetylacetone (Hacac).

To obtain experimental support for the proposed mechanism, $α$-Ga $2$O $3$ was grown with the introduction of isotopically labeled water (H $2 18$O) using N $2$ and O $2$ as process gases. By examining the $18$O/ $16$O ratio by secondary-ion mass spectrometry (SIMS), the origin of oxygen atoms in the grown films can be investigated. The resulting $18$O/ $16$O ratio almost equals the ratio of water in the source solution, taking into account the H $2 18$O naturally present in water. These results indicate that the oxygen atoms in the grown films originate from water. This result supports the proposed mechanism for phase-locked epitaxy of $α$-Ga $2$O $3$⁠.

Note that homogeneous Ga $2$O $3$ thin films of other semi-stable phases, such as $γ$- and $ε$-Ga $2$O $3$⁠, have been grown using mist CVD. Cubic-defective-spinel-structured $γ$-Ga $2$O $3$ is attractive due to its possible multifunctionality, for example, by alloying with ferrimagnetic $γ$-Fe $2$O $3$⁠. It is possible to incorporate the magnetic element Fe into Ga $2$O $3$ to produce ferromagnetic materials.¹⁰⁰ Oshima et al. applied mist CVD to grow $γ$-Ga $2$O $3$ on spinel (MgAl $2$O $4$⁠) substrates, which forms a cubic structure with a 2% lattice mismatch with $γ$-Ga $2$O $3$ .³⁴ As shown in Fig. 8, the x-ray diffraction (XRD) patterns are almost completely dominated by the peaks of $γ$-Ga $2$O $3$ except for those from the substrate if the growth temperature was 390 or 400  $°$C. At the higher growth temperature, the most stable $β$ phase is included; the growth window for the $γ$ phase is very narrow. Nevertheless, the quality of the resulting films is sufficient to determine indirect and direct bandgaps of 4.4 and 5.0 eV, respectively, and the refractive index in the visible is 2.0–2.1. The $α$ phase has been grown on sapphire substrates at similar growth temperatures, which means that, under these growth conditions, the crystal structure follows that of the underlying substrate.

$ε$-Ga $2$O $3$ forms a characteristic crystal structure. After Roy et al. suggested the $ε$ phase,⁶ it was determined by numerous investigators to form the hexagonal structure (⁠ $P 6 3mc$⁠).^{35,36,101,102} Later, through critical investigations of the crystal structure, $ε$-Ga $2$O $3$ was suggested to form the orthorhombic (⁠ $Pna 2 1$⁠) structure.⁴¹ Because $κ$-Al $2$O $3$ was already known to form an orthorhombic structure,⁴² some investigators called the Ga $2$O $3$ of orthorhombic structure “ $κ$-Ga $2$O $3$⁠.” Nevertheless, we use the name $ε$-Ga $2$O $3$ below. $ε$-Ga $2$O $3$ is known to spontaneously polarize,^102,103 which is preferred over ferroelectric devices and high-frequency transistors with a two-dimensional electron gas (2DEG).

Oshima et al. used HVPE to grow $ε$-Ga $2$O $3$ on GaN, AlN, and $β$-Ga $2$O $3$⁠.³⁵ Mist CVD was used to grow $ε$-Ga $2$O $3$ on cubic (111) MgO at 400–700  $°$C and on (111) YSZ at 600–700  $°$C.³⁷ The indirect and direct bandgap energies are approximately 4.5 and 5.0 eV, respectively. Later, Oshima et al. used mist CVD to grow $ε$-Ga $2$O $3$ on hexagonal (0001) AlN,¹⁰⁴ (0001) GaN,¹⁰⁵ and cubic (111) SrTiO $3$⁠.¹⁰⁵ When grown on cubic-structured substrates, it is reasonable that Ga $2$O $3$ preferably forms an orthorhombic structure because of the similarity of the crystal structures. Conversely, it is interesting to discuss why Ga $2$O $3$ forms an orthorhombic structure when grown on a hexagonal-structured substrate. Nishinaka et al.¹⁰⁵ showed that unit cells of $ε$-Ga $2$O $3$ are arranged so that the atomic plane of each unit cell of $ε$-Ga $2$O $3$ follows an atomic plane of the hexagonal substrate. In other words, the growth of $ε$-Ga $2$O $3$ on hexagonal substrates results from Ga $2$O $3$ nucleating by following the crystal structure of the substrate. The surface-stabilization mechanism may be the same as that reported for $ε$-Ga $2$O $3$ grown by MOCVD.¹⁰⁶ The ferroelectric properties of $ε$-Ga $2$O $3$⁠,¹⁰⁷ the growth of $ε$-(Al $x$Ga $1 − x$⁠) $2$O $3$⁠,¹⁰⁸ and the type-I band lineup of their heterointerface¹⁰⁸ are evinced in Fig. 9. These results have promoted research into 2DEG transistors.

Mist CVD is a useful technology for growing semi-stable Ga $2$O $3$⁠. When $α$-Ga $2$O $3$ was grown by mist CVD on sapphire in 2008, no other growth technology could duplicate the result. However, the knowledge that $α$-Ga $2$O $3$ is one of the key materials for new device applications has renewed efforts to use other technology to grow $α$-Ga $2$O $3$ . This is also true for other polymorphs. Not surprisingly, the most-stable $β$-Ga $2$O $3$ is grown on a variety of substrates irrespective of the growth methods and conditions, although many researchers have searched for growth windows to grow semi-stable Ga $2$O $3$⁠.

HVPE can be used to grow high-purity $α$-Ga $2$O $3$ and $ε$-Ga $2$O $3$ at high growth rates. The growth of phase-pure $α$-Ga $2$O $3$ on $c$-plane sapphire at relatively low temperatures (typically 500–600  $°$C) has been reported by several groups.^{58,59,109–111} The growth rate increases with increasing concentration of precursors, reaching over 100  $μ$m/h while maintaining a specular surface.¹¹⁰ The structural quality of HVPE-grown $α$-Ga $2$O $3$ is similar to that of mist-CVD-grown films, but the tilt angle tends to be larger (⁠ $≈100$ arc sec), probably because of the nonoptimized nucleation process.⁵⁸ The dislocation density measured by plan-view TEM is typically $≈ 10 10$ cm $− 2$⁠. Epitaxial lateral overgrown $α$-Ga $2$O $3$ has also been grown via HVPE.¹¹¹ 

Oshima et al. grew (001)-oriented $ε$-Ga $2$O $3$ on (0001) GaN and (0001) AlN at a growth temperature of about 550  $°$C for both polymorphs, finding that the use of $c$-plane GaN and AlN substrates, which have hexagonal space groups, is vitally important to obtain $ε$-Ga $2$O $3$ epitaxial layers.³⁵ This is similar to the case of mist CVD, as suggested by Nishinaka et al.¹⁰⁵ Yao et al. grew (001) $ε$-Ga $2$O $3$ on $c$-plane sapphire at 650  $°$C, which succeeded likely because of the increased Ar flow rate or increased film coverage.⁶⁸ In the study, high-resolution TEM imagery revealed the presence of a 10-nm-thick $α$-Ga $2$O $3$ interlayer between the $ε$-Ga $2$O $3$ and the substrate, with the epitaxial relationship of [ $1 ¯$100] $ε$-Ga $2$O $3$//[11 $2 ¯$0] $α$-Ga $2$O $3$//[11 $2 ¯$0] sapphire.

$α$-Ga $2$O $3$ heteroepitaxial layers grown by MOCVD tend to be in mixed-phase form. In fact, $α$-Ga $2$O $3$ sometimes forms a thin interlayer between the substrate and an overlayer of a different phase. For example, Schewski et al. observed three monolayers of pseudomorphic $α$-Ga $2$O $3$ between $β$-Ga $2$O $3$ and a (0001) sapphire substrate for three different growth techniques, namely, MOCVD, PLD, and MBE.⁶³ 

Sue et al. grew three different phases of Ga $2$O $3$ (⁠ $α$⁠, $β$⁠, and $ε$⁠) films on $c$-plane sapphire using different flow rates of HCl in MOCVD.⁶² The results show a threefold increase in the growth rate of pure-phase $β$-Ga $2$O $3$ for the HCl flow rate of 5 sccm. Upon increasing the HCl flow rate to 10 sccm, the film transitions from pure-phase $β$-Ga $2$O $3$ to a mixture of $β$- and $ε$-Ga $2$O $3$⁠. At a HCl flow rate of 30 sccm, which yields the highest growth rate of $≈$1 $μ$m/h, only $ε$-Ga $2$O $3$ forms. Furthermore, increasing the HCl flow rate produces a mixture of $α$ and $ε$-Ga $2$O $3$⁠. A combined growth and XRD study by Mezzadri et al. using water and trimethylgallium as reagents and palladium-purified H $2$ as a carrier gas revealed the growth of $ε$-Ga $2$O $3$ (001) on $c$-plane sapphire with the Ga $2$O $3$ [10 $1 ¯$0] direction parallel to the sapphire direction [11 $2 ¯$0], yielding a lattice mismatch of about 4.1%.¹⁰² 

To the best of our knowledge, no reports exist of pure $γ$-Ga $2$O $3$ films grown by MOCVD. A very thin (10–50 nm) $γ$-Ga $2$O $3$ nucleated layer was identified at the interface between an $ε$-Ga $2$O $3$ layer, grown by MOCVD, and a sapphire substrate, although the mechanism for the formation of this layer was not discussed.⁴¹ However, Bhuiyan et al. revealed the growth of pure $γ$-(Al $x$Ga $1 − x$⁠) $2$O $3$ on (010) $β$-Ga $2$O $3$ substrates with $0.37<x<1$⁠, which is likely due to increased strain due to Al incorporation inducing the rotation of $β$-phase (Al,Ga) $2$O $3$ domains, thereby promoting the formation of $γ$-phase (Al,Ga) $2$O $3$⁠.¹¹³ 

In addition to the thermodynamically stable $β$ phase, the $α$⁠, $γ$⁠, and $ε$ phases have been stabilized using MBE with heteroepitaxy. The MBE growth on (⁠ $c$-plane) sapphire (0001) frequently results in the formation of $β$-Ga $2$O $3$ (⁠ $2 ¯$01) layers with rotational domains.^52,114,115 However, a close inspection of the interface with the substrate by Schewski et al. revealed the presence of an approximately 1-nm-thick pseudomorphic $α$-Ga $2$O $3$ (0001) interfacial layer that is likely stabilized by lattice-mismatch-induced strain.⁶³ Sapphire substrates of other orientations enable the formation of thicker $α$-Ga $2$O $3$ layers before the growth of $β$-Ga $2$O $3$ proceeds on top: a combined growth and in situ XRD study by Cheng et al. demonstrated the growth of a 143-nm-thick $α$-Ga $2$O $3$ (11 $2 ¯$0) layer on $a$-plane sapphire (11 $2 ¯$0) that relaxed the in-plane stain within the first 2 nm of the growth.⁶⁴ Kracht et al. grew greater than 200-nm-thick $α$-Ga $2$O $3$ (10 $1 ¯$2) layers on (⁠ $r$-plane) sapphire (10 $2 ¯$0) before the growth of $β$-Ga $2$O $3$ (⁠ $2 ¯$01) nucleated on the (0001)-facets that emerged by roughening of the $α$-Ga $2$O $3$ (10 $1 ¯$2) layer.¹¹⁶ Oshima et al. grew $α$-Ga $2$O $3$/ $α$-Al $2$O $3$ superlattices on $r$-plane sapphire, achieving a coherent interface without misfit dislocations for up to $≈1$-nm-thick $α$-Ga $2$O $3$ (10 $1 ¯$2) layers.¹¹⁷ A combined growth and scanning transmission electron microscopy study by Jinno et al.⁶⁹ revealed the growth of 51-nm-thick $α$-Ga $2$O $3$ (10 $1 ¯$0) layers on $m$-plane sapphire (10 $1 ¯$0) that relaxes at the $α$-Ga $2$O $3$–sapphire interface. These results indicate that the $c$-plane facet enhances the growth of $β$-Ga $2$O $3$ in MBE on a sapphire crystal plane perpendicular to the $c$-plane, such as the $a$- or $m$-planes, and allows the growth of phase-pure $α$-Ga $2$O $3$ by avoiding facets.

Another example of substrate-stabilized phase formation is that of the cubic-defective-spinel $γ$-Ga $2$O $3$ (001) on spinel MgAl $2$O $4$ (001) substrates.¹¹⁸ 

Vogt et al. and Kracht et al. reported the growth of $ε$-Ga $2$O $3$ (001) on $c$-plane sapphire by metal-exchanged catalysis mediated through an additional In flux⁷³ and Sn flux,¹¹⁶ respectively. Bi et al. reported the stabilized growth of $ε$-Ga $2$O $3$ (001) on $c$-plane sapphire by Mg doping.¹¹⁹ It is conceivable that the conditions for Ga $2$O $3$ growth catalyzed by In or Sn or Mg at high growth temperature thermodynamically promote the formation of the $ε$ phase. This hypothesis is supported by the relatively small difference in formation energy between the $β$ and $ε$ phases, which decreases with increasing temperature.¹²⁰ 

ALD has been used for heteroepitaxial growth of $α$- and $ε$-Ga $2$O $3$ on $c$-plane sapphire.

Boschi et al. grew $ε$-Ga $2$O $3$ (001) on $c$-plane sapphire at 550  $°$C, obtaining uniform thickness over large surfaces.¹¹² Roberts et al. grew $α$-Ga $2$O $3$ (0001) on $c$-plane sapphire by low-temperature ALD and revealed that the film consists of (0001)-oriented $α$-Ga $2$O $3$ columns originating from the substrate.¹²¹ A TEM study showed some inclusions, most likely of the $ε$ phase and primarily at the tip of the columns, with an amorphous phase located near the surface of the film and between the $α$-Ga $2$O $3$ columns.

Wheeler et al. demonstrated the phase selectivity of $β$-, $α$-, and $ε$-Ga $2$O $3$ on $c$-plane sapphire as functions of the plasma gas composition, the gas flow and pressure during the plasma pulse, and the growth temperature.⁶⁰ Pure $α$-Ga $2$O $3$ films with an abrupt interface were produced at low temperatures and pressures, whereas high-quality, single-phase $β$-Ga $2$O $3$ films were directly deposited on $c$-plane sapphire at low pressures with an Ar-O $2$ plasma gas composition. In contrast, no single-phase $ε$-Ga $2$O $3$ heteroepitaxial films were obtained. $ε$-phase films were predominately obtained without nucleation or buffer layers by going to higher pressures and temperatures. Carbon, or other impurities, are hypothesized to help stabilize the $β$-Ga $2$O $3$ phase. Thus, to realize pure semi-stable phases, particular attention should be focused on in situ and ex situ cleaning of the sapphire substrate prior to deposition and on the initiation of growth with the metal organic precursor because ALD is conducted below the desorption temperature of most carbon species.

Table III shows examples of growth conditions used to grow semi-stable Ga $2$O $3$ by different growth methods and on different substrates. When using mist CVD, the crystal structure of the Ga $2$O $3$ epilayer is that of the underlying layer; that is, the crystal structure of the Ga $2$O $3$ epilayer is basically determined by the structure of the underlying layer, in spite of the lattice mismatch, rather than by the growth conditions. In other words, mist CVD introduces a strong tendency of phase-locking epitaxy, as discussed in Sec. III B.

Note that thick $α$-Ga $2$O $3$ was grown by HVPE for which GaCl and HCl exist in the growth atmosphere. Ga–Cl complexes and HCl are also expected in the atmosphere of mist CVD (HCl markedly influences the Ga $2$O $3$ phase in MOCVD).⁶² The similarity of the precursors as well as the existence of HCl in both mist CVD and HVPE may stabilize the growth of $α$-Ga $2$O $3$⁠. Another important factor is the low growth temperature.

Conversely, when using other growth methods, one of the common features is the growth of very thin $α$-Ga $2$O $3$ directly on $α$-Al $2$O $3$⁠. This is due to the nature of phase locking, which can dominate the very initial stage of growth on $α$-Al $2$O $3$⁠, irrespective of the growth method. However, as the growth continues, Ga $2$O $3$ seems to voluntarily nucleate to form stable $β$-Ga $2$O $3$ rather than to follow the atomic arrangement of the underlying layer. This conclusion is supported by high kinetic or thermal energy of the precursors and by the high substrate temperature.

Since $α$-Ga $2$O $3$ is grown by heteroepitaxy on $α$-Al $2$O $3$⁠, dislocation-originated defects form in the $α$-Ga $2$O $3$ thin films. Figure 10(a) shows a surface TEM image of an $α$-Ga $2$O $3$ thin film. The in-plane contrasting density is not homogeneous. By applying an inverse fast Fourier transform method to the image, clear dislocation lines (identified by red arrows) appear, as shown in Figs. 10(b) and 10(c), selecting the 11 $2 ¯$0 and 30 $3 ¯$0 diffraction spots, respectively. These dislocations are attributed to domain boundaries because inclusion of rotational domains has never been reported.³¹ 

For $α$-Ga $2$O $3$ thin films, the screw dislocations are rare in cross-sectional TEM images, whereas the edge dislocation density is calculated to be $7× 10 10$ cm $− 2$ by applying Ham’s method to the images.⁷⁸ Efforts are continuing to reduce the dislocation density. For example, Jinno et al. reduced the edge dislocation density to $6× 10 8$ cm $− 2$ by introducing quasi-graded $α$-(Al,Ga) $2$O $3$ buffer layers.¹²² ELO produced $α$-Ga $2$O $3$ films on the mask region without noticeable dislocation, as per TEM images for HVPE¹¹¹ and mist CVD.¹²³ Kawara et al. achieved the low dislocation density of less than $5× 10 6$ cm $− 2$ with the use of double-layered ELO and HVPE.¹²⁴ 

Unintentionally doped $α$-Ga $2$O $3$ has a very high resistivity, and n-type conductivity is obtained by doping with tin (Sn)^125,126 or silicon (Si)¹²⁷ in mist CVD. The carrier concentration was controlled in the range of 10 $17$–10 $19$ cm $− 3$⁠. However, the mobility was less than 24 cm $2$ V $− 1$ s $− 1$⁠, suggesting that crystal defects, such as dislocations, severely restrict the mobility. Improvement of the crystal quality is important to obtain better electrical properties.

Nevertheless, SBDs with low on-resistance and high breakdown voltage (e.g., 0.1 m $Ω$ cm $2$ and 531 V or 0.4 m $Ω$ cm $2$ and 855 V, respectively) were grown by FLOSFIA, Inc. using mist CVD.³² This achievement was followed by the fabrication of a normally off MOSFET with a maximum channel mobility of 72 cm $2$ V $− 1$ s $− 1$⁠.¹²⁸ Mist CVD is a simple, safe, and cost-effective technology that can produce low residual impurities and specular surface mobility,³² promising high-performance devices at low cost. Maeda et al. showed that the $α$-Ga $2$O $3$ SBDs had nearly ideal current–voltage characteristics in spite of the inclusion of numerous dislocation defects,¹²⁹ which is good news for the future evolution of $α$-Ga $2$O $3$ devices and encourages us to elucidate the structure and electronic characteristics of the defects.

An almost-fatal problem in wide-bandgap oxide semiconductors is the difficulty to realize both n- and p-type layers from the same material. However, corundum-structured $α$-Ga $2$O $3$ is expected to form p–n junctions with other corundum-structured p-type oxides. One of the p-type oxide candidates is $α$-Ir $2$O $3$⁠, whose p-type conductivity is attributed to the Ir $5d$ orbital, which supplies holes that are located above the O $2p$ orbital. Growth of p-type $α$-Ir $2$O $3$ and the rectification characteristics of a p–n junction made of $α$-Ir $2$O $3$ and $α$-Ga $2$O $3$ have been demonstrated, and it was shown that the lattice mismatches between $α$-Ir $2$O $3$ and $α$-Ga $2$O $3$ are as small as 0.3% and 0.6% along the $a$ and $c$ axes, respectively.¹³⁰ By alloying with Ga $2$O $3$⁠, the bandgap is enlarged and the lattice parameters approach those of Ga $2$O $3$⁠; these characteristics are preferable for p–n junctions. We fabricated p-type $α$-(Ir,Ga) $2$O $3$ films with bandgaps up to 4.3 eV and p–n junctions from n-type Ga $2$O $3$⁠.¹³¹ 

Corundum-structured $α$-In $2$O $3$ is also a semi-stable phase but was grown on sapphire substrates by mist CVD.^90,132 Given that $α$-Al $2$O $3$ is a stable phase, the corundum-structured alloy semiconductors with $α$-Al $2$O $3$⁠, $α$-Ga $2$O $3$⁠, or $α$-In $2$O $3$ [ $α$-(Al,Ga,In) $2$O $3$] enable bandgaps to be engineered over a wide range of energy. Figure 11 shows bandgap engineering by $α$-(Al,Ga) $2$O $3$ and $α$-(In,Ga) $2$O $3$ alloys from 3.7 to about 9 eV.¹³³ Note that the growth of single-phase $β$-(Al $x$Ga $1 − x$⁠) $2$O $3$ is very limited for high-Al composition (i.e., large $x$⁠) because the monoclinic phase is not a stable structure for Al $2$O $3$⁠. Conversely, a variety of heterostructures are realized by $α$-(Al $x$Ga $1 − x$⁠) $2$O $3$ with a wide range of $x$⁠. The band lineup of the $α$-(Al $x$Ga $1 − x$⁠) $2$O $3$/Ga $2$O $3$ heterostructure is shown to be type I by x-ray photoemission spectroscopy,¹³⁴ which is favorable for applications involving heterostructure transistors and multiple quantum wells.