β-Ga2O3 material properties, growth technologies, and devices: a review

Rapid progress in β-gallium oxide (β-Ga2O3) material and device technologies has been made in this decade, and its superior material properties based on the very large bandgap of over 4.5 eV have been attracting much attention. β-Ga2O3 appears particularly promising for power switching device applications because of its extremely large breakdown electric field and availability of large-diameter, high-quality wafers manufactured from melt-grown bulk single crystals. In this review, after introducing material properties of β-Ga2O3 that are important for electronic devices, current status of bulk melt growth, epitaxial thin-film growth, and device processing technologies are introduced. Then, state-of-the-art β-Ga2O3 Schottky barrier diodes and field-effect transistors are discussed, mainly focusing on development results of the author’s group.


Introduction
There is no room for doubt that silicon (Si) has been the fundamental material serving as the backbone technology of semiconductor device electronics. Generally speaking, up to now, the vast majority of electronic systems are based on Si devices such as metal-oxide-semiconductor field-effect transistors (MOSFETs), bipolar transistors, and some types of diodes. Their performance has tremendously improved by means of uninterrupted innovation in its material and device processing technologies for over the last 70 years. However, performance of Si devices has almost reached the fundamental upper limit expected from its material properties, and it is difficult to keep improving the performance at the same pace as before. In this current situation, research and development (R&D) of FETs and diodes exploiting new semiconductors, which have better physical properties for specific applications than those of Si, has been getting more active from expectations that they may open up a new field of semiconductor electronics and/or elicit characteristics that far exceed those of Si devices. Most of semiconductor physical properties are determined by their bandgap; therefore, bandgap engineering has been one of the main directions for exploring new materials, and a new semiconductor should have a different bandgap energy (E g ) from that of any existing material. Among a variety of new materials, wide bandgap (WBG) semiconductors are expected to be promising for power switching applications, since the WBG devices can offer advantages of high efficiency and power density over Si devices. The two most promising WBG semiconductors are silicon carbide (SiC) and gallium nitride (GaN); their FETs and diodes have already penetrated into the market.
Recently, a new category called "ultrawide bandgap (UWBG) semiconductor" has attracted great attention due to their strong potential providing new and exciting research opportunities for various applications, and R&D on UWBG materials and devices continues to expand over the years [1]. The UWBG semiconductors are defined by E g exceeding those of SiC (3.3 eV) and GaN (3.4 eV). High Al-content AlGaN, diamond, and cubic boron nitride are included in the UWBG materials. A new semiconductor gallium oxide (Ga 2 O 3 ), which is the main topic of this article, is a key material among the UWBG semiconductors. Ga 2 O 3 can offer two fundamental advantages over SiC and GaN: a very large E g and ease of bulk wafer production. Ga 2 O 3 is by no means a novel material; it has a long R&D history for over 70 years [2]. However, Ga 2 O 3 had been largely ignored by a majority of semiconductor researchers and engineers, which resulted in it falling behind SiC and GaN. The achievement of the first singlecrystal Ga 2 O 3 FET in 2011 is widely credited with changing the situation and invigorating Ga 2 O 3 R&D [3]. As shown in Fig. 1, the number of publications on Ga 2 O 3 significantly increases in recent years according to increased recognition among semiconductor researchers and engineers that Ga 2 O 3 has unique and attractive material properties for various optoelectronics applications.
This article provides a comprehensive review on current β-Ga 2 O 3 material and device technologies. It should be noted that although several polymorphs exist for Ga 2 O 3 as introduced in Section 2.1, discussions in this review will just focus on β-Ga 2 O 3 . Following this introduction in Section 1, Section 2 describes basic physical properties of β-Ga 2 O 3 . Sections 3 and 4 provide melt bulk and epitaxial thin-film growth technologies for β-Ga 2 O 3 , respectively. Selected elemental device processing technologies are given in Section 5. Development history of β-Ga 2 O 3 Schottky barrier diodes (SBDs) and FETs in the author's group are introduced in Sections 6 and 7, respectively, together with representative achievements reported from other institutes. Section 8 concludes this article.

Physical properties
The important material parameters of β-Ga 2 O 3 for electronic device applications are listed in Table 1, together with those of Si, SiC, and GaN. The material parameters and the other important features are discussed in this section.

Polymorph
Five types of Ga 2 O 3 polymorphs denoted by α, β, γ , δ, and were first reported in 1952 [2]. The β phase is a thermodynamically stable form, and the others are metastable. β-Ga 2 O 3 has the β-gallia monoclinic structure with lattice constants of 12.2, 3.0, and 5.8 Å in the a, b, and c axes, respectively, and the angle between the a and c axes is about 104 • [4]. A schematic of the β-Ga 2 O 3 unit cell is depicted in Fig. 2. The β-Ga 2 O 3 crystal structure is composed of two inequivalent Ga sites and three inequivalent O sites. Ga(I) and Ga(II) are tetrahedrally and octahedrally coordinated with O, respectively. O(I) and O(II) have threefold Ga coordination, while O(III) has fourfold. The important feature of β-Ga 2 O 3 in contrast to the other polymorphs is that bulk single crystals can be synthesized by melt growth methods; the details will be given in Section 3.

Band structure
β-Ga 2 O 3 has an isotropic and dispersive conduction band minimum comprised of the Ga 4s states, and an anisotropic and fairly flat valence band maximum made up of the O 2p states with contributions of the Ga 3d and 4s orbitals [5][6][7]. These band structures provide sufficient electron conduction associated with a reasonable electron effective mass of ∼0.3m 0 (m 0 : free electron mass)  [5][6][7][8] and limited hole conduction due to three factors described in Section 2.6.2.
First-principles calculations based on hybrid density functional theory predicted an E g of 4.6-4.9 eV for β-Ga 2 O 3 [5][6][7]. For a long term, there has also been intense discussion on whether β-Ga 2 O 3 is a direct or an indirect semiconductor. As an answer to this scientific question, Onuma et al. reported that energies of absorption edges in β-Ga 2 O 3 melt-grown bulk single crystals can be divided into six ranges within 4.4-5.3 eV according to the selection rules of optical transition caused by the anisotropic valence band maximum and that the direct and indirect E g were estimated to be 4.48 and 4.43 eV, respectively [9]. In other words, these experimental results indicate that the energy difference between the direct and indirect E g of β-Ga 2 O 3 is negligibly small at room temperature.

Breakdown electric field
The breakdown electric field (E br ) is an intrinsic property that each material has and corresponds to the maximum electric field that the material structure can support before avalanche breakdown. The larger E br leads to greater voltage blocking capability of devices, that is, a larger breakdown voltage (V br ). Theoretical calculation predicted that β-Ga 2 O 3 possesses an extremely large E br of over 8 MV/cm [10]. However, in fact, the breakdown events for all of the Ga 2 O 3 FETs and diodes reported so far were caused by their permanent failure due to electric field concentration at a gate or an anode electrode edge; therefore, the intrinsic E br of β-Ga 2 O 3 in terms of the avalanche breakdown has never been observed. For reference, the maximum E br experimentally derived from structural failure of a high-k dielectric/n-Ga 2 O 3 diode was ∼7 MV/cm [11].

Electron mobility
The primitive cell of β-Ga 2 O 3 gives rise to many phonon modes due to a lack of symmetry for its monoclinic structure. Among the various modes, the longitudinal optical (LO) phonon with energies of 20-40 meV has the highest impact on low-field electron mobility (μ) in β-Ga 2 O 3 bulks [12]. This is the main factor why the room temperature μ of n-Ga 2 O 3 is theoretically expected to be limited up to ∼200 cm 2 /Vs [13,14], even though the electron effective mass is comparable to those of the other WBG semiconductors. Actually, the peak room temperature and low temperature μ of n-Ga 2 O 3 films experimentally measured were 150-180 and 5000-10,000 cm 2 /Vs, respectively [15][16][17][18].

Saturation electron velocity
The high-field carrier transport is another important factor having a large impact on electronic device performance. The saturation electron velocity (v sat ) in β-Ga 2 O 3 was theoretically predicted to be 1-2×10 7 cm/s [19], which is equivalent to that in Si and about half of those in SiC and GaN. It should be noted that the v sat is sufficiently good for many of high-frequency applications.

n-type
Si, germanium (Ge), and tin (Sn) are often used as donor dopants for β-Ga 2 O 3 . These group IV elements form shallow donor states in Ga 2 O 3 [20,21], and their activation ratios at room temperature are very high (∼100% for Si). The electron density (n) in β-Ga 2 O 3 can be precisely controlled in a wide range of 10 15 -10 20 cm −3 by using the donor dopants. As for controllability of n-type conductivity, Ga 2 O 3 is absolutely the same as other semiconductors.

p-type
Contrary to the ease of n-type doping, a major drawback of Ga 2 O 3 has been a lack of hole-conductive p-type material, which is the most serious limitation for development of Ga 2 O 3 devices. In fact, there has been no report on successful p-type doping with effective hole conduction, even though there are some potential acceptor dopants: magnesium (Mg), zinc (Zn), and beryllium (Be) as cation substituting dopants and nitrogen (N) as an anion substituting one. There are three factors to make it almost impossible to realize hole-conductive p-type Ga 2 O 3 . First, it is generally difficult for single-crystal oxide semiconductors to form shallow acceptor states, since their valence band states are mainly composed of the weakly interacting O 2p orbitals as discussed in Section 2.2. Theoretical calculations predicted that all the acceptor candidates would exhibit extremely large activation energies of over 1 eV [22,23]. Another expected factor limiting hole conductivity is the very low μ and diffusion constant owing to the heavy hole effective mass originated from the flat valence band maximum [5][6][7]. Furthermore, it was also theoretically predicted that holes localize in β-Ga 2 O 3 as small polarons due to lattice distortion, as opposed to being free holes widely populating [24][25][26]. From the three factors, hole-conductive p-Ga 2 O 3 seems to be hardly realized. However, p-Ga 2 O 3 layers formed by the deep-acceptor doping are useful to form a large energy barrier of over 3 eV in n-Ga 2 O 3 by utilizing the built-in potential at the p-n junction.

Thermal conductivity
The thermal conductivity is one of the important parameters to make a great impact on performance of highpower devices, since it limits dissipation capacity of heat generated by on-state conduction loss in the channel and/or drift regions. Ga 2 O 3 has a significantly lower thermal conductivity compared to those of the other power semiconductors as listed in Table 1 [27][28][29][30]. The thermal conductivity of β-Ga 2 O 3 is anisotropic due to the monoclinic lattice structure; they were estimated to be 0.22-0.27 W/cmK in the [010] direction (highest) and 0.11-0.14 W/cmK in the [100] direction (lowest). The low thermal conductivity will not only hamper high-power device performance but also limit long-term device reliability. Hence, thermal management is one of the key technical challenges for future practical applications and industrialization of Ga 2 O 3 RF and power devices.

Device figure of merit
There are several device figures of merit (FOMs) to estimate how suitable a semiconductor material is for specific device application. Here, two FOMs are discussed: Baliga's FOM (BFOM) [31,32] and Johnson's FOM (JFOM) [33].
BFOM is commonly used as a parameter to estimate potential of semiconductor materials for power switching device applications. The BFOM of β-Ga 2 O 3 is several times larger than those of SiC and GaN. This is because BFOM is proportional to the cube of E br but only linearly proportional to μ. Therefore, we can expect from the BFOM that β-Ga 2 O 3 power devices can offer lower conduction loss compared with the other semiconductor devices listed in Table 1 when designing the structure for a given voltage.
We use JFOM to estimate material potential for RF power FET applications. The JFOM of β-Ga 2 O 3 is estimated to be the same level as or a little inferior to that of GaN. However, the actual power efficiency of FETs used in RF amplifiers is generally ∼50%, that is, the other half loss converts to heat. Thus, it is expected that the poor heat dissipation capacity of β-Ga 2 O 3 attributed to its low thermal conductivity becomes much more serious for applications to RF devices rather than power switching devices.

Melt bulk growth
The availability of affordable native wafers manufactured from melt-grown bulk single crystals is one of the distinctive features for β-Ga 2 O 3 and offers a significant advantage of Ga 2 O 3 over SiC and GaN in terms of wafer size, crystal quality, and production cost, because SiC and GaN bulk crystals require alternative synthesis techniques using higher pressure and temperature. β-Ga 2 O 3 bulk single crystals have been synthesized by various melt growth methods illustrated in Fig. 3, such as floating zone [34,35], Czochralski (CZ) [36,37], vertical Bridgman [38], and edge-defined film-fed growth (EFG) [35,39]. Si or Sn is usually used as a donor dopant to control n, and semi-insulating Ga 2 O 3 bulks are producible through compensation of residual donors with Fe or Mg doping. Now, n-type and semi-insulating β-Ga 2 O 3 wafers manufactured from CZ and EFG bulks are made commercially available to the public, which are of sufficiently high quality for development of various Ga 2 O 3 devices.

Molecular beam epitaxy
Molecular beam epitaxy (MBE) is an epitaxial growth technique conducted in ultrahigh vacuum and enables precise control of layer thickness. As other compound semiconductors, MBE has been the most commonly used for thin-film growth of Ga 2 O 3 in the early days of R&D. Figure 4 depicts a schematic of a typical MBE machine for Ga 2 O 3 growth. An effusion cell is used to sublimate a Ga metal source in order to form molecular beam flux that travels in a line of sight toward a substrate. Ga 2 O 3 epitaxial growth happens on the substrate by means of oxidization of supplied Ga atoms using oxidant such as ozone (O 3 ) and oxygen (O) radicals. Two-step reaction happens between Ga and O during Ga 2 O 3 MBE growth [40]. In the first stage of growth, volatile gallium suboxides (Ga 2 O) are formed by oxidization of supplied Ga atoms on the growth surface. Then, in the second stage, Ga 2 O desorbs from the growth surface or is incorporated into a Ga 2 O 3 epitaxial layer by becoming Ga 2 O 3 through a further oxidization process. Due to this unique growth kinetics, the growth rate strongly depends on growth parameters such as O/Ga flux ratio, growth temperature, and substrate orientation, since it is determined by competition between the Ga 2 O desorption and oxidization processes. Intentional n-type doping has been demonstrated by using Si, Sn, and Ge as donor dopants [41][42][43][44]. However, it is difficult to control a low doping density on the order of 10 16 cm −3 or less, because the surface of the dopant source in an effusion cell is easily oxidized during the MBE growth by background O species, resulting in the dopant flux hardly controlled. Furthermore, there is a specific issue only for Sn atoms that they tend to segregate on the growth surface without being incorporated into the epitaxial film [45]. At present, MBE-grown β-Ga 2 O 3 epitaxial wafers are mainly utilized for fabrication of lateral FETs, and both of the MBE methods using O 3 and O radicals can provide high-quality β-Ga 2 O 3 films sufficient for the device development.

Halide vapor phase epitaxy
High-speed β-Ga 2 O 3 thin-film growth attained by halide vapor phase epitaxy (HVPE) was first reported in 2014 [46]. The crystal quality of the HVPE-grown β-Ga 2 O 3 films was excellent from the beginning. Figure 5 illustrates a schematic of an HVPE reactor for Ga 2 O 3 growth. Gallium monochloride (GaCl) and O 2 are typically used as Ga and O precursors, respectively. The growth rate of β-Ga 2 O 3 thin films can be increased up to 20 μm/h without any degradation of crystal quality. The background n of unintentionally doped (UID) β-Ga 2 O 3 (001) films is very low, which is less than 1×10 13 cm −3 [47]. Si doping performed by simultaneous supply of SiCl 4 during the Ga 2 O 3 growth can provide accurate control of n in a wide range from 10 15 to 10 19 cm −3 [15]. The maximum electron μ in HVPE-grown Si-doped n-Ga 2 O 3 films is ∼150 cm 2 /Vs at room temperature reported so far [15]. In recent years, epitaxial wafers with HVPE-grown n-Ga 2 O 3 layers are mass produced and widely utilized for development of vertical FETs and SBDs around the world.

Metalorganic chemical vapor deposition
Metalorganic chemical vapor deposition (MOCVD) is a well-established epitaxial growth technique for compound semiconductors and is often employed for production of high-quality epitaxial wafers on an industrial scale. There used to be some limitations for Ga 2 O 3 MOCVD growth, such as low growth rate, high-density background n, and low electron μ. However, most of the technical issues have been resolved within the last 5 years, and MOCVD can provide high-quality β-Ga 2 O 3 films required for device development now. In a Ga 2 O 3 MOCVD reactor as shown in Fig. 6, traditional Ga precursors such as trimethylgallium (TMGa) and triethylgallium (TEGa), and high-purity O 2 and H 2 O are generally used for the growth. Si donor doping can be done by simultaneous supply of Si-containing precursors.

Ion implantation doping
For β-Ga 2 O 3 , ease of ion implantation doping has gained less attention compared with other features; however, this is definitely important for mass production of β-Ga 2 O 3 devices in the future. Here, ion implantation doping processes developed for β-Ga 2 O 3 by the author's group are presented.

n-type
Si-ion implantation doping process was developed to fabricate ohmic electrodes on n-Ga 2 O 3 [52]. Implanted Si atoms start to be activated by annealing at 800 • C, and the activation ratio becomes the highest at 900-1000 • C. Note that the activation ratio also depends on the Si doping density. The highest n of Si-doped β-Ga 2 O 3 layers formed by implantation was 3×10 19 cm −3 for a Si density of 5×10 19 cm −3 , which is high enough to form decent ohmic electrodes with a low contact resistance. The typical room-temperature electron μ of n-Ga 2 O 3 layers with an n of low-to mid-10 17 cm −3 formed by Si-ion implantation is ∼100 cm 2 /Vs [53], which is comparable with those of n-Ga 2 O 3 layers prepared by in-situ donor doping during epitaxial growth. The Si-ion implantation doping is useful for formation of not only ohmic electrodes but also FET channels; the author's group has been using this technique to form n-Ga 2 O 3 active regions in various types of devices.

Deep acceptor p-type
The author's group attempted to develop ion implantation doping processes for deep acceptor species to form a large energy barrier in n-Ga 2 O 3 provided by the built-in potential of the p-n junction rather than a hole-conductive p-Ga 2 O 3 layer. We first tried Mg as a deep-acceptor candidate, because it is predicted to be the most stable cation-site acceptor with the lowest formation energy in β-Ga 2 O 3 [22,23]. Mg-ion implantation doping into n-Ga 2 O 3 bulk crystals and subsequent annealing converts the n-type conductivity into semi-insulating, which is attributed to compensation of the donors by the doped Mg [54]. However, implanted Mg atoms significantly diffuse during the activation annealing; in consequence, it is difficult to precisely control their depth profile.
We chose N as an alternative candidate of deep acceptor for β-Ga 2 O 3 . N is the closest to O in terms of ionic size and electronic structure and thus can be expected to be the most possible anion-substituting acceptor. Implanted N atoms in n-Ga 2 O 3 effectively activate by annealing at temperatures of higher than 1100 • C and contribute to compensation of the background donors [54]. Furthermore, the N atoms exhibit much lower diffusivity in β-Ga 2 O 3 than Mg atoms, which helps to keep the doping profile even after high-temperature activation annealing.
Test structures with a Mg-or a N-doped current blocking layer (CBL) as shown in Fig. 7a were fabricated by ion implantation processes to compare their leakage characteristics. The N-doped structure provided much lower leakage current compared with the Mg-doped one [ Fig. 7b], because the N profile was maintained even after activation annealing due to its less diffusivity. These results indicate that N-ion implantation is more useful for device fabrication than Mg-ion implantation.

Schottky contact
A variety of Schottky metals including tungsten (W), copper (Cu), nickel (Ni), iridium (Ir), platinum (Pt), and gold (Au) have been reported for n-Ga 2 O 3 [55][56][57]. Pt and Ni are the most common ones used for device fabrication among them. The Schottky contacts on n-Ga 2 O 3 are usually of high quality, which is typified by an ideality factor close to unity. The barrier height between the metal and n-Ga 2 O 3 reported so far is 1.0-1.5 eV, depending on the metal species and the crystal orientation of β-Ga 2 O 3 surface. Layered oxide metal PdCoO 2 also forms a high-quality Schottky junction with n-Ga 2 O 3 [58].

Ohmic contact
Titanium (Ti)-based metal stacks typified by Ti/Au have been commonly used to fabricate ohmic electrodes on n-Ga 2 O 3 . It should be noted that Ti-based electrodes formed on n-Ga 2 O 3 can provide ohmic characteristics only in case of n > 10 18 cm −3 ; therefore, formation of n + -Ga 2 O 3 is absolutely necessary to fabricate ohmic contacts. Si-ion implantation doping and regrowth have been utilized to form n + -Ga 2 O 3 ohmic regions [52,59,60]. The post-metallization annealing provides short-range intermixing of Ti and Ga 2 O 3 at the interface, resulting in further reduction in contact resistance and improvement in reliability and endurance [61][62][63]. The routine process developed in the author's group based on the Si-ion implantation doping described in Section 5.1.1 and postmetallization annealing can provide a low specific contact resistance of less than 1×10 −5 cm 2 . Transparent amorphous oxides such as indium tin oxide and zinc oxide have also been tested as ohmic electrodes [64][65][66].

Etching
Dry etching processes such as BCl 3 -based reactive-ion etching (RIE) and inductively coupled plasma RIE (ICP-RIE) are most commonly used for fabrication of β-Ga 2 O 3 devices [67,68]. It is also possible to perform wet etching of β-Ga 2 O 3 using hot acid such as H 3 PO 4 and H 2 SO 4 [69,70]. Dry etching often causes plasma damage on the Ga 2 O 3 surface [71], and wet etching is effective to remove the damaged region [72]. Metal-assisted chemical etching, which is a technique to use Pt having a catalytic effect as an etching mask, is unique and useful to fabricate mesa structures and fin arrays with a high aspect ratio [73].

Direct wafer bonding
The performance of many types of semiconductor devices is limited by their heat dissipation capacity, since heat generation under high-power operation is inevitable even for high-efficiency power devices, and an electrical resistance of the drift layer increases with rising operation temperature due to a decrease in electron Fig. 7 a Schematic cross section of test structure with CBL fabricated using Mg-or N-ion implantation doping and b its vertical leakage current characteristics μ. Therefore, for Ga 2 O 3 device technologies, the low thermal conductivity is a serious potential weakness, and thermal management is one of the most important R&D challenges. One of the effective ways to improve heat dissipation from Ga 2 O 3 devices is heterogeneous integration of Ga 2 O 3 active layers with foreign substrates having high thermal conductivity such as SiC and diamond by direct bonding [74][75][76][77]. The author's group performed surface activated bonding between single-crystal β-Ga 2 O 3 and cost-effective polycrystalline SiC substrates. The Ga 2 O 3 /SiC bonded substrate demonstrated superior characteristics such as a large mechanical bonding strength of over 10 MPa, a small specific electrical resistance of 2×10 −4 cm 2 , and a negligibly small thermal resistance at the bonding interface [74].

SBD
WBG-and UWBG-semiconductor SBDs can offer some advantages against Si diodes mainly based on their large BFOMs: higher V br , lower on-resistance (R on ), higher efficiency, and faster switching speed. A variety of Ga 2 O 3 SBDs have been developed over the last decade. Primitive Ga 2 O 3 SBDs were fabricated with a β-Ga 2 O 3 bulk wafer synthesized by EFG [78]. At the time, β-Ga 2 O 3 epitaxial wafers with a thick drift layer having n < 10 17 cm −3 were hardly available, since MBE was the only technique that could provide high-purity epitaxial layers but was not able to provide n-Ga 2 O 3 layers with a low n satisfying requirement for fabrication of highvoltage devices due to technical limitation of donor doping [45]. The HVPE technology became a breakthrough for the development of vertical β-Ga 2 O 3 devices, because high-quality, low-doped n-Ga 2 O 3 drift layers were made available thanks to it. Simple β-Ga 2 O 3 SBDs with HVPEgrown drift layers demonstrated excellent current-voltage output characteristics following the thermionic emission and thermionic field emission models at forward-and reverse-bias conditions, respectively [79], which are typical behaviors for WBG and UWBG semiconductors. The next step was enhancement of V br to prove that β-Ga 2 O 3 SBDs are useful for high-voltage applications. First demonstration of Ga 2 O 3 SBDs with V br > 1 kV was achieved by applying a field plate to mitigate electric field concentration at the edge of the anode electrode [80]. The field-plated SBDs also showed good on-state device characteristics such as an R on of 5.1 m cm 2 and an ideality factor of 1.03. Another edge-termination structure often employed for power SBDs is a guard ring. Our latest SBDs with both a field plate and a N-implanted guard ring, which is schematically illustrated in Fig. 8, achieved an enhancement of V br to 1.43 kV with keeping the same level of R on [81]. Mg-ion implantation doping and fluorine treatment were also employed to form guard rings [82,83]. As another attempt to increase V br , β-Ga 2 O 3 trench SBDs have also been developed [84][85][86]. An anode electrode was formed to cover trench sidewalls on which an Al 2 O 3 dielectric was deposited. At the reverse-bias condition, the depletion region laterally spreads from the sidewalls, leading to an increase in V br . A very high V br of over 2 kV was demonstrated for the trench SBDs with keeping a low R on of ∼10 m cm 2 .

MESFET
The first demonstration of single-crystal Ga 2 O 3 transistors was achieved in the form of metal-semiconductor FET (MESFET) as shown in Fig. 9, which was fabricated using a 300-nm-thick Sn-doped n-Ga 2 O 3 channel layer grown on a Mg-doped semi-insulating β-Ga 2 O 3 (010) substrate by MBE [3]. The MESFET met a basic standard Fig. 9 Schematic cross section of Ga 2 O 3 MESFET. Adapted with permission [3]. Copyright 2012, AIP required as a transistor: drain current (I d ) modulation from the on-state to the off-state by gate voltage (V g ) swing. Although it was the first experimental attempt on fabricating Ga 2 O 3 FETs, fundamental device characteristics were reasonably good, such as a maximum I d of 15 mA/mm at a drain voltage (V d ) of 40 V, a large threeterminal off-state V br of ∼250 V, and a high I d on/off ratio of over four orders of magnitude. However, two shortcomings of poor source/drain ohmic contacts and small surface leakage were noted for the MESFET.

Depletion-mode MOSFET
To improve the ohmic contact, the Si-ion implantation doping process was developed [52], as described in Sections 5.1.1 and 5.2.2. As a next step of β-Ga 2 O 3 FET development, we fabricated depletion-mode (Dmode) MOSFETs with Si-implanted n-Ga 2 O 3 channel and n + -Ga 2 O 3 ohmic regions, and an Al 2 O 3 gate dielectric formed by plasma-assisted atomic layer deposition (ALD), as illustrated in Fig. 10 [87]. Thanks to a large reduction in the ohmic contact resistance, the maximum I d was largely increased to 65 mA/mm. The gate dielectric contributed to increasing the off-state V br to ∼400 V and drastically suppressing an off-state leakage current, which led to a large increase in the I d on/off ratio to over 10 10 .

Field-plated MOSFET
Based on the D-mode MOSFETs, β-Ga 2 O 3 field-plated MOSFETs were fabricated to further enhance V br by preventing electric field concentration at the drain-side edge portion of the gate metal. Figure 11 shows a crosssectional schematic of the Ga 2 O 3 MOSFET structure with a gate-connected field plate [88]. The field plate successfully contributed to a large enhancement of V br to ∼750 V without any degradation of on-state device characteristics from those of the D-mode MOSFETs. The field-plated MOSFETs also exhibited no DC-RF dispersion by virtue of the effect of the thick SiO 2 passivation and high operation stability up to 300 • C. Due to the continued development, the V br of lateral β-Ga 2 O 3 field-plated MOSFETs has increased to over 2 kV [89,90]. From the material properties typified by the large E g , β-Ga 2 O 3 is expected to have high resistance against stresses caused by high temperature, radiation, and corrosive gases, as well as being suited for power devices. Hence, it can be considered that β-Ga 2 O 3 FETs have a large potential even for wireless communications and signal processing circuits in harsh environments in which it is difficult for existing semiconductor devices to keep operating for a long term. Based on the idea, we also investigated gamma-ray tolerance of the Ga 2 O 3 fieldplated MOSFETs [91]. The MOSFETs showed only small degradation of DC I d -V d output characteristics even after high-dose cumulative gamma-ray irradiation up to 1.6 MGy. Note that the irradiation dose is at a level of exceeding requirements for typical space applications. These results indicate a great potential of β-Ga 2 O 3 FETs for applications to high-temperature and/or radiation-hard electronics.

Normally-off MOSFET
Enhancement-mode (E-mode) β-Ga 2 O 3 MOSFETs realizing normally-off operation were fabricated by decreasing a thickness and/or a doping concentration of a channel layer to enable full channel depletion under a gate at V g = 0 V [92]. β-Ga 2 O 3 MOSFETs having an unintentionally N-doped Ga 2 O 3 channel layer grown by plasma-assisted MBE as illustrated in Fig. 12 demonstrated normally-off operation with a large turn-on threshold V g of over +8 V [93]. The channel layer with UID N and Si densities of respectively 1×10 18 and 2×10 17 cm −3 is equivalent to p-type, since deep-acceptor N impurities compensate donor Si ones. These results would imply formation of an

High-frequency MOSFET
The author's group fabricated highly scaled β-Ga 2 O 3 MOSFETs as shown in Fig. 13, which aimed for applications to high-frequency wireless communications and logic circuits operating in harsh environments [94]. The MOSFETs with a gate length (L g ) of 200 nm demonstrated a current-gain cutoff frequency (f T ) of 9 GHz and a maximum oscillation frequency (f max ) of 27 GHz. Note that comparable RF small-signal characteristics were also reported from other institutes [95,96]. These RF device characteristics indicate that β-Ga 2 O 3 FETs can be used for wireless communications at frequencies up to ∼10 GHz.

Vertical FET
As of the time of this writing, there have been a few groups developing vertical β-Ga 2 O 3 FETs in the world. Two representative developments are discussed in this subsection.

Current-aperture FET
Vertical FETs with a current aperture were often selected as a first target in the history of vertical transistor developments due to its simple structure. Figure 14 shows a schematic cross section of a vertical D-mode β-Ga 2 O 3 MOSFET with a current aperture that is bounded laterally by N-implanted CBLs [97]. It should be noted that three ion implantation processes were conducted to form the N-doped p-Ga 2 O 3 CBLs, a Si-doped n-Ga 2 O 3 channel, and heavily Si-doped n ++ -Ga 2 O 3 source ohmic contact regions. Decent device characteristics such as a maximum I d of 0.42 kA/cm 2 and a specific R on of 31.5 m cm 2 were demonstrated. The I d on/off ratio of over 10 8 was recorded. The three-terminal off-state V br was limited to less than 30 V owing to large electric field in the Al 2 O 3 gate dielectric due to the high Si doping density of the channel.
E-mode operation of vertical β-Ga 2 O 3 MOSFETs was also demonstrated [98]. The device fabrication process and structure of the E-mode MOSFETs were almost the same as those of the D-mode MOSFETs. There were two modifications from the D-mode structure: a decrease in Si doping density of the channel from 1.5×10 18 to 5.0×10 17 cm −3 and formation of an n ++ -region-gate overlap to avoid channel depletion in the gate-source access region at thermal equilibrium. A turn-on threshold V g of larger than +3 V and an I d on/off ratio of over 10 6 were attained. The off-state V br was enhanced to ∼250 V thanks to the reduction in the Si doping density in the channel.

Fin-channel FET
Top-and side-gate configuration formed on mesa trenches of fin-channel FETs allows superior gate controllability compared with that of planar FETs. The finchannel FET structures are often developed for lowvoltage, high-speed applications; however, the gate configuration is also effective for vertical power FETs on electric-field management. The same group of Cornell University that developed the trench SBDs reported vertical β-Ga 2 O 3 FETs with a sub-μm fin channel [99]. Emode operation was achieved for the devices by means of full channel depletion from the sidewalls at V g = 0 V. The fin-channel FETs demonstrated excellent device characteristics such as a large I d of 1 kA/cm 2 , a specific R on of 10-20 m cm 2 , an I d on/off ratio of ∼10 8 , and an off-state V br of ∼1 kV.

Conclusions
This article presents an overview of current R&D status of β-Ga 2 O 3 science and engineering, which covers material properties, melt bulk growth, epitaxial thin-film growth, device processing, and state-of-the-art device technologies. The extremely large BFOM and the availability of high-quality, large-diameter single-crystal wafers produced from melt-grown bulks are especially attractive from practical application perspectives. Over the past 10 years, all the aspects of β-Ga 2 O 3 material and device technologies have significantly progressed, and several development milestones on the way to industrialization and commercialization of β-Ga 2 O 3 FETs and SBDs have already been achieved. It is gratifying that Ga 2 O 3 has gained a lot of popularity for the last 10 years thanks to the accomplishments and is now largely accepted as one of the promising UWBG materials in the semiconductor community. However, there is no room for doubt that the current β-Ga 2 O 3 device technologies are still immature, and that we still have a long way to start to introduce them into markets. Future intense efforts are indispensable to develop advanced device structures to improve their characteristics. Overcoming the two shortcomings related to fundamental material properties, which are the absence of hole-conductive p-Ga 2 O 3 and the poor thermal conductivity, should be tough challenges. The potential of β-Ga 2 O 3 devices is worth putting the effort into the developments, because it can be expected that β-Ga 2 O 3 power and harsh-environment devices will contribute to global energy saving and creation of new application fields of semiconductor device electronics, respectively. The continued efforts on material and device engineering should lead to generating far-reaching technological and socioeconomic impacts. The author read and approved the final manuscript.

Funding
Part of the work presented in this article was supported by Council for Science, Technology and Innovation (CSTI), Cross-ministerial Strategic Innovation Promotion Program (SIP), Next generation power electronics (funding agency: New Energy and Industrial Technology Development Organization), and the Strategic Information and Communications R&D Promotion Program (SCOPE) of the Ministry of Internal Affairs and Communications, Japan.