# Fileset

[Progress and challenges in the development of ultra-wide.pdf](https://mdr.nims.go.jp/filesets/0391468d-45ab-44f7-92ef-879954d63a83/download)

## Creator

[Yuichi Oshima](https://orcid.org/0000-0001-8293-4891), Elaheh Ahmadi

## Rights

This content may be downloaded for personal use only. Any other use requires prior permission of the author and the publisher. This material originally appeared in Yuichi Oshima et al., Appl. Phys. Lett. 121, 260501 (2022) and may be found at https://doi.org/10.1063/5.0126698.[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Progress and challenges in the development of ultra-wide bandgap semiconductor α-Ga2O3 toward realizing power device applications](https://mdr.nims.go.jp/datasets/0f47750c-077e-4f42-a950-c1919ea5f2b6)

## Fulltext

Progress and challenges in the development of ultra-wide bandgap semiconductor α-Ga2O3 toward realizing power device applicationsViewOnlineExportCitationCrossMarkRESEARCH ARTICLE |  DECEMBER 29 2022Progress and challenges in the development of ultra-widebandgap semiconductor α-Ga2O3 toward realizing powerdevice applications Yuichi Oshima   ; Elaheh Ahmadi Appl. Phys. Lett. 121, 260501 (2022)https://doi.org/10.1063/5.0126698 CHORUS 29 January 2024 06:12:56https://pubs.aip.org/aip/apl/article/121/26/260501/2834878/Progress-and-challenges-in-the-development-ofhttps://pubs.aip.org/aip/apl/article/121/26/260501/2834878/Progress-and-challenges-in-the-development-of?pdfCoverIconEvent=citehttps://pubs.aip.org/aip/apl/article/121/26/260501/2834878/Progress-and-challenges-in-the-development-of?pdfCoverIconEvent=crossmarkjavascript:;https://orcid.org/0000-0001-8293-4891javascript:;https://orcid.org/0000-0002-8330-9366javascript:;https://doi.org/10.1063/5.0126698https://pubs.aip.org/aip/apl/article-pdf/doi/10.1063/5.0126698/16489113/260501_1_accepted_manuscript.pdfhttps://servedbyadbutler.com/redirect.spark?MID=176720&plid=2291284&setID=592934&channelID=0&CID=842343&banID=521636251&PID=0&textadID=0&tc=1&scheduleID=2211497&adSize=1640x440&data_keys=%7B%22%22%3A%22%22%7D&matches=%5B%22inurl%3A%5C%2Fapl%22%5D&mt=1706508776547948&spr=1&referrer=http%3A%2F%2Fpubs.aip.org%2Faip%2Fapl%2Farticle-pdf%2Fdoi%2F10.1063%2F5.0126698%2F16489112%2F260501_1_online.pdf&hc=4bf7d69e06aac3980aa98218a0730506e59af19d&location=Progress and challenges in the developmentof ultra-wide bandgap semiconductor a-Ga2O3toward realizing power device applicationsCite as: Appl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698Submitted: 16 September 2022 . Accepted: 6 December 2022 .Published Online: 29 December 2022Yuichi Oshima1,a) and Elaheh Ahmadi2AFFILIATIONS1Optical Single Crystals Group, National Institute for Materials Science, Tsukuba, Ibaraki 3050044, Japan2Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, Michigan 48109, USAa)Author to whom correspondence should be addressed: OSHIMA.Yuichi@nims.go.jpABSTRACTUltra-wide-bandgap (UWBG) semiconductors, such as Ga2O3 and diamond, have been attracting increasing attention owing to theirpotential to realize high-performance power devices with high breakdown voltage and low on-resistance beyond those of SiC and GaN.Among numerous UWBG semiconductors, this work focuses on the corundum-structured a-Ga2O3, which is a metastable polymorph ofGa2O3. The large bandgap energy of 5.3 eV, a large degree of freedom in band engineering, and availability of isomorphic p-type oxides toform a hetero p–n junction make a-Ga2O3 an attractive candidate for power device applications. Promising preliminary prototype devicestructures have been demonstrated without advanced edge termination despite the high dislocation density in the epilayers owing to theabsence of native substrates and lattice-matched foreign substrates. In this Perspective, we present an overview of the research and develop-ment of a-Ga2O3 for power device applications and discuss future research directions.Published under an exclusive license by AIP Publishing. https://doi.org/10.1063/5.0126698I. INTRODUCTIONGiven the strong demand to suppress CO2 emissions and saveenergy, it is urgent to promote efficient energy consumption alongwith the use of reusable energy sources. It is essential to reduce energyloss in power converters to efficiently utilize electrical energy. Powerconverters are used in virtually all electrical equipment. Currently,most power converters employ power devices based on Si semicon-ductors. However, their efficiency is close to the material limit. Theuse of wide bandgap semiconductors is a promising method to exceedthe Si limit. This is because the on-resistance (Ron) of a drift layer in apower semiconductor device is inversely proportional to the cube ofthe critical field for breakdown (EC), and EC tends to increase with theincrease in bandgap energy Eg, as shown in Fig. 1.1 Accordingly, Ron ofa drift layer is expressed asRon ¼4V2BelE3c; (1)where VB indicates the break down voltage, e symbolizes relativedielectric constant, and l corresponds to electron mobility. Thedenominator elE3c is referred to as Baliga’s figure of merit (BFOM).2The BFOM indicates how small Ron can be in principle. Table I sum-marizes the material properties and BFOM of the selected semicon-ductors. For instance, the BFOM of 4H–SiC is 340, which implies thatRon of a 4H–SiC device can be 1/340 of that of Si. It is worth mention-ing that intrinsic experimental values of e, l, and Ec for a-Ga2O3 areunavailable at the moment. Among existing wide bandgap semicon-ductors, 4H–SiC and GaN are the most well-developed materials forpower device applications, and these devices have gradually penetratedthe market.a-Ga2O3, the target material of this study, has a larger Eg thanthose of 4H–SiC and GaN, and as a result, a higher Ec can be expected.The first-principles calculations indicate that the energy gap should beindirect.6,7 Moreover, a-Ga2O3 can be grown on isomorphic sapphire(a-Al2O3), of which large-scale wafers are commercially available at areasonable price. Moreover, a-Ga2O3 has a large degree of freedom inband engineering and conductivity control owing to an abundance ofcorundum-structured oxides, and it is relatively easy to create solidsolutions with them. Additionally, a-Ga2O3 is a potential candidate forsolar-blind UV detectors because Eg¼ 5.3 eV corresponds tok¼ 238nm, which is shorter than the shortest wavelength of the solarspectrum on the surface of the Earth. Herein, we present an overviewAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-1Published under an exclusive license by AIP PublishingApplied Physics Letters PERSPECTIVE scitation.org/journal/apl 29 January 2024 06:12:56https://doi.org/10.1063/5.0126698https://doi.org/10.1063/5.0126698https://doi.org/10.1063/5.0126698https://www.scitation.org/action/showCitFormats?type=show&doi=10.1063/5.0126698http://crossmark.crossref.org/dialog/?doi=10.1063/5.0126698&domain=pdf&date_stamp=2022-12-29https://orcid.org/0000-0001-8293-4891https://orcid.org/0000-0002-8330-9366mailto:OSHIMA.Yuichi@nims.go.jphttps://doi.org/10.1063/5.0126698https://scitation.org/journal/aplof the a-Ga2O3-related technologies and discuss future perspectives byfocusing on power device applications. Regarding UV applications, acomprehensive review was conducted by Biswas and Nishinaka.8II. MATERIAL PROPERTIES AND POTENTIALFOR POWER DEVICE APPLICATIONS OF a-Ga2O3Ga2O3 crystalizes into five different structures.9,10 Among theGa2O3 polymorphs, the corundum-structured a-phase is the secondmost well developed material for semiconductor device applications,next to the most thermally stable b-phase. In addition to the largest Egamong the Ga2O3 polymorphs, a-Ga2O3 exhibits several advantages.First, there are many corundum-structured oxides, as illustrated inFig. 2, and a-Ga2O3 can be used to create solid solutions.11 Therefore,a-Ga2O3 has a higher degree of freedom for band engineering, and itmay even be possible to provide multi functionality. For instance,a-(AlxGa1-x)2O3 can be grown without a compositional limitation12–14compared to the case of b-Ga2O3, and Eg can be controlled over awide range (Fig. 3). As depicted in Fig. 3, the Eg of a-(AlxGa1-x)2O3can be increased to 8.8 eV, which cannot be reached byb-(AlxGa1-x)2O3, indicating a possibility for the realization of furtherhigh-performance power devices beyond a-Ga2O3. In particular, nearlydefect-free devices can be fabricated when the Al content is sufficientlyhigh to enable the growth of coherently strained a-(AlxGa1-x)2O3 onsapphire. Ferromagnetism can be formed by alloying with a-Fe2O3.15Second, isomorphic p-type oxides with relatively small lattice mismatch(<0.3%), such as a-(GaxIr1-x)2O3 and a-(GaxRh1-x)2O3,11 can beFIG. 2. Relationship between the bandgap and lattice constant along the a-axis ofcorundum-structured oxides. Reproduced with permission from Kaneko et al., Jpn.J. Appl. Phys., Part 1 57, 02CB18 (2018). Copyright 2018 The Japan Society ofApplied Physics.11FIG. 3. Bandgap energy in terms of averaged bond length/lattice constant alongthe a-axis of a-(AlxGa1-x)2O3 and a-(InxGa1-x)2O3. Reprinted with permission fromS. Fujita and K. Kaneko, J. Cryst. Growth 401, 588–592 (2014). Copyright 2014Elsevier.12FIG. 1. Bandgap dependence of Ec. Reproduced with permission from Higashiwakiet al., Appl. Phys. Lett. 100, 013504 (2012). Copyright 2012 AIP Publishing.1TABLE I. Material properties and BFOM of semiconductors. The values are fromRef. 3 except those for a-Ga2O3.Si 4H–SiC GaN b-Ga2O3 a-Ga2O3 DiamondEg (eV) 1.1 3.3 3.4 4.5 5.3 5.5e 11.8 9.7 9 10 12.8a 5.5l (cm2/V s) 1400 1000 1200 200 200b 2000EC (MV/cm) 0.3 2.5 3.3 6.5 9.5c 10BFOM (vs Si) 1 340 870 1231 4921 24 661aTheoretical value obtained via first-principles calculations.4bAssumed to be the same as that for b-Ga2O3. The highest reported value for a-Ga2O3is 65 cm2/V s.5cExpected value from the empirical curve shown in Fig. 1.Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-2Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/aplutilized to create a hetero-p–n junction16–18 unlike the case of b-Ga2O3.Recently, Zhang et al. reported a b-Ga2O3 hetero-p–n-junction diodeusing NiO as the p-type layer, and the performance surpassed thematerial limits of SiC and GaN despite the crystal structure differencebetween b-Ga2O3 and NiO.19 This fact may mean that the necessity formatching the crystal structure and lattice parameters is debatable.Another advantage is that a single-crystalline a-Ga2O3 film can begrown on a large-scale using cost-effective sapphire substrates.However, the dislocation density in the epilayer is extremely high ifcountermeasures are not taken. The crystal plane of the epilayer can becontrolled by the choice of the crystal plane of sapphire. In the case ofb-Ga2O3, a b-Ga2O3 film can also be grown heteroepitaxially.However, the crystal plane is virtually limited to (201), on which stack-ing faults are readily formed, and the epilayer includes in-plane rota-tional domains20 although such domains can be suppressed to a certaindegree by growing films on off-angled substrates.21 Therefore, anexpensive singe crystal b-Ga2O3 substrate needs to be used even if theepilayer does not have to be of ultra-high quality.The major drawbacks of a-Ga2O3 are as follows. First, the dislo-cation density in an epilayer is extremely high because of the absenceof native a-Ga2O3 and lattice-matched foreign substrates. Moreover,the use of foreign substrates (sapphire in most cases) results in thebowing and cracking of the epi-wafer caused by thermal stress owingto the difference in the thermal expansion coefficients. Additionally,a-Ga2O3 epilayers grown on insulating sapphire substrates require acomplicated fabrication process to produce vertical-structured devices.Second, a-Ga2O3 is a metastable phase, which turns to the b-phase athigh temperatures of approximately 550 �C or higher.22 Therefore,damage recovery by thermal annealing would be difficult even if ionimplantation is performed. Fortunately, thermal annealing of electro-des should not be problematic, because optimal contact performancecan be achieved below the transition temperature, and thermal treat-ment at higher temperatures degrades contact performance, similar tothat of b-Ga2O3.23,24 Interestingly, the transition temperature tends toincrease as the film thickness decreases, as shown in Fig. 4.25Moreover, it has been reported that the transition initiates from thefilm surface.25 Under ultra-high pressure, a-Ga2O3 is thermodynami-cally the most stable among the Ga2O3 polymorphs.26 The increasedthermal stability of thin a-Ga2O3 films is most likely caused by thecompressive strain owing to the lattice mismatch, which effectively sta-bilizes the a-phase. Formation of a capping layer on the surface of ana-Ga2O3 film effectively increases the thermal stability.27 McCandlessreported that the transition temperature of a 57-nm-thick a-Ga2O3with a capping layer of SiO2 or Mo increased to �800 �C while that ofthe control sample without a capping layer was �600 �C.27 Cappingwith a crystalline Al2O3 layer further increased the transition tempera-ture to�900 �C, which can be attributed to the thermal diffusion of Alinto the underlying a-Ga2O3 layer. Thus, alloying with a-Al2O3dramatically increases the thermal stability. Jinno et al. clarified theAl composition dependence of thermal stability of c-planea-(AxGa1-x)2O3 films.28 As displayed in Fig. 5, the transition tempera-ture is 650 �C when x¼ 0, which increases to 950 �C when x¼ 0.45.No transition was observed when x> 0.6, even after annealing at1100 �C. Hence, ion implantation and the following thermal annealingfor damage recovery may be possible in high-Al-composition films.Third, thermal conductivity of a-Ga2O3 should be low, which in prin-ciple would have an adverse impact on heat dissipation from powerdevices. Although experimental thermal conductivities of a-Ga2O3have not been reported, Yang et al. theoretically estimated the thermalconductivities up to 800K, as displayed in Fig. 6, by combining thefirst principles calculation and iteratively solving the Boltzmann trans-port equation.29 Based on their calculation, the thermal conductivitiesshould be 11.61W/mK along [100], 9.38W/mK along [010], and8.04W/mK along [001], which are considerably lower than those ofGaN (210W/mK30) and 4H–SiC (347W/mK31). Note that the experi-mental thermal conductivity of b-Ga2O3 is the highest along [010]FIG. 4. Thermal stability of a-Ga2O3 films as a function of the film thickness. Opensymbols indicate samples that maintained the a-phase. The solid symbols indicatesamples that completely converted to the b-phase. Samples drawn in red and bluecolors were grown using Ga(acac)3 and GaCl3 as a Ga precursor. The circular, tri-angular, and rhomboid symbols illustrate the growth temperature of 500, 600, and700 �C, respectively. Reproduced with permission from Jinno et al., AIP Adv. 10,115013 (2020). Copyright 2020 AIP Publishing.25FIG. 5. Thermal stability of the c-plane a-(AxGa1-x)2O3 films as a function of the Alcomposition. The blue crosses illustrate the samples that converted to the b-phasewhile the red circles represent the samples that maintained the a-phase.Reproduced with permission from Jinno et al., Jpn. J. Appl. Phys., Part 1 60,SBBD13 (2021). Copyright 2021 The Japan Society of Applied Physics.28Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-3Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/apl(27W/mK) and lowest along [010] (10.9W/mK).32 The low thermalconductivity is a common problem associated with Ga2O3.III. EPITAXYThe development of epitaxial growth methods is essential forrealizing high-performance a-Ga2O3 power devices. Epitaxial growthof a-Ga2O3 has been primarily investigated via mist chemical vapordeposition (mist CVD), halide vapor phase epitaxy (HVPE), molecularbeam epitaxy (MBE), and metalorganic vapor phase epitaxy(MOVPE). A wide range of growth temperatures (350–880 �C) havebeen reported because the thermal stability of a-(AlxGa1-x)2O3depends on the film thickness and Al composition, as described in Sec.II. In this section, we present an overview of the current situation ofthese growth methods and then discuss detailed technical aspects suchas substrates, conductivity control, and defect control.A. Growth methods1. Mist CVDIn this method, an aqueous solution containing a gallium source,such as gallium (III) acetyl acetonate or GaCl3, is atomized via ultra-sonic vibration.34,35 Subsequently, the mist is transferred to the reactortogether with a carrier gas, such as N2 or O2, and a-Ga2O3 is grown ona substrate by exploiting the chemical reaction between the galliumsource and H2O or O2. The method realized the first epitaxial growthof a-Ga2O3.34 The growth conditions have been well established togrow phase-pure a-Ga2O3 epilayers. Moreover, lm-order-thick filmscan be grown at a typical growth rate of approximately 1lm/h.Additionally, it is possible to grow solid solutions such as a-(AlxGa1-x)2O3,12 and the growth of a-(AlxGa1-x)2O3/a-(AlyGa1-y)2O3super lattice was demonstrated.36Residual impurity concentrations in mist-CVD-grown a-Ga2O3tend to be high. For instance, Uno et al. reported that their UID filmincluded H, C, and Si at following concentrations: [H]¼ 2� 1019 cm�3,[C]¼ 7� 1017 cm�3, and [Si]¼ 3� 1018 cm�3.37 Plausible sources of H,C, and Si include H2O, quartz (SiO2), and acetylacetone, respectively.SiO2 and a high concentration of H2O could react with each other athigh temperatures to release Si-included species as contaminants.Therefore, it is essential to establish a technique to suppress the residualimpurities to grow a drift layer for a high- VB device. Note that it is occa-sionally stated that the mist-CVD growth apparatus is cost-effective thanthose of metalorganic chemical vapor deposition (MOCVD) or halidevapor phase epitaxy (HVPE). However, this is probably based on thelab-level simple structure of the mist-CVD apparatus. If an industry-levelapparatus with gas lines for multiple growth precursors and dopants isfully equipped with appropriate components, such as an optimized lami-nar flow channel, mass-flow controllers, air-operated valves, program-mable logic controllers, and an exhaust system, the machine cost wouldbe comparable to those of MOCVD or HVPE.2. HVPEIn this method, a-Ga2O3 is grown using GaClx and O2 as precur-sors.38–40 GaClx (x¼ 1 or 3) is produced upstream in the reactor in thechemical reaction involving Ga and HCl/Cl2 (hereinafter denoted asHCl-HVPE38,39 and Cl2-HVPE,40 respectively). Rapid growth is one ofthe most attractive points of HVPE. In the case of HCl-HVPE, thegrowth rate exceeds 100lm/h when GaCl3 and O2 are used as the pre-cursors, as illustrated in Fig. 7.39 Such a high growth rate is beneficialfor growing a thick a-Ga2O3 film for a drift a layer or freestandinga-Ga2O3 wafer. To date, there have been no reports on the HVPE ofa-(AlxInyGa1-x-y)2O3 solid solutions. However, it should be possiblebecause the HVPE of In2O3 has already been demonstrated,41 and thesupply of AlCl3 is a well-established technique in the HVPE of AlN.42The most major impurities in HCl-HVPE-grown UID a-Ga2O3include H and Cl, and the concentrations tend to increase at highgrowth rates. For instance, in our case, [H] is below the detectionlimit of SIMS measurement (5� 1016 cm�3), and [Cl] is typically5� 1016 cm�3 at a growth rate of 15lm/h. When the growth ratewas increased to 100lm/h, [H] and [Cl] increased to 2.8� 1017and 1.4� 1018 cm�3, respectively.39 In b-Ga2O3, Cl is theoreticallyFIG. 7. HVPE growth rate of a-Ga2O3 as a function of the GaCl supply vapor pres-sure under an additional HCl supply to convert GaCl–GaCl3. The O2 supply wasfixed at 3.125 kPa. The calculated equilibrium vapor pressures of H2, HCl, GaCl,and GaCl3 are also displayed. The calculated results reveal that GaCl should beeffectively converted to GaCl3. Reproduced with permission from Oshima et al.,Semicond. Sci. Technol. 35, 055022 (2020). Copyright 2020 IOP Science.39FIG. 6. Calculated lattice thermal conductivity of a-Ga2O3 as a function of the tem-perature. Calculated and experimental values for b-Ga2O3 are also shown for com-parison. Reproduced with permission from Yang et al., J. Vac. Sci. Technol. A 40,052801 (2022). Copyright 2022 The American Vacuum Society (AVS).29Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-4Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/aplpredicted to be a shallow donor.43 It is likely that Cl is a shallow donor ina-Ga2O3 despite the lack of theoretical calculations. In reality, HVPE-grown UID a-Ga2O3 films are highly resistive, and n-type conductionowing to Cl impurity has not been reported yet. Extended research isrequired to clarify whether Cl in a-Ga2O3 is inactive or just compensated.The background Si concentration is below the detection limit of SIMSmeasurement (� 1015 cm�3), even at a growth rate of 100lm/h, despitethe existence of H2 (by-produced by the reaction of Ga and HCl) in thequartz reactor tube probably owing to the low growth temperature.3. MBEIn this method, a-Ga2O3 is grown using gallium vapor flux andplasma-activated oxygen (presumably atomic oxygen) as precursors. a-(AlxGa1-x)2O3 can be grown across the entire compositional range.13 Inthe case of MBE, there is a critical thickness beyond which the film iscontaminated by the b-phase. There have been no previous reports onsuch critical thickness in the cases of mist CVD or HVPE. The criticalthickness depends on the crystal plane. Reported maximum thicknessesfor phase-pure MBE-grown a-Ga2O3 films are 3.3nm for (0001),4414.3nm for (1120),44 51nm for (1010),13 and 217nm for (1012).45Moreover, residual impurity concentrations in MBE-growna-Ga2O3 have not been reported yet. In the case of b-Ga2O3, the primaryresidual donor is Si, which is believed to be originated from quartz partsused in the plasma cell.46 It is likely that MBE-grown a-Ga2O3 exhibit asimilar tendency.4. MOVPEMOVPE is a commonly utilized growth method in the III–Vsemiconductor industry. MOVPE has been applied to R&D ofb-Ga2O3, and high-quality epilayers and promising device prototypeshave been demonstrated.47,48 More recently, MOVPE of phase-purea-Ga2O3 was reported by Bhuiyan et al.14 They grew a a-Ga2O3 filmon (1010) sapphire using trimethylgallium and O2 as the precursors.14Moreover, they reported the growth of a-(AlxGa1-x)2O3 in a full Alcomposition range using trimethylaluminum as the Al source.14Residual impurities in MOVPE-grown a-Ga2O3 have not beenreported yet. Generally, MOVPE-grown films tend to include carbonas a residual impurity, which originates from the metalorganic precur-sors. For example, the carbon concentration in MOVPE-grown UIDGaN exhibits a tendency to increase as the growth temperature or V/III ratio decreases.49 Depending on the growth condition, MOVPE-grown b-Ga2O3 includes carbon impurities. For example, Ikenagaet al. reported that the carbon concentration in their MOVPE-grownb-Ga2O3 film was below the detection limit (3� 1016 cm�3) when thegrowth temperature was 1000 �C; however, it increased to�1.5� 1018 cm�3 at 800 �C.50 In the case of a-Ga2O3, the growth tem-perature is even lower (650 �C,14 for instance). The low growth tem-perature can suppress the combustion of hydrocarbon and enhancethe incorporation efficiency of carbon impurity, resulting in a highercarbon concentration in the crystal.B. SubstratesSapphire, synthetic mica51 LiNbO3, and LiTaO352 have beenreported as potential substrates for a-Ga2O3. Among them, sapphire isthe only realistic choice for industrial use at present owing to itsisomorphic structure and availability of large-diameter and high-quality wafers at a reasonable price. Moreover, a-Ga2O3 can be grownon various crystal planes of sapphire such as (0001), (1120), and(1010). Higher electron mobility has been reported in m-plane filmsthan in c-plane films5 although it is unclear whether the difference isintrinsic or because of the crystal-orientation-dependent defect charac-ter/distribution.C. Conductivity controlGroup-IV elements, such as Si, Ge, and Sn, which substitute Gasites, are used as shallow donors in a-Ga2O3, and the related dopingtechniques are well established. In the case of mist CVD, a dopantmaterial, such as SnCl2 or ClSi(CH3)2((CH2)2CN), is dissolved in theaqueous solution of the growth precursors.53,54 A previous studyreported on n-type conductivity control via F doping.55 In the case ofHVPE, dopant gas sources, such as SiH4 and GeCl4, have beenreported.56,57 Moreover, electron mobility decreases at low carrier con-centrations probably because of the scattering by dislocations.5,56 Sonet al. experimentally demonstrated that electron mobility increased asthe dislocation density decreased (reflected to the FWHM of the x-rayrocking curve), as illustrated in Fig. 8.56 Based on a theoretical estima-tion, the scattering by dislocations should be negligible even in thenon-degenerated region when the dislocation density is less than107–108 cm�2.59 Moreover, p-type conductivity control of a-Ga2O3would be virtually impossible besides the case of b-Ga2O3, in whichthe causes of the difficulty are believed to be the absence of shallowacceptors, large effective hole mass, and formation of small polarons.60Fortunately, as mentioned in Sec. II, corundum-structured p-typeoxides, such as a-(GaxIr1-x)2O3, with relatively small lattice mismatchare available to form a hetero-p–n junction.IV. DEFECT CONTROLA. DislocationsIn a conventional a-Ga2O3 film, the dislocation density is as highas 1010 cm�2 because of the lattice mismatch between a-Ga2O3 andFIG. 8. Electron mobility as a function of the free electron density in HVPE-growna-Ga2O3 films with different dislocation densities, reproduced with permission fromof Son et al., ECS J. Solid State Sci. Technol. 9, 055005 (2020). Copyright 2020IOP Science.56 The dislocation densities were calculated based on the x-ray rock-ing curve widths by using a method described in Ref. 58.Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-5Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/aplsapphire (Da/a �4.5%, Dc/c �3.3%).61,62 Among the dislocations,edge dislocations comprise the vast majority in many cases for c-planea-Ga2O3 on sapphire. It is reflected to the broad FWHM of an x-rayrocking curve measured in skew-symmetric geometry while theFWHM is extremely narrow when measured in symmetric geometry,as depicted in Fig. 9. Accordingly, XRC measurements must be per-formed in both geometries to accurately estimate the mosaicity of thefilm. As mentioned in Sec. III, dislocations with such high density scat-ter electrons significantly to reduce the mobility. Accordingly, the dis-location density should be decreased to 107–108 cm�3 or less.59 Notethat the properties of crystal defects, including dislocations, grainboundaries, and stacking faults, are at the moment poorly understoodfor any phase of Ga2O3. Hence, the impact of dislocations on the elec-trical characteristics or device performance is probably not limited tothe reduction in electron mobility. It is likely that dislocations ina-Ga2O3 also act as the current leak path or diffusion path forimpurities.Several techniques have been reported to decrease the dislocationdensity, including super lattice buffer layers,36 use of patterned sap-phire substrates (PSS),63 and epitaxial lateral overgrowth technique(ELO).35,62,64–66 Son et al. performed the HVPE of a-Ga2O3 on a PSS,and the dislocation density was reported to be 8.4� 109 cm�2 whilethat in a control sample was 1.6� 1010 cm�2.63 It is probably becauseof the lateral growth of a-Ga2O3 at the initial 3D-growth stage.However, the effect was limited because of the short period (�1lm)of the pattern. Jinno et al. grew an a-Ga2O3 layer via mist CVD on aquasi-graded a-(AlxGa1-x)2O3 super lattice buffer layer and obtainedreduced screw and edge dislocation densities of 3� 108 and6� 108 cm�2, respectively.36 Oshima et al. performed the ELO ofa-Ga2O3 by HVPE and reported that the dislocation density was lessthan 5� 106 cm�2 in the laterally over grown area on the mask (Fig.10).62 However, the dislocation density was still extremely high on thewindow areas because the dislocations in the seed layer propagatedinto the regrown a-Ga2O3 through the windows and reached the topsurface. Additionally, a line of dislocations was newly formed at a coa-lesced boundary.67 There are several strategies to solve the problem,which can be learned from the case of GaN. The modification of themask pattern is one such method. The fraction of the defective area,i.e., window area and coalesced boundaries, can be decreased bymaking the mask width wider, and window width narrower. Thickfilm growth is another effective method.68 During the thick filmgrowth, dislocations with Burgers vectors of opposite signs attract eachother to annihilate by making a dislocation loop, and dislocations withBurgers vectors of the same sign should disperse to uniformly distrib-ute because of the repulsive interaction. Ma et al. demonstrated thereduction in the dislocation density by thick film growth in mist-CVD-grown a-Ga2O3, as illustrated in Fig. 11.58 Combining the modi-fication of the mask pattern and thick film growth will result in a betterresult. HVPE is a suitable growth method to shorten the time for thickfilm growth. Double ELO, in which the windows of the second maskare located to avoid the defective areas of the first ELO layer, is anotherplausible option;65 however, it may not be cost effective because theprocess requires two cycles of photolithography and growth.FIG. 9. (a) Typical x-ray rocking curves for 0006 and 1012 diffractions measured insymmetric and skew-symmetric geometries, respectively. Moreover, the FWHMs for0006 and 1012 are illustrated. The sample was an HVPE-grown a-Ga2O3 film on(0001) sapphire. (b) Crystal structure of a-Ga2O3 showing (0001) and (1012)planes.FIG. 10. SEM and TEM images of ELO-a-Ga2O3 stripes. (a) Plan-view SEM image,(b) cross-sectional TEM image, and (c) plan-view TEM image. Reproduced withpermission from Oshima et al., APL Mater. 7, 022503 (2019). Copyright 2019 AIPPublishing.FIG. 11. (a) and (b) XRCs of 0006 and 1014 diffractions for a-Ga2O3 epilayers withdifferent thicknesses, respectively, and (c) edge and screw dislocation densities (Deand Ds, respectively), calculated based on XRCC-FWHMs as a function of the filmthickness. The dashed line represents a least squares fit to the relationship of De/ 1/h. Reproduced with permission from Ma et al., Appl. Phys. Lett. 115, 182101(2019).Copyright 2019 AIP Publishing.58Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-6Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/aplB. Deep levelsDeep levels significantly impact device performance through thetrapping and releasing of carriers. Hence, it is essential to comprehendwhat type of traps exist, their origins, how they affect device perfor-mance, and how the density can be controlled. Some reports discussthe theoretical and experimental investigations of deep levels ina-Ga2O3, although the number of the reports is limited compared tothat of b-Ga2O3. Koayashi et al. reported the formation energies ofnative point defects [vacancies (VGa and VO), interstitials, and vacancypairs] and charge transition levels of the native point defects in a-Ga2O3based on first-principle calculations.7 Experimental investigations havebeen conducted by means of deep-level transient spectroscopy (DLTS),deep-level optical spectroscopy (DLOS), photocapacitance, and photoin-duced current.69–71 For instance, Takane et al. investigated the deep lev-els in a mist-CVD-grown UID a-Ga2O3 film on (1010) sapphire viaDLOS and photocapacitance.71 Consequently, three types of deep levels,namely, E1 (CBM–2.0 eV, 3.5� 1014 cm�3), E2 (CBM–2.5 eV,3.6� 1014 cm�3), and E3 (CBM–3.2 eV, 6.2� 1015 cm�3), were detected,as illustrated in Fig. 12, where CBM refers to conduction band mini-mum. Takane et al. compared their results with the theoretical calcula-tions of Kobayashi et al.7 and found that E3 is likely to originate fromVGa or VGa–VO complex. Refer to the excellent review by Polyakov et al.for more details and other reports.72V. DEVICESA. SBDsa-Ga2O3-based SBDs have been primarily reported by the groupsof Kyoto University and FLOSFIA, Inc.33,73–75 As depicted in Fig. 13,the research group fabricated vertical-structured SBDs (GaOTMSBD)by transferring a mist-CVD-grown a-Ga2O3 film on a metal substratewith high thermal conductivity.73 Unfortunately, the details of thetransfer process have not been disclosed. However, it should be notedthat the mechanical bonding strength between an a-Ga2O3 film and asapphire substrate is relatively weak, and the film can be easily sepa-rated from the substrate via thermal stress when the film is thickerthan approximately 5lm. It is likely that this feature is beneficial forthe transfer process. The SBDs comprises a simple planar structurewithout guarding layers and passivation layers. The Ron and VB were0.1 mX cm2 and 531V for one of the SBDs (SBD1), and 0.4 mX cm2and 855V for another SBD (SBD2), respectively. The Ron of SBD1 isapproximately 1/7 of a commercially available SiC SBD and beyondthe SiC limit (Fig. 14). The relatively small thickness of the a-Ga2O3layer (�10lm) and the use of a metal substrate contributed to mini-mize the Ron as well as maximizing the heat dissipation. Consequently,an extremely low heat resistance of 2.7 �C/W, which is as low as thatof a commercial SiC device, was reported even for an ampere-classSBD mounted in a standard TO220 package.33 The reverse recoverycharacteristics of an ampere-class SBD were reported to be equivalentor even faster than SiC devices, as illustrated in Fig. 15.33 Remarkably,it has been reported that the forward and reverse I–V characteristics ofthe a-Ga2O3-based SBDs were comprehensively explained by ther-moionic and thermoionic field emission models, respectively, from294 to 423K, despite the high dislocation density.75 The ideality factorwas 1.03, and it was independent of the temperature through294–423K.75 FLOSIA has been shipping engineering samples of theirSBDs (VB¼ 600V, Imax¼ 10A77) and PFC power supply boards forevaluation. Table II lists the specifications of the PFC power supplyboard.78 Figure 16 depicts the PFC power supply board combinedwith a DC–DC converter, in which a-Ga2O3-based SBDs are also usedfor driving an LED light.B. PN diodesModulation doping based on a p–n junction is one of themost effective methods for improving VB and Ron. Hence, it is neces-sary to clarify the characteristics of the a-(IrxGa1-x)2O3/a-Ga2O3 het-ero p–n junction. Kan et al. fabricated a p–n junction diode using anFIG. 12. (a) Relationship between DCand photon energy. The inset illustratesthe enlarged view. (b) The optical crosssection and the fitting results using theP€assler model (red lines). (c) Distributionof deep traps in the bandgap and theirconcentration. Reproduced with permis-sion from Takane et al., Phys. StatusSolidi B 258, 2000622 (2021). Copyright2021 Wiley.FIG. 13. Fabrication process of a GaOTMSBD. A thin a-Ga2O3 film is lifted off a sapphire substrate and transferred onto a supporting metal substrate. Reproduced with permis-sion from Oda et al., Appl. Phys. Express 9, 021101 (2016). Copyright (2016) The Japan Society of Applied Physics.Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-7Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/apla-Ir2O3/a-Ga2O3 hetero structure and confirmed the rectifying I–Vcharacteristics, as displayed in Fig. 17.16 Moreover, they clarified thetype-II band alignment of the junction by XPS, as shown in Fig. 18.The band offset for holes was as large as 4.7 eV.16 Kaneko et al. investi-gated a-(GaxIr1-x)2O3 as a p-type layer to increase the bandgap anddecrease the lattice mismatch and band offset.18 When x was increasedfrom 0 to 0.6, the bandgap increased from 3.0 to 4.2 eV, and the latticemismatch decreased from 0.3% to 0.12%. Although the decrease in theband offset for holes has not been described in the report, it shoulddecrease because the valence band maximum (VBM) decreased by�1 eV, as illustrated in Fig. 19.18 However, Mg doping was required toprovide clear p-type conduction to the a-(Ga0.6Ir0.4)2O3 layer despitethe simultaneous increase in resistivity. The behavior was not straight-forward. Further analysis is required to clarify the nature ofa-(GaxIr1-x)2O3.C. Field effect transistors (FETs)The first report on a a-Ga2O3-based FET was by Dang et al. in2015.79 The MESFET was fabricated using a mist-CVD-grown Sn-doped (0001) a-Ga2O3 layer on sapphire with a AgOx Schottky gate.The rectification ratio, VB, and on-off ratio were 6� 106, 19.6V, and2� 107, respectively. More recently, an a-Ga2O3-based MOSFET witha VB of 2.3 kV was reported.80 The MOSFET with a HfO2 gate wasfabricated using an HVPE-grown Si-doped (0001) a-Ga2O3 layer onFIG. 14. Benchmarking of the GaOTMSBDs. Data for b-Ga2O3 vertical SBDsreported in Ref. 76 were added for comparison. Reproduced with permission fromT. Shinohe, in Proceedings of the International Power Electronics Conference(IPEC-Himeji 2022-ECCE Asia), Himeji, Japan (IEEE, 2022), pp. 627–631.Copyright 2022 Institute of Electrical Engineers of Japan.33FIG. 15. Reverse recovery characteristics of an ampere-class GaOTMSBD.Reproduced with permission from T. Shinohe, in Proceedings of the InternationalPower Electronics Conference (IPEC-Himeji 2022-ECCE Asia), Himeji, Japan (IEEE,2022), pp. 627–631. Copyright 2022 Institute of Electrical Engineers of Japan.33TABLE II. Specifications of the PFC power supply board.78Item SpecificationInput voltage AC 90–242V (Max. 4A@AC 100V,2A@AC 200V)Input frequency 50–60HzBias voltage DC 12VOutput voltage Typ. DC 390V (Max. 0.9 A)Output power Max. 360W (Max. 0.9 A)Switching frequency Variable (100–240 kHz, Typ. 120 kHz)FIG. 16. Engineering samples displayed by FLOFIA at Techno-Frontier 2020 (July20–22, 2022, Tokyo, Japan). (a) GaOTMSBDs mounted in standard TO-220 pack-ages. (b) Evaluation boards driving an LED light. (c) Magnified image of the evalua-tion boards.FIG. 17. (a) Schematic of the cross section of the a-Ir2O3/a-Ga2O3 hetero-p–n-junction diode. (b) I–V characteristics of the diode. Reproduced with permission offrom Kan et al., Appl. Phys. Lett. 113, 212104 (2018). Copyright 2018 AIPPublishing.16Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-8Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/aplsapphire. The epilayer comprised a 300-nm-thick Si-doped top layerand a 900-nm-thick UID buffer layer. Ti/Al/Ni/Au was used assource and drain electrodes instead of conventional Ti/Au, and as aresult, the contact resistance was reduced by a factor of 10 or more.However, Ron was as high as 335 mX cm2, and Ec was approximately1MV/cm, which was much lower than the expected material limit(�9.5 eV). Hence, there is still much room for improvement, asshown in Fig. 20. A vertical structure should be employed to funda-mentally reduce Ron.The FETs described above are normally on devices; however,normally off devices are desirable for safety. The research group ofKyoto University and FLOFIA demonstrated normally off operationof a preliminary MOSFET with a corundum-structured p-well layer(GaOTMMOSFET)33 although the drain current was exceedingly small.The threshold gate voltage was as high as 7.9V. The device structureand output characteristics are displayed in Figs. 21 and 22, respec-tively. The channel mobility of the MOSFET was as high as 72 cm2/V s(Fig. 21),33 which was superior to that of commercially available SiCdevices.VI. SUMMARY AND FUTURE PERSPECTIVESA. Growth methodsCurrently, mist CVD and HCl-HVPE are the most maturedgrowth methods for thin and thick a-Ga2O3 films, respectively. Theprimary reason is most likely the much wider growth windows tosecure the phase purity than those in Cl2-HVPE, MBE, and MOCVD.The existence of H2 and/or H2O in the growth atmosphere is a charac-teristic of mist CVD and HCl-HVPE. These facts indicate that H2and/or H2O play an instrumental role in expanding the growth win-dow. If the mechanism is clarified, and the MBE/MOVPE process isimproved based on the mechanism, these methods could be consid-ered mainstream to grow a-Ga2O3 thin films. From the perspectives ofthe high growth rate and residual impurity concentration, HVPEwould be the best choice to grow a thick film, such as a drift layer for ahigh-VB device, whose thickness can be several ten micrometers ormore depending on the VB. Multi-wafer planetary HVPE technologyfor GaN would be applicable to a-Ga2O3 for mass production.FIG. 19. CBM and VBM positions of a-(GaxIr1-x)2O3 in terms of Ga composition x.VBM of a-Ga2O3 is considered as the standard of energy (E¼ 0). Reproduced withpermission from Kaneko et al., Appl. Phys. Lett. 118, 102104 (2021). Copyright2021 AIP Publishing.18FIG. 18. Energy-band diagrams of (a) isolated a-Ir2O3 and a-Ga2O3. (b) a-Ir2O3/a-Ga2O3 hetero junction at thermal equilibrium. Reproduced with permission offrom Kan et al., Appl. Phys. Lett. 113, 212104 (2018). Copyright 2018 AIPPublishing.16FIG. 20. Benchmarking of power transistors. Unlabeled data points are for b-Ga2O3. The a-Ga2O3 limit and the data for a-Ga2O3 MOSFET reported in Ref. 80were added. Reproduced with permission from Green et al., APL Mater. 10,029201 (2022). Copyright 2018 AIP Publishing.81FIG. 21. (a) Schematic of the normally off GaOTMMOSFET. (b) Optical micrographof the MOSFET. Reproduced with permission from https://flosfia.com/struct/wp-con-tent/uploads/79cd9d2dfa54a771f642e008cc4f9cb0.pdf for “News Release FromFLOSFIA and Kyoto University (2018).” Copyright 2018 FLOSFIA and KyotoUniversity.82Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-9Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://flosfia.com/struct/wp-content/uploads/79cd9d2dfa54a771f642e008cc4f9cb0.pdfhttps://flosfia.com/struct/wp-content/uploads/79cd9d2dfa54a771f642e008cc4f9cb0.pdfhttps://scitation.org/journal/aplB. SubstratesCurrently, a-Ga2O3 has been grown heteroepitaxially on foreignsubstrates. However, freestanding a-Ga2O3 substrates are preferable. Inthe case of GaN, the melt growth requires ultra-high pressure.Consequently, it is virtually impossible to grow a practically useful bulkcrystal for wafer fabrication, similar to the case of a-Ga2O3.Accordingly, freestanding GaN wafers are mass-produced by HVPE ofa thick GaN layer on a foreign substrate and removal of the substratefollowing the growth.83 It is likely that freestanding a-Ga2O3 wafers aredeveloped in the future using the same methodology, as depicted inFig. 23. It has been reported that a-Ga2O3 powder was grown via a fluxmethod under a high pressure of 44 kbar.84 If seeding and temperaturegradient control under such ultra-high pressure are realized, a bulka-Ga2O3 crystal could be grown to produce freestanding a-Ga2O3wafers, although it would be a considerable challenge to accomplish it.C. Defect controlAt present, it is unclear how crystal defects in a-Ga2O3, includingdislocations and point defects, affect device performance, what are thekiller defects, and how low the concentrations should be. Such an investi-gation should be conducted in mutual feedback of the improvement ofcrystal growth technologies and the analysis of the device performance.1. DislocationsAlthough the dislocation density in heteroepitaxial a-Ga2O3 canbe reduced by ELO, high-dislocation-density areas on window regionsand formation of dislocations at the coalesced boundary remain prob-lematic. However, we believe that such difficulties will eventually beovercome using the measures described in Sec. IVA. Indeed, the prob-lem of production cost needs to be considered. Note that allowable dis-location density has not been experimentally clarified yet and needs tobe explored in device development because the allowable dislocationdensity should be dependent on the device structure and operatingconditions. There seems to be a tendency that wider bandgapsemiconductors are more robust to crystal defects. For example,GaN-based laser diodes (LDs) with a dislocation density of 106 cm�3efficiently operate while the dislocation density needs to be muchlower in GaAs-based LDs. Despite the high dislocation density, therelatively high VB and excellent ideality factor of current a-Ga2O3SBDs may indicate that this tendency is applicable to a-Ga2O3.2. Deep levelsCurrently, the understanding of deep levels in a-Ga2O3 is insuffi-cient despite the practical significance. Systematic investigation isrequired using samples grown by various growth methods under vari-ous growth conditions or with post-growth treatment such as thermalannealing and irradiation of particle beams. Furthermore, the originsof the deep levels need to be clarified by comparing the experimentalresults and theoretical predictions. Influence of the deep levels on thedevice performance, allowable trap density, and how to control thedensity should also be investigated.D. Device-related issues1. p-type layerThe hetero p–n-junction of a-(IrxGa1-x)2O3/a-Ga2O3 with smalllattice mismatch is promising; however, the development is still in itsinfancy and there are many technical issues that need to be investi-gated such as carrier concentration control, point defects, and interfacequality.2. Device structure and fabrication processDevice structures of a-Ga2O3-based devices have been ratherprimitive without advanced edge-termination technologies developedfor other preceding power semiconductor devices. Consequently, suchimmature device fabrication technologies may be a primary cause oflow Ec,80 which is far from the expected value. Accordingly, the devicefabrication process needs to be further developed alongside improvingthe crystal quality. So far, it is difficult to use ion implantation, asexplained in Sec. II, and the device structure needs to be developed byepitaxial growth and etching.85–87 Fabrication of 3D-structures, suchas trenches and fins by selective area growth without plasma damage,would also be an interesting choice, as demonstrated for b-Ga2O3.883. Thermal managementSelf-heating of a power device driven under high-power-densitycondition is a serious problem that degrades device performance anddestroys the device structure. In the case of Ga2O3, the problem is moreserious than the case of SiC or GaN, primarily because the thermal con-ductivity is much lower, and higher power density is projected. Package-level thermal engineering alone is insufficient to enable efficient heatFIG. 22. (a) Output characteristics of the normally off GaOTMMOSFET. (b) Channelmobility of the MOSFET as a function of the gate voltage. Reproduced with permis-sion from T. Shinohe, in Proceedings of the International Power ElectronicsConference (IPEC-Himeji 2022-ECCE Asia), Himeji, Japan (IEEE, 2022), pp.627–631. Copyright 2022 Institute of Electrical Engineers of Japan.33FIG. 23. Fabrication of a freestanding a-Ga2O3 wafer by HVPE thick film growthand removal of the starting substrate.Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-10Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://scitation.org/journal/apldissipation to lower the junction temperature to an acceptable level undersevere conditions, and device-level engineering based on electro-thermalcodesign is essential. In the case of b-Ga2O3, it has been predicted that thejunction temperature lower than 200 �C should be possible for a lateralMOSFET driven under a power density of 10W/mm by applying variousmeasures, including the thinning of the b-Ga2O3 substrate, flip-chip het-erointegration on a high-thermal-conductivity carrier wafer, and passiv-ation using nano-crystalline diamond.89 For more details regarding thethermal management of UWBG semiconductor power devices, refer tothe excellent review by Choi et al.90In the case of a-Ga2O3, thermal management has not been com-prehensively investigated. Basically, similar strategies for b-Ga2O3should be followed. However, it is unclear whether it is possible to real-ize sufficiently low thermal resistance because of the predicted thermalconductivity being even lower than those for b-Ga2O3 and the lowthermal stability of a-Ga2O3. Fundamental thermal properties ofa-Ga2O3 need to be experimentally explored for precise simulations.E. Marketing strategiesAlthough the building blocks are in steady progress, it will takemore time to realize commercially available high-quality a-Ga2O3templates or freestanding substrates for the fabrication of power devi-ces involving high-voltage operation. Accordingly, it would be a goodstrategy to initiate preliminary development in low to middle voltagedevice applications (<600V), in which crystal defects would not haveserious impact on the device performance. In this application area,including AC adopters, data server power supplies, and home applian-ces, a-Ga2O3 devices should have cost competitiveness, even to Si-based devices, because it should be possible to considerably reduce theepilayer thickness and device chip area owing to the higher Ec andlower Ron. In the future, it would also be possible to penetrate high-voltage-use or cost-effective-use market by solving the related techni-cal problems and optimizing the mass-production technologies duringthe preliminary production period.Considering the superior material properties of a-Ga2O3 and theprogress of various technologies to compensate the drawbacks, such ashigh dislocation density and the difficulty in p-type conduction, webelieve that a-Ga2O3-based devices occupy a certain position in the futurepower device market. The race has just started, and there are still manychallenging technical problems to be overcome; however, it is worth a try.ACKNOWLEDGMENTSThis work was supported by the Air Force Office of ScientificResearch (Program Manager, Dr. Ali Sayir) through Program No.FA9550–20-1–0045 and the National Science Foundation underGrant No. 2043803.AUTHOR DECLARATIONSConflict of InterestThe authors have no conflicts to disclose.Author ContributionsYuichi Oshima: Conceptualization (equal); Writing – original draft(lead). Elaheh Ahmadi: Conceptualization (equal); Writing – review &editing (lead).DATA AVAILABILITYData sharing is not applicable to this article as no new data werecreated or analyzed in this study.REFERENCES1M. Higashiwaki, K. Sasaki, A. Kuramata, T. Masui, and S. Yamakoshi, Appl.Phys. Lett. 100, 013504 (2012).2B. J. Baliga, Fundamentals of Power Semiconductor Devices (Springer,2008).3M. Higashiwaki, Gallium Oxide-Materials Properties, Crystal Growth, andDevices (Springer, 2020), Chap. 1.5.2.4F. Bechstedt and J. Furthmuller, Appl. Phys. Lett. 114, 122101 (2019).5K. Akaiwa, K. Ota, T. Sekiyama, T. Abe, T. Shinohe, and K. Ichino, Phys.Status Solidi A 217, 1900632 (2020).6J. Furthmuller and F. Bechstedt, Phys. Rev. B 93, 115204 (2016).7T. Kobayashi, T. Gake, Y. Kumagai, F. Oba, and Y. Matsushita, Appl. Phys.Express 12, 091001 (2019).8M. Biswas and H. Nishinaka, APL Mater. 10, 060701 (2022).9R. Roy, V. G. Hill, and E. F. Osborn, J. Am. Chem. Soc. 74, 719 (1952).10H. Y. Playford, A. C. Hannon, E. R. Barney, and R. I. Walton, Chem. Eur. J. 19,2803 (2013).11K. Kaneko, S. Fujita, and T. Hitora, Jpn. J. Appl. Phys., Part 1 57, 02CB18(2018).12S. Fujita and K. Kaneko, J. Cryst. Growth 401, 588–592 (2014).13R. Jinno, C. S. Chang, T. Onuma, Y. Cho, S.-T. Ho, D. Rowe, M. C. Cao, K. Lee,V. Protasenko, D. G. Schlom, D. A. Muller, H. G. Xing, and D. Jena, Sci. Adv.7, eabd5891 (2021).14A. F. M. A. U. Bhuiyan, Z. Feng, H.-L. Huang, L. Meng, J. Hwang, and H.Zhao, APL Mater. 9, 101109 (2021).15K. Kaneko, T. Nomura, I. Kakeya, and S. Fujita, Appl. Phys. Express 2, 075501(2009).16S. Kan, S. Takemoto, K. Kaneko, I. Takahashi, M. Sugimoto, T. Shinohe, and S.Fujita, Appl. Phys. Lett. 113, 212104 (2018).17J. G. Hao, H. H. Gong, X. H. Chen, Y. Xu, F. Ren, S. L. Gu, R. Zhang, Y. D.Zheng, and J. D. Ye, Appl. Phys. Lett. 118, 261601 (2021).18K. Kaneko, Y. Masuda, S. Kan, I. Takahashi, Y. Kato, T. Shinohe, and S. Fujita,Appl. Phys. Lett. 118, 102104 (2021).19J. Zhang, P. Dong, K. Dang, Y. Zhang, Q. Yan, H. Xiang, J. Su1, Z. Liu, M. Si, J.Gao, M. Kong, H. Zhou, and Y. Hao, Nat. Commun. 13, 3900 (2022).20E. G. V�ıllora, K. Shimamura, K. Kitamura, and K. Aoki, Appl. Phys. Lett. 88,031105 (2006).21Y. Oshima, E. G. V�ıllora, and K. Shimamura, J. Cryst. Growth 410, 53–58(2015).22S. Lee, Y. Ito, K. Kaneko, and S. Fujita, Jpn. J. Appl. Phys., Part 1 54, 030301(2015).23F. Massabuau, D. Nicol, F. Adams, J. Jarman, J. Roberts, A. Kovacs, P. Chalker,and R. Oliver, J. Phys. D 54, 384001 (2021).24Y. Yao, R. F. Davis, and L. M. Porter, J. Electron. Mater. 46, 2053 (2017).25R. Jinno, K. Kaneko, and S. Fujita, AIP Adv. 10, 115013 (2020).26D. Machon, P. F. McMillan, B. Xu, and J. Dong, Phys. Rev. B 73, 094125(2006).27J. P. McCandless, C. S. Chang, K. Nomoto, J. Casamento, V. Protasenko, P.Vogt, D. Rowe, K. Gann, S. T. Ho, W. Li, R. Jinno, Y. Cho, A. J. Green, K. D.Chabak, D. G. Schlom, M. O. Thompson, D. A. Muller, H. G. Xing, and D.Jena, Appl. Phys. Lett. 119, 062102 (2021).28R. Jinno, K. Kentaro, and S. Fujita, Jpn. J. Appl. Phys., Part 1 60, SBBD13(2021).29G. Yang, P. R. Romeo, A. Apostoluk, and B. Vilquina, J. Vac. Sci. Technol. A40, 052801 (2022).30Y. Oshima, T. Suzuki, T. Eri, Y. Kawaguchi, K. Watanabe, M. Shibata, and T.Mishima, J. Appl. Phys. 98, 103509 (2005).31R. Wei, S. Song, K. Yang, Y. Cui, Y. Peng, X. Chen, X. Hu, and X. Xu, J. Appl.Phys. 113, 053503 (2013).32Z. Guo, A. Verma, X. Wu, F. Sun, A. Hickman, T. Masui, A. Kuramata, M.Higashiwaki, D. Jena, and T. Luo, Appl. Phys. Lett. 106, 111909 (2015).Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-11Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://doi.org/10.1063/1.3674287https://doi.org/10.1063/1.3674287https://doi.org/10.1063/1.5084324https://doi.org/10.1002/pssa.201900632https://doi.org/10.1002/pssa.201900632https://doi.org/10.1103/PhysRevB.93.115204https://doi.org/10.7567/1882-0786/ab3763https://doi.org/10.7567/1882-0786/ab3763https://doi.org/10.1063/5.0085360https://doi.org/10.1021/ja01123a039https://doi.org/10.1002/chem.201203359https://doi.org/10.7567/JJAP.57.02CB18https://doi.org/10.1016/j.jcrysgro.2014.02.032https://doi.org/10.1126/sciadv.abd5891https://doi.org/10.1063/5.0065087https://doi.org/10.1143/APEX.2.075501https://doi.org/10.1063/1.5054054https://doi.org/10.1063/5.0047710https://doi.org/10.1063/5.0027297https://doi.org/10.1038/s41467-022-31664-yhttps://doi.org/10.1063/1.2164407https://doi.org/10.1016/j.jcrysgro.2014.10.038https://doi.org/10.7567/JJAP.54.030301https://doi.org/10.1088/1361-6463/ac0d28https://doi.org/10.1007/s11664-016-5121-1https://doi.org/10.1063/5.0020464https://doi.org/10.1103/PhysRevB.73.094125https://doi.org/10.1063/5.0064278https://doi.org/10.35848/1347-4065/abde25https://doi.org/10.1116/6.0001870https://doi.org/10.1063/1.2132508https://doi.org/10.1063/1.4790134https://doi.org/10.1063/1.4790134https://doi.org/10.1063/1.4916078https://scitation.org/journal/apl33T. Shinohe, in Proceedings of the International Power Electronics Conference(IPEC-Himeji 2022-ECCE Asia), Himeji, Japan (IEEE, 2022), pp. 627–631.34D. Shinohara and S. Fujita, Jpn. J. Appl. Phys., Part 1 47, 7311–7313 (2008).35R. Jinno, N. Yoshimura, K. Kaneko, and S. Fujita, Jpn. J. Appl. Phys., Part 1 58,120912 (2019).36R. Jinno, T. Uchida, K. Kaneko, and S. Fujita, Appl. Phys. Express 9, 071101(2016).37K. Uno, M. Ohta, and I. Tanaka, Appl. Phys. Lett. 117, 052106 (2020).38Y. Oshima, E. G. Villora, and K. Shimamura, Appl. Phys. Express 8, 055501(2015).39Y. Oshima, K. Kawara, T. Oshima, M. Okigawa, and T. Shinohe, Semicond.Sci. Technol. 35, 055022 (2020).40K. Goto, H. Nakahata, H. Murakami, and Y. Kumagai, Appl. Phys. Lett. 117,222101 (2020).41R. Togashi, S. Numata, M. Hayashida, T. Suga, K. Goto, A. Kuramata, S.Yamakoshi, P. Paskov, B. Monemar, and Y. Kumagai, Jpn. J. Appl. Phys., Part 155, 1202B3 (2016).42Y. Kumagai, T. Yamane, and A. Kokitu, J. Cryst. Growth 281, 62–67 (2005).43J. B. Varley, J. R. Weber, A. Janotti, and C. G. Van de Walle, Appl. Phys. Lett.97, 142106 (2010).44Z. Cheng, M. Hanke, P. Vogt, O. Bierwagen, and A. Trampert, Appl. Phys.Lett. 111, 162104 (2017).45M. Kracht, A. Karg, M. Feneberg, J. Bl€asing, J. Sch€ormann, R. Goldhahn, andM. Eickhoff, Phys. Rev. Appl. 10, 024047 (2018).46E. Ahmadi and Y. Oshima, J. Appl. Phys. 126, 160901 (2019).47Z. Feng, A. F. M. A. U. Bhuiyan, M. R. Karim, and H. Zhao, Appl. Phys. Lett.114, 250601 (2019).48M. J. Tadjer, F. Alema, A. Osinsky, M. A. Mastro, N. Nepal, J. M. Woodward,R. L. Myers-Ward, E. R. Glaser, J. A. Freitas, Jr., J. Hajzus, M. Ebrish, T. J.Anderson, K. D. Hobart, J. K. Hite, and C. R. Eddy, Jr., J. Phys. D 54, 034005(2021).49D. D. Koleske, A. E. Wickenden, R. L. Henry, and M. E. Twigg, J. Cryst.Growth 242, 55–69 (2002).50K. Ikenaga, N. Tanaka, T. Nishimura, H. Iino, K. Goto, M. Ishikawa, H.Machida, T. Ueno, and Y. Kumagai, J. Cryst. Growth 582, 126520 (2022).51Y. Arata, H. Nishinaka, K. Shimazoe, and M. Yoshimoto, J. Soc. Mater. Sci.,Jpn. 70, 738–744 (2021).52K. Shimazoe, H. Nishinaka, Y. Arata, D. Tahara, and M. Yoshimoto, AIP Adv.10, 055310 (2020).53T. Kawaharamura, G. T. Dang, and M. Furuta, Jpn. J. Appl. Phys., Part 1 51,040207 (2012).54T. Uchida, K. Kaneko, and S. Fujita, MRS Adv. 3(3), 171–177 (2018).55S. Morimoto, H. Nishinaga, and M. Yoshimoto, Thin Solid Films 682, 18–23(2019).56H. Son, Y. Choi, J. Park, B. Ryu, and D. Jeon, ECS J. Solid State Sci. Technol. 9,055005 (2020).57Y. Oshima, K. Kawara, M. Kasu, T. Shinohe, and T. Hitora, in 60th ElectronicMaterials Conference (EMC), Santa Barbara, 2018.58T. C. Ma, X. H. Chen, Y. Kuang, L. Li, J. Li, F. Kremer, F.-F. Ren, S. L. Gu, R.Zhang, Y. D. Zheng, H. H. Tan, C. Jagadish, and J. D. Ye, Appl. Phys. Lett.115, 182101 (2019).59H. Takane, H. Izumi, T. Wakamatsu, K. Tanaka, and K. Kaneko, in 69th JSAPSpring Meeting (Sophia University, Tokyo, Japan, 2022).60J. B. Varley, Gallium Oxide -Materials Properties, Crystal Growth, and Devices(Springer, 2020), Chap. 18.3.5.61K. Kaneko, H. Kawanowa, H. Ito, and S. Fujita, Jpn. J. Appl. Phys., Part 1 51,11PJ03 (2012).62Y. Oshima, K. Kawara, T. Shinohe, T. Hitora, M. Kasu, and S. Fujita, APLMater. 7, 022503 (2019).63H. Son, Y. Choi, S. Hong, J. Parka, and D. Jeon, IUCrJ 8, 462–467 (2021).64H. Son, Y. Choi, J. Ha, S. Jung, and D. Jeon, Cryst. Growth Des. 19, 5105(2019).65K. Kawara, Y. Oshima, M. Okigawa, and T. Shinohe, Appl. Phys. Express 13,075507 (2020).66Y. Oshima, S. Yagyu, and T. Shinohe, J. Appl. Phys. 130, 175304 (2021).67Y. Oshima, S. Yagyu, and T. Shinohe, J. Cryst. Growth 576, 126387 (2021).68K. Fujito, S. Kubo, H. Nagaoka, T. Mochizuki, H. Namita, and S. Nagao,J. Cryst. Growth 311, 3011–3014 (2009).69A. Y. Polyakov, N. B. Smirnov, I. V. Shchemerov, E. B. Yakimov, V. I.Nikolaev, S. I. Stepanov, A. I. Pechnikov, A. V. Chernykh, K. D. Shcherbachev,A. S. Shikoh, A. Kochkova, A. A. Vasilev, and S. J. Pearton, APL Mater. 7,051103 (2019).70D. Jeon, H. Son, J. Hwang, A. Y. Polyakov, N. B. Smirnov, I. V. Shchemerov,A. V. Chernykh, A. I. Kochkova, S. J. Pearton, and I.-H. Lee, APL Mater. 6,121110 (2018).71H. Takane, K. Kaneko, T. Shinohe, and S. Fujita, Phys. Status Solidi B 258,2000622 (2021).72A. Y. Polyakov, V. I. Nikolaev, E. B. Yakimov, F. Ren, S. J. Pearton, and J. Kim,J. Vac. Sci. Technol. A 40, 020804 (2022).73M. Oda, R. Tokuda, H. Kambara, T. Tanikawa, T. Sasaki, and T. Hitora, Appl.Phys. Express 9, 021101 (2016).74D. Yang, B. Kim, J. Oh, T. H. Lee, J. Ryu, S. Park, S. Kim, E. Yoon, Y. Park, andH. Jang, ACS Appl. Mater. Interfaces 14, 5598–5607 (2022).75T. Maeda, M. Okigawa, Y. Kato, I. Takahashi, and T. Shinohe, AIP Adv. 10,125119 (2020).76J. Yang, C. Fares, R. Elhassani, M. Xian, F. Ren, S. J. Pearton, M. Tadjer, and A.Kuramata, ECS J. Solid State Sci. Technol. 8, Q3159 (2019).77See https://www.kyoei.co.jp/news/2020/20200626.html for “Website of KyoeiSangyo Co., Ltd.”78See https://www.kyoei.co.jp/news/2020/20200626.html for “News ReleaseFrom Kyoei Sangyo Co., Ltd. (2020).”79G. T. Dang, T. Kawaharamura, M. Furuta, and M. W. Allen, IEEE Trans.Electron Devices 62, 3640–3644 (2015).80Y. Jeong, J. Park, M. Yeom, I. Kang, J. Yang, H. Kim, D. Jeon, and G. Yoo,Appl. Phys. Express 15, 074001 (2022).81A. J. Green, J. Speck, G. Xing, P. Moens, F. Allerstam, K. Gumaelius, T. Neyer,A. Arias-Purdue, V. Mehrotra, A. Kuramata, K. Sasaki, S. Watanabe, K. Koshi,J. Blevins, O. Bierwagen, S. Krishnamoorthy, K. Leedy, A. R. Arehart, A. T.Neal, S. Mou, S. A. Ringel, A. Kumar, A. Sharma, K. Ghosh, U. Singisetti, W.Li, K. Chabak, K. Liddy, A. Islam, S. Rajan, S. Graham, S. Choi, Z. Cheng, andM. Higashiwaki, APL Mater. 10, 029201 (2022).82See https://flosfia.com/struct/wp-content/uploads/79cd9d2dfa54a771f642e008cc4f9cb0.pdf for “News Release From FLOSFIA and Kyoto University(2018).”83Y. Oshima, T. Eri, M. Shibata, H. Sunakawa, K. Kobayashi, T. Ichihashi, and A.Usui, Jpn. J. Appl. Phys., Part II 42, L1–L3 (2003).84J. P. Remeika and M. Marezio, Appl. Phys. Lett. 8, 87–88 (1966).85Z. A. Jian, Y. Oshima, S. Wright, K. Owen, and E. Ahmadi1, Semicond. Sci.Technol. 34, 035006 (2019).86J. Um, B. Choi, D. Jeong, H. Choi, S. Hwang, D. Jeon, and J. Kim, Electron.Mater. Lett. 17, 142–147 (2021).87J. H. Um, B. S. Choi, W. S. Jang, S. Hwang, D.-W. Jeon, J. K. Kim, and H. Cho,Korean J. Met. Mater. 59, 121–126 (2021).88T. Oshima and Y. Oshima, Appl. Phys. Express 15, 075503 (2022).89B. Chatterjee, K. Zeng, C. D. Nordquist, U. Singisetti, and S. Choi, IEEE Trans.Compon., Packag. Manuf. Technol. 9, 2352 (2019).90S. Choi, S. Graham, S. Chowdhury, E. R. Heller, M. J. Tadjer, G. Moreno, andS. Narumanchi, Appl. Phys. Lett. 119, 170501 (2021).Applied Physics Letters PERSPECTIVE scitation.org/journal/aplAppl. Phys. Lett. 121, 260501 (2022); doi: 10.1063/5.0126698 121, 260501-12Published under an exclusive license by AIP Publishing 29 January 2024 06:12:56https://doi.org/10.1143/JJAP.47.7311https://doi.org/10.7567/1347-4065/ab55c6https://doi.org/10.7567/APEX.9.071101https://doi.org/10.1063/5.0014056https://doi.org/10.7567/APEX.8.055501https://doi.org/10.1088/1361-6641/ab7843https://doi.org/10.1088/1361-6641/ab7843https://doi.org/10.1063/5.0031267https://doi.org/10.7567/JJAP.55.1202B3https://doi.org/10.1016/j.jcrysgro.2005.03.011https://doi.org/10.1063/1.3499306https://doi.org/10.1063/1.4998804https://doi.org/10.1063/1.4998804https://doi.org/10.1103/PhysRevApplied.10.024047https://doi.org/10.1063/1.5123213https://doi.org/10.1063/1.5109678https://doi.org/10.1088/1361-6463/abbc96https://doi.org/10.1016/S0022-0248(02)01348-9https://doi.org/10.1016/S0022-0248(02)01348-9https://doi.org/10.1016/j.jcrysgro.2022.126520https://doi.org/10.2472/jsms.70.738https://doi.org/10.2472/jsms.70.738https://doi.org/10.1063/5.0006137https://doi.org/10.1143/JJAP.51.040207https://doi.org/10.1557/adv.2018.45https://doi.org/10.1016/j.tsf.2019.04.051https://doi.org/10.1149/2162-8777/ab96achttps://doi.org/10.1063/1.5120554https://doi.org/10.7567/JJAP.51.11PJ03https://doi.org/10.1063/1.5051058https://doi.org/10.1063/1.5051058https://doi.org/10.1107/S2052252521003389https://doi.org/10.1021/acs.cgd.9b00454https://doi.org/10.35848/1882-0786/ab9fc5https://doi.org/10.1063/5.0068097https://doi.org/10.1016/j.jcrysgro.2021.126387https://doi.org/10.1016/j.jcrysgro.2009.01.046https://doi.org/10.1063/1.5094787https://doi.org/10.1063/1.5075718https://doi.org/10.1002/pssb.202000622https://doi.org/10.1116/6.0001701https://doi.org/10.7567/APEX.9.021101https://doi.org/10.7567/APEX.9.021101https://doi.org/10.1021/acsami.1c21845https://doi.org/10.1063/5.0028985https://doi.org/10.1149/2.0211907jsshttps://www.kyoei.co.jp/news/2020/20200626.htmlhttps://www.kyoei.co.jp/news/2020/20200626.htmlhttps://doi.org/10.1109/TED.2015.2477438https://doi.org/10.1109/TED.2015.2477438https://doi.org/10.35848/1882-0786/ac7431https://doi.org/10.1063/5.0060327https://flosfia.com/struct/wp-content/uploads/79cd9d2dfa54a771f642e008cc4f9cb0.pdfhttps://flosfia.com/struct/wp-content/uploads/79cd9d2dfa54a771f642e008cc4f9cb0.pdfhttps://doi.org/10.1143/JJAP.42.L1https://doi.org/10.1063/1.1754500https://doi.org/10.1088/1361-6641/aafeb2https://doi.org/10.1088/1361-6641/aafeb2https://doi.org/10.1007/s13391-020-00267-4https://doi.org/10.1007/s13391-020-00267-4https://doi.org/10.3365/KJMM.2021.59.2.121https://doi.org/10.35848/1882-0786/ac75c8https://doi.org/10.1109/TCPMT.2019.2923356https://doi.org/10.1109/TCPMT.2019.2923356https://doi.org/10.1063/5.0056271https://scitation.org/journal/apl