# Fileset

[jjap_65_13_138001.pdf](https://mdr.nims.go.jp/filesets/b8d85125-ea02-47bd-a7b3-5e45615cfb66/download)

## Creator

[Takayoshi Oshima](https://orcid.org/0000-0001-8550-9735), [Yuichi Oshima](https://orcid.org/0000-0001-8293-4891)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[(110)- and (−134)-faceted surface morphology of halide vapor phase epitaxy-grown                    <i>β-</i>                    Ga                    <sub>2</sub>                    O                    <sub>3</sub>                    homoepitaxial layers on (011) substrates](https://mdr.nims.go.jp/datasets/7f0ea44e-cf8e-43ed-ad27-c73280ce3fbf)

## Fulltext

(110)- and (−134)-faceted surface morphology of halide vapor phase epitaxy-grown β-Ga2O3 homoepitaxial layers on (011) substratesaaa(110)- and (−134)-faceted surface morphology of halide vapor phaseepitaxy-grown β-Ga2O3 homoepitaxial layers on (011) substratesTakayoshi Oshima* and Yuichi OshimaResearch Center for Electronic and Optical Materials, National Institute for Materials Science, Tsukuba, Ibaraki 305-0044, Japan*E-mail: OSHIMA.Takayoshi@nims.go.jpReceived May 25, 2026; revised June 23, 2026; accepted June 29, 2026; published online July 8, 2026The surface morphology of the homoepitaxial layer grown on a (011) β-Ga2O3 substrate by HCl-based halide vapor phase epitaxy wasinvestigated using atomic force microscopy (AFM) and transmission electron microscopy (TEM). The surface consisted of (110) and (−134) facetselongated along [1−11], with facet coverage ratios of 0.207:0.793 (AFM) and 0.210:0.790 (TEM), in close agreement with the geometricallyexpected ratio of 0.219:0.781. Thus, the surface can be regarded as being composed of (110) macrosteps and (−134) terraces. These findingsindicate that growth kinetics and impurity incorporation should be discussed based on the actual (110) and (−134) faceted morphology, ratherthan the nominal (011) substrate orientation. © 2026 The Author(s). Published on behalf of The Japan Society of Applied Physics by IOPPublishing Ltdβ-Ga2O3 has gained significant attention as a promisingultrawide-bandgap semiconductor for high-power applica-tions.1) This interest is primarily driven by its breakdownelectric field, which is estimated to be approximately8MV cm−1, resulting in a Baliga’s figure of merit signifi-cantly higher than those of wide-bandgap SiC and GaN.2)Furthermore, the availability of high-quality, large-areasubstrates up to 6 inches in diameter, enabled by high-speedmelt growth methods, provides a distinct cost advantage forindustrial scalability.1,3,4)In high-voltage and high-current applications, verticaldevice architectures are preferred, requiring thick epitaxialdrift layers with low impurity concentrations. Halide vaporphase epitaxy (HVPE) has become a key technique forforming such layers due to its high growth rates and ability toproduce high-purity epitaxial films.5–10) Accordingly,HVPE-grown epiwafers have been widely used as platformsfor demonstrating vertical high-power devices, includingSchottky barrier diodes (SBDs),11) metal-oxide-semicon-ductor field-effect transistors (MOSFETs),12) and hetero-junction p-n diodes.13)The (001)-oriented plane has primarily been chosen forepiwafers due to its relatively high growth rate in HVPE.8)However, (001) surfaces often suffer from significant mor-phological issues, such as deep pits or streaky grooveselongated along the [010] direction.8) These pits, whichoriginate from dislocations propagating from the substrate,often require costly chemical mechanical polishing (CMP)before device fabrication.7,14) Notably, nearly half of theepilayer thickness may need to be removed by CMP toplanarize the surface.14) Additionally, chlorine (Cl), whichacts as a shallow donor,15) is readily incorporated into (001)epitaxial layers during HVPE, making it difficult to achievelow-donor concentrations less than 8 × 1015 cm−3.16)Recently, the (011) plane has emerged as a superioralternative to overcome these morphological and Cl-incor-poration issues. Goto et al. discovered that a pit-freehomoepitaxial surface can be achieved on (011) substratesby Cl2-based HVPE, in which GaCl precursor is produced bythe reaction between Ga and Cl2, because the (011) plane isnearly parallel to the dislocations that generate surface pitson the (001) plane.8) This suppresses the inheritance ofsubstrate defects into the epitaxial layer. Although thegrowth rate of Cl2-based HVPE is limited to approximately2.8 μm h−1 to prevent the deposition of particles produced byparasitic reactions,8) the rate can be increased to approxi-mately 14 μm h−1 using HCl-based HVPE, in which GaClprecursor is produced by the reaction between Ga and HCl,together with the additional introduction of HCl etching gasto suppress parasitic reactions.9,17,18) However, this addi-tional HCl supply in HCl-based HVPE induces slit-like pitsthat are not observed in Cl2-based HVPE,9) and thus the HClflow rate must be optimized to suppress both particledeposition and pit formation. Moreover, for both Cl2- andHCl-based HVPE, the (011) orientation effectively sup-presses Cl incorporation, enabling low Cl concentrations ofaround 1 × 1015 cm−3, roughly one order of magnitudelower than those typically measured for the (001)orientation.9,16)These advantages have already led to high-performancedevice demonstrations on (011) epiwafers, including verticalmulti-fin MOSFETs with breakdown voltages exceeding 10kV16) and vertical SBDs reaching 3.7 kV with ultra-lowleakage current.19) Furthermore, other studies related to the(011) orientation, such as bulk crystal growth,20) metal–organic chemical vapor epitaxy,21) plasma-free HCl-gasetching,22) and killer-defect characterization,23) have beenreported, highlighting the growing interest in the (011)orientation as a promising platform for β-Ga2O3 epitaxyand device fabrication.In this study, we focus on the surface morphology ofHVPE-grown (011) homoepitaxial layers. In our previousstudy, we observed that the surface exhibited step-bunchedmacrosteps and terraces rather than an atomically flatmorphology, even in pit-free flat areas, as evidenced byscanning electron microscopy (SEM) observations [Fig. 1].9)This observation suggests that crystallographic facets otherthan the (011) plane develop on the surface. However, thefaceted structures were not investigated in detail. Therefore,in this study, we further investigated the surface morphologyContent from this work may be used under the terms of the Creative Commons Attribution 4.0 license. Any further distribution ofthis work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.138001-1© 2026 The Author(s). Published on behalf ofThe Japan Society of Applied Physics by IOP Publishing LtdJapanese Journal of Applied Physics 65, 138001 (2026) BRIEF NOTEhttps://doi.org/10.35848/1347-4065/ae83f5https://crossmark.crossref.org/dialog/?doi=10.35848/1347-4065/ae83f5&domain=pdf&date_stamp=2026-07-08https://orcid.org/0000-0001-8550-9735https://orcid.org/0000-0001-8293-4891mailto:OSHIMA.Takayoshi@nims.go.jphttps://creativecommons.org/licenses/by/4.0/https://doi.org/10.35848/1347-4065/ae83f5using nanometer-scale characterization techniques, includingatomic force microscopy (AFM) and transmission electronmicroscopy (TEM), to identify the facets forming themacrosteps and terraces.For this investigation, we used a 3.6 μm thick homoepitaxiallayer grown on the (011) substrate using HCl-based HVPE,with detailed growth conditions reported previously.9) The layerwas single-crystalline, and its tilt and twist spreads werecomparable to those of the substrate. Furthermore, unintentionalCl incorporation was sufficiently low, with a concentration of1.7 × 1015 cm−3. Therefore, this homoepitaxial layer isconsidered suitable for power device applications.The surface morphology of the homoepitaxial layer was firstexamined by AFM, as shown in Fig. 2. The surface consisted ofmacrosteps and terraces elongated along the [1 1̄ 1] direction.Owing to the presence of macrostep-and-terrace structures, theroot-mean-square (RMS) roughness values were somewhatlarge for a homoepitaxial layer, with values of 8.0 and7.2 nm for 3 μm × 3 μm and 1 μm × 1 μm scan areas,respectively [Figs. 2(a) and 2(b), respectively]. In contrast, themacrostep and terrace surfaces were relatively smooth, withlocal RMS roughness values of 0.2–0.6 nm and 0.1–0.5 nm,respectively, indicating the development of well-defined crys-tallographic facets.To estimate the possible macrostep and terrace facets, aheight profile perpendicular to the [111] direction wasextracted from Fig. 2(b), as shown in Fig. 2(c). Linear fittingof the corresponding regions yielded inclination angles of23° and 7° for the macrosteps and terraces, respectively.Although these inclination angles may include uncertaintiesdue to the finite cantilever-tip size and slight scan-trackingdeviations, they are still useful as screening indicators fornarrowing down possible facet orientations.The possible macrostep and terrace planes were consid-ered as follows. Assuming that the intersection line betweenthe macrostep and terrace planes is parallel to the [111]direction, these (hkl) planes can be regarded as zone planesbelonging to the [111] zone axis. Thus, they satisfy the zonelaw, h − k + l = 0, or equivalently, k = h + l. Among therelatively low-index planes satisfying this relationship in thevicinity of (011), (110) and (132) are possible step-planecandidates, with calculated inclination angles of 28.27° and8.74°, respectively; whereas (134) and (112) are possibleterrace-plane candidates, with calculated inclination anglesof 7.66° and 19.73°, respectively. Here, the lattice para-meters used for the calculations are a = 12.214 Å,b = 3.0371 Å, c = 5.7981 Å, α = 90°, β = 103.83°, and γ =90°.24) The selection of these candidate planes can beintuitively understood from the TEM diffraction patternshown later [Fig. 3(c)]. Based on their consistency with theexperimentally observed inclination angles, the (110) and(134) planes are considered to be the most plausiblecandidates for the macrostep and terrace facets, respectively.Additionally, the surface coverage ratio of the macrostepand terrace facets was calculated by analyzing the AFMtopography data. Figure 2(d) shows a histogram of the localsurface inclination angles over the entire scan area shown inFig. 2(b). The histogram exhibited two peaks at 23.1° and7.1°, consistent with the inclination angles of the macrostepsand terraces obtained from the line profile. The corre-sponding peak areas (blue and red regions), obtained aftersubtraction of the linear background, were used to estimatethe respective areas projected onto the (011) plane. Theprojected area (Apro) can be converted into the actual surfacearea (Aact) using Aact = Apro/cos θ, where θ is the inclinationangle. The resulting actual surface coverage ratio of themacrosteps and terraces was determined to be 0.207:0.793.To further examine the macrostep and terrace planes, weanalyzed the cross-sectional surface profile of the homoepitaxiallayer using TEM, as shown in Fig. 3. Figures 3(a) and 3(b) arebright-field TEM images acquired with low and high magnifica-tions, respectively. The electron incidence direction was [111],and the acceleration voltage was 200 kV. The observed macro-step and terrace profiles were linear, indicating that these facetswere flat.Figure 3(c) shows the corresponding electron diffractionpattern, in which each hkl diffraction spot represents thedirection normal to the (hkl) plane. By comparing the normaldirections of the observed macrostep and terrace facets in theTEM image with the directions of the hkl diffraction spots,their facets could be assigned to (110) and (134), respec-tively. Here, the angular deviation between the step-surfacenormal and the 110-diffraction spot was ~0.7°, whereasthe angular deviation between the terrace-surface normal andthe 134-diffraction spot was ~0.8°. These assignments areconsistent with the facet candidates inferred from the AFManalysis.The (110) and equivalent (110) facets are known to appear onthe surfaces of homoepitaxial layers grown on (010)substrates,25) as the (110) surface is energetically more stablethan the (010) surface.26) In addition, smooth homoepitaxialgrowth of (110) β-Ga2O3 with an RMS roughness of 0.4 nm hasbeen reported.27) Thus, the development of the (110) facet on ourepilayer surface is reasonable. In contrast, the (134) facet has notyet been reported in β-Ga2O3 studies. Because the (134) plane isa high-index vicinal plane, its surface atomic structure, includingpossible surface terminations and dangling-bond configurations,is not straightforwardly defined. Therefore, further theoreticalstudies, including detailed surface-energy calculations, arerequired to clarify the origin of its stabilization.The surface coverage ratio of the (110) and (134) facetswas also evaluated from the cross-sectional TEM image. Thewidths of the macrostep and terrace regions in Fig. 3(a) wereseparately summed and compared, using seven step–terraceFig. 1. Plan-view SEM image of a homoepitaxial layer grown on a(011) β-Ga2O3 substrate.138001-2© 2026 The Author(s). Published on behalf ofThe Japan Society of Applied Physics by IOP Publishing LtdJpn. J. Appl. Phys. 65, 138001 (2026) BRIEF NOTEFig. 2. AFM images of the homoepitaxial layer grown on the (011) β-Ga2O3 substrate, acquired over scan areas of (a) 3 μm × 3 μm and(b) 1 μm × 1 μm. These topographic images were leveled over the entire area. (c) Height profile taken along the arrow indicated in (b). (d) Relativefrequency of local inclination angles obtained from (b).Fig. 3. Cross-sectional TEM images of the homoepitaxial layer on the (011) β-Ga2O3 substrate viewed along the [111] direction, acquired at differentmagnifications: (a) low-magnification image and (b) high-magnification image. Note that hkl* denotes the direction normal to the (hkl) plane. The sevenstep–terrace pairs indicated by a curly brace in (a) were used for the calculation of the step–terrace surface coverage ratio (see the text for details).(c) Electron diffraction pattern acquired from the same specimen orientation as that used for the TEM observations.138001-3© 2026 The Author(s). Published on behalf ofThe Japan Society of Applied Physics by IOP Publishing LtdJpn. J. Appl. Phys. 65, 138001 (2026) BRIEF NOTEpairs for which both ends were fully contained within theimage. The surface coverage ratio was determined to be0.210:0.790.Finally, to examine the validity of the surface coverageratios experimentally determined from the AFM and TEMresults, we compared these values with the geometricallyexpected ratio. As shown in Fig. 4, we considered atriangular cross section bounded by the (011), (110), and(134) planes. Based on a simple trigonometric calculation,the length ratio of the (110) and (134) segments wascalculated to be 0.219:0.781, which agrees well with theexperimentally determined values of 0.207:0.793 (AFM) and0.210:0.790 (TEM), supporting the validity of our facetidentification.In conclusion, we found that the homoepitaxial surface onthe (011) β-Ga2O3 substrate was characterized by (110)macrosteps and (134) terraces elongated along the [111]direction, with facet coverage ratios of 0.207:0.793 (AFM)and 0.210:0.790 (TEM), in good agreement with the theoreticalratio of 0.219:0.781. This faceted morphology is expected toinfluence growth kinetics and impurity incorporation, includingCl incorporation, during HVPE growth. These findings there-fore provide useful insights into the morphology and growthbehavior of HVPE-grown (011) β-Ga2O3 homoepitaxial layers.Furthermore, the present results suggest that facet-level flatnessmay also be attainable in HVPE-grown homoepitaxial layers on(110)- and (134)-oriented substrates, offering a promising routetoward CMP-free epiwafers.AcknowledgmentsThis work was supported by the Nanofabrication Unit of theNational Institute for Materials Science (NIMS) through theAdvanced Research Infrastructure for Materials andNanotechnology (ARIM), which is supported by the Ministryof Education, Culture, Sports, Science and Technology(MEXT), Japan (No. JPMXP1226NM5028). This work wasalso financially supported through a project commissioned bythe New Energy and Industrial Technology DevelopmentOrganization (NEDO), under the Ministry of Economy, Tradeand Industry (METI), Japan (No. JPNP22007).ORCID iDsTakayoshi Oshima https://orcid.org/0000-0001-8550-9735Yuichi Oshima https://orcid.org/0000-0001-8293-48911) K. Sasaki, Appl. Phys. Express 17, 090101 (2024).2) M. Higashiwaki, K. Sasaki, A. Kuramata, T. Masui, and S. Yamakoshi,Appl. Phys. Lett. 100, 013504 (2012).3) T. Igarashi, Y. Ueda, K. Koshi, R. Sakaguchi, S. Watanabe, S. Yamakoshi,and A. Kuramata, Phys. Status Solidi 262, 2400444 (2025).4) X. Gao, Z. Jin, D. Wu, J. He, Y. Yan, Y. Liu, K. Ma, N. Xia, H. Zhang, andD. Yang, J. Alloys Compd. 1063, 187699 (2026).5) H. Murakami et al., Appl. Phys. Express 8, 015503 (2015).6) K. Goto, K. Konishi, H. Murakami, Y. Kumagai, B. Monemar,M. Higashiwaki, A. Kuramata, and S. Yamakoshi, Thin Solid Films 666,182 (2018).7) Q. T. Thieu, D. Wakimoto, Y. Koishikawa, K. Sasaki, K. Goto, K. Konishi,H. Murakami, A. Kuramata, Y. Kumagai, and S. Yamakoshi, Jpn. J. Appl.Phys. 56, 110310 (2017).8) K. Goto, H. Murakami, A. Kuramata, S. Yamakoshi, M. Higashiwaki, andY. Kumagai, Appl. Phys. Lett. 120, 102102 (2022).9) Y. Oshima and T. Oshima, Sci. Technol. Adv. Mater. 26, 2585551 (2025).10) Y. Oshima and T. Oshima, Semicond. Sci. Technol. 38, 105003 (2023).11) F. Otsuka, H. Miyamoto, A. Takatsuka, S. Kunori, K. Sasaki, andA. Kuramata, Appl. Phys. Express 15, 016501 (2022).12) D. Wakimoto, C.-H. Lin, Q. T. Thieu, H. Miyamoto, K. Sasaki, andA. Kuramata, Appl. Phys. Express 16, 036503 (2023).13) J. Zhang et al., Nat. Commun. 13, 3900 (2022).14) J. Yang, S. Ahn, F. Ren, S. J. Pearton, S. Jang, J. Kim, and A. Kuramata,Appl. Phys. Lett. 110, 192101 (2017).15) J. B. Varley, J. R. Weber, A. Janotti, and C. G. Van de Walle, Appl. Phys.Lett. 97, 142106 (2010).16) D. Wakimoto, C.-H. Lin, K. Ema, Y. Ueda, H. Miyamoto, K. Sasaki, andA. Kuramata, Appl. Phys. Express 18, 106502 (2025).17) Y. Oshima, K. Kawara, T. Oshima, M. Okigawa, and T. Shinohe,Semicond. Sci. Technol. 35, 055022 (2020).18) T. Oshima and Y. Oshima, Appl. Phys. Express 15, 075503 (2022).19) E. J. Hollar and E. Farzana, Appl. Phys. Lett. 128, 053503 (2026).20) B. Chen, W. Mu, Y. Liu, P. Wang, X. Ma, J. Zhang, X. Dong, Y. Li, Z. Jia,and X. Tao, CrystEngComm 25, 2404 (2023).21) M. M. Hossain Sarkar, D. S. Yu, S. Alam, M. Hassan, J. Hwang, andH. Zhao, ACS Appl. Electron. Mater. 8, 1380 (2026).22) T. Oshima and Y. Oshima, Jpn. J. Appl. Phys. 64, 018003 (2025).23) S. Sdoeung, Y. Otsubo, K. Sasaki, A. Kuramata, and M. Kasu, Appl. Phys.Lett. 123, 122101 (2023).24) J. Åhman, G. Svensson, and J. Albertsson, Acta Crystallogr. C 52, 1336(1996).25) P. Mazzolini, P. Vogt, R. Schewski, C. Wouters, M. Albrecht, andO. Bierwagen, APL Mater. 7, 022511 (2019).26) M. Wang, S. Mu, and C. G. Van de Walle, Phys. Rev. Mater. 7, 064603(2023).27) T. Itoh, A. Mauze, Y. Zhang, and J. S. Speck, Appl. Phys. Lett. 117, 152105(2020).Fig. 4. Schematic illustrating the orientation relationship among the(011), (110), and (134) planes.138001-4© 2026 The Author(s). Published on behalf ofThe Japan Society of Applied Physics by IOP Publishing LtdJpn. J. Appl. Phys. 65, 138001 (2026) BRIEF NOTEhttps://orcid.org/0000-0001-8550-9735https://orcid.org/0000-0001-8550-9735https://orcid.org/0000-0001-8293-4891https://orcid.org/0000-0001-8293-4891https://doi.org/10.35848/1882-0786/ad6b73https://doi.org/10.1063/1.3674287https://doi.org/10.1002/pssb.202400444https://doi.org/10.1016/j.jallcom.2026.187699https://doi.org/10.7567/APEX.8.015503https://doi.org/10.1016/j.tsf.2018.09.006https://doi.org/10.1016/j.tsf.2018.09.006https://doi.org/10.7567/JJAP.56.110310https://doi.org/10.7567/JJAP.56.110310https://doi.org/10.1063/5.0087609https://doi.org/10.1080/14686996.2025.2585551https://doi.org/10.1088/1361-6641/acf241https://doi.org/10.35848/1882-0786/ac4080https://doi.org/10.35848/1882-0786/acc30ehttps://doi.org/10.1038/s41467-022-31664-yhttps://doi.org/10.1063/1.4983203https://doi.org/10.1063/1.3499306https://doi.org/10.1063/1.3499306https://doi.org/10.35848/1882-0786/ae0d2ahttps://doi.org/10.1088/1361-6641/ab7843https://doi.org/10.35848/1882-0786/ac75c8https://doi.org/10.1063/5.0309746https://doi.org/10.1039/D3CE00052Dhttps://doi.org/10.1021/acsaelm.5c02627https://doi.org/10.35848/1347-4065/ada706https://doi.org/10.1063/5.0170398https://doi.org/10.1063/5.0170398https://doi.org/10.1107/S0108270195016404https://doi.org/10.1107/S0108270195016404https://doi.org/10.1063/1.5054386https://doi.org/10.1103/PhysRevMaterials.7.064603https://doi.org/10.1103/PhysRevMaterials.7.064603https://doi.org/10.1063/5.0027884https://doi.org/10.1063/5.0027884 Acknowledgments A2