Floral design GaN crystals: low-resistive and low-dislocation-density growth by oxide vapor phase epitaxy

GaN crystal growth mode in the oxide vapor phase epitaxy (OVPE) method, which simultaneously provides low electrical resistance and low threading dislocation density (TDD), has been investigated in detail. The results clarified that these qualities can be achieved by the expression of numerous inverted pyramidal pits, called three-dimensional (3D) growth mode. This mode reduced TDD from 3.8 × 106 cm−2 to 2.0 × 104 cm−2 for 1 mm thick growth because the threading dislocations (TDs) converged to the center of each pit. Moreover, when the crystal surface after polishing was observed by photoluminescence measurement, peculiar floral designs reflecting the distribution of oxygen concentration were observed over the entire surface. In addition, the etch pits exhibited TDs in the center of each floral design. On the basis of our results, we proposed that the 3D-OVPE-GaN will serve as a key material for improving the performance of vertical GaN devices.


Introduction
Electronic devices using gallium nitride (GaN) 1,2) are a key technology for improving energy efficiency toward the achievement of sustainable development goals. [3][4][5] Vertical GaN power devices are promising for high-power applications because they offer higher breakdown voltages and higher current operation on smaller chip sizes than horizontal structures. Despite such high potential, the development speed of GaN devices has been slower than that of silicon carbide devices. One reason for this is the lack of highquality and low-cost freestanding GaN wafers. 6) To fully draw out the potential of vertical GaN power devices and to implement them in society, it will be necessary to establish a manufacturing technology for GaN wafers featuring low dislocation density, low electrical resistance, large diameter, and low cost. 7,8) To reduce threading dislocation density (TDD), the hydride vapor phase epitaxy (HVPE) method using a technique called dislocation elimination by epitaxialgrowth with inverse-pyramidal pits, 9) the epitaxial lateral overgrowth (ELO) 10) or maskless-3D technique, 11) the Naflux method with multipoint seed technique, 12,13) and the ammonothermal method 14,15) have been proposed. To reduce resistance, Ge and Si doping techniques in the HVPE method have been reported. [16][17][18][19] Further, to increase the diameter, 6-7 inch GaN wafers have been achieved by the HVPE and Naflux methods. 13,[20][21][22] We have been developing the oxide vapor phase epitaxy (OVPE) method to manufacture high-performance, low-cost GaN wafers. 23) This method can be used to manufacture large bulk GaN crystals with a simple apparatus for long-term growth, since no NH 4 Cl is generated as a solid by-product, unlike the case with the HVPE method, which is the mainstream GaN manufacturing technology. Therefore, we expect the OVPE method to be able to produce low-cost GaN wafers. We have also reported the fabrication of low-TDD (TDD: the order of 10 4 cm −2 ) and low-resistance (resistivity: the order of 10 −4 Ω cm) 2 inch GaN wafers. Moreover, the characteristics of the p-n diodes fabricated on the OVPE-GaN wafers include a high breakdown voltage of 1.8 kV, and extremely low on-resistance of 0.08 mΩ cm 2 was demonstrated by conductivity modulation with highly efficient photon recycling. 24) However, the origins of the reduction of TDs and of the low resistivity have not yet been presented. This paper discusses for the first time the characteristics of OVPE-GaN crystals, with a focus on the growth mode. 3D-OVPE growth was the key technique for achieving simultaneously low TDD and low resistance. Furthermore, distinctive floral design patterns were observed in the overall polished surfaces of the OVPE-GaN crystals grown by the 3D growth mode.

Crystal growth conditions
We constructed the vertical reactor shown in Fig. 1 for use in this experiment. The reactor consisted of a source zone and a growth zone, each with electric resistance heaters. The reactor materials were quartz and ceramic. The temperature of each zone was controlled by the respective heaters. The source zone had four gas lines (lines 1-4) and a Ga boat. The growth zone had a seed substrate set on the seed holder, which had a rotation system. In the OVPE method in this experiment, Ga 2 O gas as a Ga source and GaN crystals were generated using the following Eqs. (1) and (2).
The following materials were used in this crystal growth. Metallic Ga (6N) was used as a starting gallium source, and NH 3 gas (5N) was used as a nitrogen source. H 2 gas (7N) and N 2 gas (6N) were supplied as carrier gases, and O 2 gas (6N) was supplied to generate H 2 O gas. H 2 O gas was generated upstream of the metallic Ga in Fig. 1. As seed substrates, commercially available GaN substrates manufactured by the HVPE method (+c-plane with 0.3 ± 0.2°off-angle toward the a-axis direction; 0.4 mm thick, f 2 inch) were used. The TDD of the HVPE-GaN substrates was 1 -5 × 10 6 cm −2 . OVPE-GaN crystals with different growth modes were prepared by changing the V/III (nitrogen source flow rate/ gallium source flow rate) ratio during growth and were used as analytical samples for investigating the structural and electrical properties. According to previous research, 2D growth was promoted in the case of a high V/III ratio by the OVPE method on the +c-plane, and 3D growth was advanced in the low V/III ratio. 25) In this experiment, the temperatures of the source and growth zones were set to 1130°C and 1200°C, respectively. The input conditions of the gas flow rates in the case of high and low V/III ratios are shown in Table I. Here, the conditions of the high V/III ratio and the low V/III ratio are named conditions A and B, respectively. Under condition A, H 2 , N 2 , and H 2 O flows were flowed at rates of 7.2, 1.8, and 0.014 l min −1 from gas line 1. N 2 was flowed at 10 l min −1 from line 2. From line 3, H 2 and NH 3 were introduced at 10 and 20 l min −1 , and N 2 was flowed at 24 l min −1 from line 4. Under condition B, H 2 , N 2 , and H 2 O were flowed at 4, 1, and 0.04 l min −1 from gas line 1. H 2 and N 2 were put in at 2.5 and 2.5 l min −1 from line 2. N 2 and NH 3 were introduced at 13-14 and 1 -2 l min −1 , respectively, from line 3. The total flow rate of N 2 and NH 3 in line 3 was set to 15 l min −1 . H 2 and N 2 were flowed at 12.5 and 12.5 l min −1 from line 4. As indicated in Eq. (1), the amount of Ga 2 O generated is proportional to the amount of H 2 O supplied when there is sufficient liquid Ga in the Ga boat. Therefore, the V/III (NH 3 flow rate/H 2 O flow rate) ratios were 1429 and 25-50 at conditions A and B, respectively, when the V/III ratio was replaced by the ratio of the nitrogen source to the oxygen source.

Crystal evaluation methods
To evaluate the quality of the grown crystal, the following were analyzed: surface morphology, growth rate, growth mode, emission spectrum, impurity concentration, lattice constant, radius of lattice curvature, electrical properties, TDD, and TD behavior. Surface morphology was observed by scanning electron microscopy (SEM) (JEOL JSM-7610F). Growth rates were evaluated by measuring weight change. Growth modes of grown crystals were confirmed by the cross-sectional SEM and the depth observation of two-photon excitation photoluminescence (2PPL) (Nikon A1RMP) from the crystal surfaces, and the emission spectra were also measured by 2PPL. Impurity concentrations were measured by secondary ion mass spectrometry (SIMS) (CAMECA IMS-6f and IMS-7f). The lattice constant and the radius of the lattice curvature were measured by X-ray diffraction (Bruker D8 DISCOVER). The electrical properties were examined by Hall-effect measurement (TOYO ResiTest 8300) at room temperature. TDD was evaluated by measuring etch pit density. Etching was carried out with a mixed melt of NaOH-KOH (NaOH: KOH = 1: 1) at 450°C for 10 min. 26) The TDD of the grown crystal was evaluated after surface mechanical polishing (MP) and chemical mechanical polishing (CMP). The behavior of dislocation propagation was observed by transmission X-ray topography (XRT) of a cross section of a thinning sample. The XRT experiments were performed at the Hyogo beamline (BL24XU) of Spring-8. 27) 3. Results and discussion 3.1. Comparison of 2D and 3D growth modes 3.1.1. Appearance and surface morphology.  Fig. 2(a), abnormal points at the surface are indicated by the arrows in the dashed frame. These polycrystals were likely created by a high supersaturation ratio under the growth condition. The relationship between the supersaturation ratio and the occurrence of polycrystals was reported in Ref. 23; under these growth conditions, the supersaturation ratios of conditions A and B were 605 and 2 -8, respectively. As can be seen from Fig. 2(c), the grown crystal surface under condition A had a 2D morphology featuring step bunching. On the other hand, as shown in Fig. 2(d), the surface under condition B had a 3D morphology with numerous inverted pyramidal pits. We named the GaN crystals prepared in 2D and 3D modes as 2D-and 3D-OVPE-GaN, respectively. In this experiment, the growth rates of the 2D-and 3D-OVPE-GaN were 26 and  Table I. Gas flow rates and zone temperatures of growth conditions A and B.

Condition
Line 1 (l min −1 ) Line 2 (l min −1 ) Line 3 (l min −1 ) Line 4 (l min −1 ) Temperature (°C) 60 μm h −1 , and the growth rates were controlled by the partial pressure of Ga 2 O supplied to the substrate surface within these experimental conditions. 3.1.2. Impurity concentration. The impurity concentration of each crystal grown for 3 h after surface smoothing was evaluated by SIMS at 2-10 μm depth from the surface in the 75 μm × 75 μm range. The SIMS results appear in Table II. The average concentrations of oxygen and silicon in the 2D-OVPE-GaN were 3.8 × 10 17 and 1.4 × 10 19 atoms cm −3 , respectively, and carbon was below the detection limits. In the 3D-OVPE-GaN, the concentrations of oxygen and silicon were 4.8 × 10 20 and 2.0 × 10 19 atoms cm −3 , respectively, and carbon was below the detection limit. The oxygen concentration in the 3D-OVPE-GaN was extremely high relative to that in the 2D-OVPE-GaN because the main growth plane in the latter is a (0001) polar plane, whereas that in the 3D-OVPE-GaN is semipolar. Compared with the (0001) polar plane and the semipolar plane of the ideal surfaces, the (0001) polar plane is a Ga-rich surface while the semipolar plane is often an N-rich surface. It was reported that oxygen atoms are more easily incorporated into the N site than the Ga site. 28) Consistent with this, oxygen incorporation was much higher in the 3D-OVPE-GaN than in the 2D-OVPE-GaN. Moreover, the silicon concentration was almost equally high in both growth modes because quartz was used as the reactor material. In this experiment, the reactor was heated to over 1130°C, so the quartz evaporated and a large amount of silicon was mixed in the OVPE-GaN crystals. 3.1.3. Lattice constant and lattice curvature radius. In Fig. 3(a), the lattice constants of the 2D-and 3D-OVPE-GaN grown for 3 h are depicted with the values of the HVPE-GaN seeds. The average a-axis and c-axis lattice constants of the 2D-OVPE-GaN were 3.1895 and 5.1857 Å, respectively. These values were almost the same as those with the HVPE-GaN seed. In contrast, the values of the 3D-OVPE-GaN were 3.1898 and 5.1873 Å. The c-axis expanded remarkably compared with the seed. This is attributable to the fact that the a-axis lattice constant of OVPE-GaN is close to the value of the seed substrate because it grows under the restraint of the lattice spacing of the seed substrate a-axis, while the expansion of the c-axis lattice constant is due to volume change by a high concentration of oxygen. Additionally, the radii of the lattice curvatures were measured by 0002 GaN XRC mapping as shown in Fig. 3(b). The radius of the lattice curvature calculated from the following Eq. (3), here, Dx is the distance between the measurement positions, and w D is the difference between the w values.
The R values of the 2D-and 3D-OVPE-GaN were concave 6.01 m and convex 7.82 m, respectively. The seed lattice curvature radii were 8.27 and 5.74 m concave. In the 3D-OVPE, the radius of the lattice curvature changed from   29,30) These results suggested that Ga vacancies in the 3D-OVPE-GaN would promote the self-compensation effect of the free carrier.    shows, the dashed line is the ideal curve showing the relationship between carrier concentration and electron mobility when the compensation ratio θ (Na/Nd) is 0; as the compensation ratio increases, the mobility value with respect to the carrier concentration decreases. 32) As these results show, the compensation ratio in the relationship between the carrier concentration and the mobility did not increase significantly in 3D-OVPE-GaN. This might be due to the crystallographic feature of the 3D-OVPE-GaN. Research into this result is ongoing. 3.1.5. Threading dislocation density. Figures 6(a)-6(c) contain surface SEM images after NaOH-KOH etching of the HVPE-GaN seed substrate and of OVPE-GaN crystals grown for 3 h in each growth mode after polishing to a thickness of 50 μm. When the threading dislocations (TDs) were estimated from the etch pit densities of the SEM images, the TDDs of the HVPE-GaN seed substrate, the 2D-OVPE-GaN, and the 3D-OVPE-GaN were 3.8 × 10 6 cm −2 , 3.5× 10 6 cm −2 , and 7.7 × 10 5 cm −2 , respectively, as shown in Fig. 6(d).
According to these results, while the TDD of the 2D-OVPE-GaN was almost equivalent to that of the HVPE-GaN seed substrate, the TDD of the 3D-OVPE-GaN was reduced to about 20% of that of the HVPE-GaN with only 50 μm thick growth.
3.1.6. Photoluminescence spectrum. Figure 7 presents the measurement results of the PL spectra of the 2D-and 3D-OVPE-GaN at room temperature. The measured samples are the same as the 50 μm thick OVPE-GaN described in the previous section. The integration times were 3 and 100 ms in each 2D-and 3D-OVPE-GaN measurement, respectively, and the vertical axis in Fig. 7 represents values normalized by the integration times. The peak top, the peak intensity, and the full width at half maximum (FWHM) of the 2D-and 3D-OVPE-GaN were 3.418 and 3.492 eV, 17328 and 198, and 93 and 236 meV, respectively. The PL peak top in each crystal was shifted from a pure GaN band-gap energy of 3.39 eV to the higher-energy side. This is mainly the Burstein-Moss effect, which is more pronounced at high carrier concentrations. 29) Therefore, the peak top of the 3D-OVPE-GaN shifted to a value larger than that of the  2D-OVPE-GaN. The PL intensity of the 3D-OVPE-GaN was much lower than that of the 2D-OVPE-GaN, and the FWHM of 3D-OVPE-GaN was larger than that of the 2D-OVPE-GaN. It can be inferred that these results were caused by the 3D-OVPE-GaN having more Ga vacancies and composite defects with Ga vacancies than 2D-OVPE-GaN, as explained in part of Fig. 4.

3.2.
Origin of 3D growth mode and the growth facet 3.2.1. Origin of 3D-OVPE-GaN. The change in the OVPE-GaN growth mode revealed that the 3D growth mode can realize ultralow-resistance GaN crystals with extremely high carrier concentrations and can dramatically reduce the TDD with only 50 μm thick growth. Additionally, 3D-OVPE-GaN is grown under the condition of a low V/III ratio, which is favorable for a high rate of growth. The origin and the forming planes of the pits were then identified in order to elucidate the 3D growth mode in detail. The interface between the seed substrate and the 3D-OVPE-GaN grown layer was observed to confirm the starting point of the pits with 2PPL. , and one of the pits is indicated by a white arrow. Luminescence was significantly reduced when the pits were formed in the 3D growth mode, as explained by the PL spectrum results in Fig. 7. Additionally, the pits grew as the thickness increased, as exhibited in Figs. 8(c) to 8(d). Here, the densities of the dark spots and pits in Fig. 8 were 5 × 10 6 cm −2 at each depth, showing equivalent values. It is therefore assumed that the pits are generated starting from TDs. In an attempt to directly confirm the connection between TDs and pits, the TDs of the seed substrates were determined as shown by the red and yellow arrows in Fig. 8(a), and the 2D PL images were confirmed from the cross-sectional direction on the red line and yellow line containing the TDs illustrated by red and yellow arrows, respectively. The cross-sectional PL images appear in Figs. 8(e) and 8(f). These cross-sectional images were created by integrating 150 images of 2D-PL taken at equally spaced distances from z = −7 to 6 μm. Therefore, the plurality of horizontal lines are from the noise due to the integrated image. Also, the yellow, blue, and red lines and frames in Figs. 8(e) and 8(f) correspond to the observation positions in Fig. 8(a). As the images show, the pits were generated on the TDs as indicated by the red and yellow arrows, which are the same TDs as in Fig. 8(a). Additionally, a 3D image, with inverted light and dark, is shown in Fig. 8(g). The image confirmed that the pits were opened mainly on the TDs like flowers blooming on stems. These results suggested that the TDs of the seed substrate were the origin of the formation of the 3D-OVPE-GaN.

3.2.2.
Growth facet of 3D-OVPE-GaN. Next, the surface morphology of the 3D-OVPE-GaN grown for 6 h in condition B was observed in order to identify the surface crystal planes. Figure 9(a) exhibits a bird's-eye view SEM image of In addition, the difference in oxygen incorporation at each plane was evaluated. Figure 10(a) shows an optical microscope image of the position where SIMS was performed on the crystal surface after OVPE-GaN was surface polished to a thickness of 300 μm. The in-plane distribution image of oxygen concentration in the area inside the green frame in Fig. 10(a) is shown in Fig. 10(b). The growth history of the pits formed in 3D-OVPE was observed even on the surface after polishing as shown in Fig. 10(a). Figure 10(b) shows that the oxygen concentration of β{3034} was higher than that of α{1122}. The average oxygen concentration at β{3034} in Fig. 10(b) was 5.1 × 10 20 atoms cm −3 , and that at α{1122} was 3.2 × 10 20 atoms cm −3 . Each average value was calculated for a 2 μm 2 area.
It is possible to consider the nitrogen density of each grown surface in order to discuss this oxygen concentration difference. At the ideal surfaces, the nitrogen densities of the { } 1122 and { } 3034 were 1.22 × 10 15 cm −2 and 1.01 × 10 15 cm −2 considering the surface area in Ref. 33. However, in the case of the semipolar planes of GaN crystal, the ideal surfaces are not stable in a growth atmosphere,  according to Akiyama et al. and Kawamura et al. 33,34) Therefore, when comparing semipolar planes such as { } 1122 and { } 3034 , it is necessary to consider surface reconstruction in the growth atmosphere on each plane. In addition, the step-and-terrace structures of the different facets during growth are not the same. Hence, it is expected that more detail on these semipolar planes will be reported in the near future, making it possible to discuss this oxygen concentration difference. 3.2.3. Features of 3D-OVPE-GaN on +c-plane. The 3D-OVPE-GaN grown for 6 h was polished to a thickness of 300 μm. The polished surface after NaOH-KOH etching was observed by SEM and CL as shown in Figs. 11(a) and 11(b). The etch pits in Fig. 11(a), indicated by yellow arrows, correspond to the center of the pit history in Fig. 11(b). The TDD was 7.7 × 10 4 cm −2 for the 300 μm thick OVPE-GaN calculated by etch pit density; the seed TDD was 1-5 × 10 6 cm −2 . In 3D growth mode, since the surfaces exposed during growth were mainly { } 1122 and { } 3034 , TDs tended to propagate in the direction toward these inclined facets from the +c-plane. Consequently, the TDs converged at the bottom of each pit, thus facilitating dislocation coalescence and annihilation. In other words, it is possible to estimate the TDD by counting the numbers of pit histories of polished crystals or the growth pits of as-grown crystals. Moreover, Fig. 12 shows a surface PL image of the 3D-OVPE-GaN over a wide range. The image confirms that the floral designs formed uniformly on the entire surface and that TD reduction occurred within the wafer.

Relationship between growth thickness and TDDs
The relationship between the 3D-OVPE-GaN growth thickness and TDDs was also investigated. Figure 13 shows the TDDs of the OVPE-GaN crystals on the HVPE-GaN substrates as a function of growth thickness. Here, the reference value for TDD reduction with the r-FIELO technique reported by H. Geng et al. 35) is also described. In this experiment, each growth was carried out under condition B in Table I. The TDDs of OVPE-GaN were evaluated at approximately 1, 50, 200, 300, 400, 700, and 1000 μm thicknesses by counting the dark-spot density with 2PPL, the etch pit density of NaOH + KOH etched crystals, and the growth pit density of as-grown crystals using SEM. As shown in Fig. 13, the TDDs of the OVPE-GaN were reduced to 2 × 10 4 cm −2 with 1 mm growth from the order of 10 6 cm −2 . The relationship between TDDs and growth thickness is suggested as shown in Eqs. (4) and (5). 36) In those equations, r is the TDD, K is the dislocation reaction cross section, and h is the growth thickness. The parameter  h corresponds to the initial TDD r 0 at some thickness h 0 .
The solid line in Fig. 13 is the fitting curve derived from the experimental results. The values of the parameters optimized by fitting and the coefficient of determination R 2 are shown in the inset. Both fitting curves had a sufficiently high R 2 . It is noteworthy that the K of the OVPE-GaN was about 100 times that of the HVPE-GaN. The OVPE-GaN grown in 3D mode constantly exposed a large number of pits formed by the planes inclined from the c-plane. These pits coalesced with each other as the OVPE-GaN growing, resulting in the convergence of the TDs. Therefore, the K of 3D-OVPE growth reached a dramatically large value compared to that of the HVPE-GaN by the r-FIELO technique. Moreover, a cross-sectional thinning sample of the 3D-OVPE-GaN was produced so as to observe TD propagation. The sample was extracted as shown in Fig. 14(a). First, a 5 × 5.3 mm tip was taken from the 2 inch, 400 μm thick 3D-OVPE-GaN grown on the HVPE-GaN substrate, which was also 400 μm thick, after which the thinning sample was sliced from the tip. The sample size (length × width × thickness) was 0.8 mm × 5.3 mm × 0.1 mm. Both the front and back surfaces of the sample were treated by MP and CMP to remove the damaged layer. Figure 14 and the image in the 2110 diffraction reflects a-type and (a+c)-type dislocations, suggesting that 3D-OVPE growth decreased all types of TDs. Also, the interface between the HVPE-GaN substrate and the 3D-OVPE-GaN crystal in the image of 0002 diffraction is significantly darker than in the image of 2110 diffraction. This is because of the lattice distortion of the substrate and the grown layer. As shown in Fig. 3(a), the lattice constants of the 3D-OVPE-GaN crystal in the a-axis and c-axis directions were 3.1898 and 5.1873 Å, respectively. On the other hand, the lattice constants of the HVPE-GaN substrate were 3.1895 and 5.1858 Å. The lattice constants of the c-axis and a-axis of 3D-OVPE-GaN increased by 0.0290% and 0.0099%, respectively, compared to HVPE-GaN. Because the c-axis showed a larger change, it is considered that the strain at the interface is more prominent in the c-axis direction.

Challenges for large-diameter 3D-OVPE-GaN
We previously reported on 2 inch OVPE-GaN wafers. 23) Fabrication of larger-diameter wafers is needed to meet the needs of the GaN power device industry. By optimizing growth temperature, gas flow rate, and growth reactor design, we succeeded in fabricating-wafers of up to 6 inches with 3D-OVPE growth. Figure 16 displays 2, 4, and 6 inch OVPE-GaN as grown. The 2 and 4 inch OVPE-GaN were grown on freestanding GaN wafers, whereas the 6 inch OVPE-GaN was grown on a GaN template (GaN on silicon). Hence, thick (>400 μm) 2 and 4 inch OVPE-GaN were grown, while the 6 inch OVPE-GaN was grown as a thin film of 100 μm or less. 3D-OVPE-GaN will be a key technology that will greatly accelerate the development of next-generation GaN power devices, because it enables larger-diameter wafers with low resistance and low TDD.

Conclusions
In this study, we discussed GaN crystal growth technology by the OVPE method, which simultaneously realizes low electrical resistance and low TDD and which had not been reported in detail until now, with a focus on the growth mode. In growth using this method, the uptake of high-concentration oxygen into the crystal was promoted and the carrier concentration became high with the 3D growth mode, which   GaN crystals having extremely low resistivity were obtained as a result. Also in the 3D growth mode, since the growth surface was formed by a large number of inverted pyramidal pits, we clarified that the TDD was reduced from 3.8 × 10 6 cm −2 to 2.0 × 10 4 cm −2 with 1 mm thick growth, and the dislocation reaction cross-section K was a large value of 2.663 μm. To investigate the starting point of the pits for 3D growth, the seed substrate and the grown layer interface were observed using 2PPL, enabling us to confirm that the pits of the OVPE-GaN layer began to open starting from the TDs of the seed substrate. Moreover, PL observation of the grown crystal surface after polishing revealed a uniform floral pattern on the entire surface. This was attributed to the difference in oxygen incorporation between { } 1122 and { } 3034 . In addition, etch pit and CL observation in the same region of the crystal surface confirmed the presence of the threading dislocation in the center of each floral pattern. In other words, 3D-OVPE-GaN has a uniform TDD distribution over the entire surface. These results revealed that utilizing the 3D mode in OVPE-GaN growth is important for preparing crystals with low resistance and low TDD. We believe that this technology will be important for accelerating the development of next-generation GaN power devices because it enables large-diameter growth with low resistance and low dislocation density in 3D mode.

Acknowledgments
A part of this work was supported by the "Technical Innovation Project to Create a Future Ideal Society and Lifestyle" of the Ministry of Environment of Japan and by the "Advanced Low Carbon Technology Research and Development Program; ALCA; JPMJAL1201" of the Japan Science and Technology Agency. The synchrotron radiation experiments were performed at BL24XU of SPring-8 with approval from the Japan Radiation Research Institute