Growth of GaN on a three-dimensional SCAATTM bulk seed by tri-halide vapor phase epitaxy using GaCl3

GaN with a film thickness of 200–600 μm was grown on the as-grown three-dimensional supercritical acidic ammonia technology (SCAATTM) bulk seed that comprised only semipolar { 10 1 ¯ 1 ¯ } planes at a temperature as high as 1390 °C by tri-halide vapor phase epitaxy; further, the GaN film was also characterized. The FWHM value of the X-ray rocking curves was ∼40″, which was almost similar to the value of the used seed. The curvature radii were as large as 40–64 m. Further, the carrier concentrations were observed to be as small as 5.1 × 1017–9.1 × 1017 cm−3. However, the basal plane stacking fault densities were observed to be 3.4 × 101–5.4 × 101 cm−1 and were observed to increase during the growth process because of the as-grown SCAATTM seed surface condition.


Introduction
Recently, electronic devices have been extensively investigated using free-standing bulk c-GaN substrates because of their lower threading dislocation densities (TDDs) as compared to those of the devices that are fabricated using foreign substrates. [1][2][3][4] The potential of GaN semiconductors for both radio frequency devices and power electronics exceeds that of SiC, which has been used for several applications in power electronics, including air conditioners, power conditioners of solar cell systems, and railway vehicle systems. Generally, bulk c-GaN substrates that are fabricated using hydride vapor phase epitaxy (HVPE) are used in electronic devices, which exhibit a TDD of approximately 1 × 10 6 cm −2 . [5][6][7][8] However, several groups have observed that an increase in TDD caused various device performance issues such as decreased electron mobility, deterioration of transconductance gm, and increased current leakage. [9][10][11] Reference 12 reported that the screw component of threading dislocation would cause leakage current and that the edge and mixed components would not cause leakage current by using conductive atomic force microscopy. Therefore, HVPE-GaN materials that are grown on true bulk GaN crystals are fabricated using solution growth methods, including the ammonothermal method and the Na-flux method. These methods could potentially balance low TDD values of 1 × 10 4 cm −2 or the low and high growth rates that were obtained using HVPE at more than 100 μm h −1 . [13][14][15][16] Furthermore, the growth of GaN along the c-axis faces a significant challenge with respect to the growth of bulk GaN having a thickness of a few tens of millimeters; the diameter of the GaN crystal decreases with its growth because its edge contains tilted planes comprising {1011}, {1012}, or {1122}, which grow in the inward direction of the wafer. 17) Therefore, the maximum thickness of GaN is limited, which results in high costs during the fabrication of a GaN wafer.
GaN that is grown on the supercritical acidic ammonia technology (SCAAT TM ) bulk seed by tri-halide vapor phase epitaxy (THVPE) using GaCl 3 is a promising method for fabricating high-quality GaN substrates with TDDs of 1 × 10 4 cm −2 or lower because THVPE enables high growth temperatures of ∼1400°C. [18][19][20] 21) These results show that THVPE growth denotes surface orientation selectivity for the growth of GaN. Furthermore, planar growth with a film thickness of ∼1.0 mm by THVPE on GaN substrates, which were sliced from a bulk crystal grown by SCAAT TM , revealed that crystal characteristics, such as the FWHM and the basal plane stacking fault (BSF) density of the GaN film on the semipolar ( ) 1011 plane, were superior to those of the nonpolar m-plane ( ) 1010 . 22) Hence, we propose the growth process in which the GaN crystal diameter could increase during the growth. Figure 1 illustrates a flowchart for the production of GaN wafers with a high crystal quality by combining THVPE with the SCAAT TM bulk seed. It was expected that the crystal size would increase during the growth because the SCAAT TM bulk seed comprised only of the { } 1011 planes. After growth, the crystal boule was sliced. Further, GaN wafers with TDDs of 1 × 10 4 cm −2 or lower could be obtained. The top-part of the crystal boule could be recycled for performing the forthcoming growths.
In this study, we conducted GaN growth on the as-grown three-dimensional SCAAT TM bulk seed comprising semipolar { } 1011 planes with six-fold symmetry at a temperature as high as 1390°C. Two samples with different growth times were grown using THVPE. Further, the growth rate, FWHM of the X-ray rocking curve (XRC), curvature radius, carrier concentration, impurity concentration, cathodoluminescence (CL), and photoluminescence (PL) of the GaN films were evaluated.

Experimental section
A three-dimensional SCAAT TM bulk seed that comprised semipolar { } 1011 planes with six-fold symmetry was employed as depicted in Fig. 2(a). This seed exhibited a shape that extended laterally because it was sliced from the GaN boule grown by SCAAT TM , mainly toward the m-direction. The seed was dipped in a 1 M HCl aqueous solution for a few seconds and was followed by washing with water and drying. The XRC-FWHM and TDD were approximately 40 arcsec and 1 × 10 3 -1 × 10 4 cm −3 , respectively.
For the THVPE of GaN, GaCl 3 was used as a group-III precursor, which was generated by the two-step reaction between Cl 2 gas and Ga metal, as denoted in Eqs. (1) and (2). In the deposition zone, NH 3 was introduced to contact the substrate, and GaN was grown using the reaction between GaCl 3 gas and NH 3 gas, as depicted in Eq. (3). Further, the details of THVPE used in this experiment have been reported in a previous study. 21) The schematic of the reactor was depicted in Fig. 3. Optimized THVPE growth conditions were used in this study. The growth was conducted by employing N 2 as the carrier gas at 1390°C using a residence heating susceptor. Two growth conditions (named sample #1 and sample #2) with different growth times of 3 and 10 h, respectively, were used to observe the influence of film thickness on the number of properties of GaN. The partial input pressures of NH 3 and the generated GaCl 3 were 1.2 × 10 −1 and 2.3 × 10 −3 atm, respectively. The film thickness was determined using crosssectional fluorescence (FL) microscopy. Further, the crystal quality was evaluated by XRC-FWHM using double crystal X-ray diffraction (XRD). The curvature radius toward á ñ 1120 was evaluated by ω-scan peak dependence on the crystal position using XRD. 23) The carrier concentration was determined using Raman spectroscopic measurements.
Additionally, impurity concentrations of oxygen and silicon that were incorporated into GaN were investigated using secondary-ion mass spectrometry (SIMS). The BSF was characterized using a scanning electron microscope (SEM) equipped with CL and a liquid nitrogen cooling stage at low temperatures (82-83 K). 24) Further, PL measurements were conducted to evaluate the optical properties of the materials.

Results and discussion
Figures 2(a) and 2(b) depict the photographs of the SCAAT TM bulk crystal seed before and after growth. Prior to growth, the seed displayed a yellow color because impurities were incorporated during acidic ammonothermal growth. After growth, the color of the seed became almost transparent, and it could be observed that the surface morphology displayed some undulations along á ñ 1120 . Further, the differential interference contrast microscopy images of the sample surfaces are depicted in Fig. 4.   A smooth morphology was observed in the case of sample #1. Sample #2 also exhibited a smooth surface with some triangle hillocks. Figure 5 depicts the cross-sectional FL images for the GaN film grown on the SCAAT TM bulk seed for (a) sample #1 and (b) sample #2. Further, we observed the as-sliced samples that were sliced along the a-face. It was possible to distinguish the epilayer from the SCAAT TM bulk seed because the seed was brighter than the epilayer owing to the high concentration of impurities. 18) The film thickness of sample #1 was ∼167-208 μm depending on the crystal position; we observed that the higher the position, the higher the film thickness will be. Note that the source gas including NH 3 and GaCl 3 flowed from above the sample. The film thickness of sample #2 was ∼519-615 μm, which increased with the growth time. The facets consisting of ( ) 0001 and { 1010} planes were observed in some parts of the crosssection in Fig. 5(b) (e.g. the part inwhite circle), which were possibly caused by the growth and coalescence of hillocks in Fig. 4. The upper layer of the sample toward [ ] 0001 was Vshaped. This could be attributed to a slower growth rate of the ( ) 0001 plane in relation to that of the { } 1011 planes. It is expected to find optimized growth conditions in which the growth rate on the ( ) 0001 plane is faster than that on the { } 1011 planes. Table I presents several properties of samples #1 and #2. The XRC-FWHMs were 42″ and 41″ for samples #1 and #2, respectively. These results indicate that the crystal quality did not degrade during the THVPE growth process. The curvature radii were as large as 64 and 36 m corresponding to the off-angle of ∼0.001°in terms of the f45 wafer. Figure 6 depicts the Raman spectra for (a) sample #1 and (b) sample #2. A 1 (LO) indicates a longitudinal optical mode, which forms the LO phonon and plasmon-coupled (LOPC) mode. The peak shift of A 1 (LO) can be explained by the formation of the LOPC mode, which is dependent on the carrier concentration. The carrier concentration n was calculated using the A 1 (LO) frequency w + as follows: 25 where w p is the plasmon frequency and e is the elementary charge. e ¥ (=5.35) is the optical dielectric constant of GaN, and * m (=0.2m 0 ) is the effective mass of the electron. 26) w L (=735 cm −1 ) is the LO phonon frequency in the uncoupled sample. The carrier concentrations of samples #1 and #2 were calculated to be 9.1 × 10 17 and 5.1 × 10 17 cm −3 , respectively, using these equations. Furthermore, the impurity concentrations of oxygen (O) for samples #1 and #2 were measured using SIMS and were observed to be 8 × 10 17 and 3 × 10 17 cm −3 , respectively, while the impurity concentration of silicon (Si) for both samples was 1 × 10 17 cm −3 . Thus, the difference between carrier concentrations could be attributed to the incorporation of oxygen in GaN based on the amount of H 2 O degasification in the reactor during growth. Note that hydrogen concentrations were below the detection limit for SIMS measurement. Furthermore, SIMS measurement for SCAAT TM bulk seed exhibited that the impurity concentrations of oxygen, silicon, and hydrogen were 2 × 10 18 cm −3 , below the detection limit, and 5 × 10 19 cm −3 , respectively, which shows that THVPE growth on SCAAT TM bulk seed could lead to a significant reduction of impurities in GaN except for silicon.
BSFs appear considerably often during the GaN growth of nonpolar and semipolar planes. The BSFs in wurtzite GaN can be categorized into three types, I 1 (…ABABCBCB…), I 2 (…ABABCACA…), and E (…ABABCABA…), with each exhibiting different atomic sequences of c-plane stacking. 27) The BSF for type I 1 , which has the lowest formation energy, is the most dominant in wurtzite GaN. BSFs exhibit a local deviation from the hexagonal wurtzite to the cubic zincblende crystal structure. 28) This wurtzite/zinc-blende heterostructure results in a zinc-blende quantum well in the wurtzite matrix, leading to an emission with a unique wavelength that depends on the BSF type. GaN films were investigated to characterize the BSFs using monochromatic low-temperature  CL (LT-CL) measurements at a wavelength of 364 nm for type I 1 BSFs. Further, typical images of each sample are depicted in Fig. 7. The bright lines in the LT-CL image indicate the BSFs that were generated during homoepitaxial growth by THVPE. The CL images were obtained at five different points over a sample wafer for evaluating the BSF density. The average BSF densities are presented in Table I. Further, the average BSF densities for samples #1 and #2 were 3.4 × 10 1 and 5.4 × 10 1 cm −1 , respectively. Samples with longer growth times resulted in larger BSF densities, indicating that the BSF densities increased during the growth process. Figure 8 depicts the PL spectra of samples #1 and #2, which are measured at 4 K. The PL spectra comprised a dominant donor-bound exciton (D 0 X) at 3.478 eV. BSFs in the GaN layer manifest in the PL spectra in the form of an emission band at ∼3.42 eV, which corresponds to type I 1 BSF. 29) An emission band was observed at 3.425 eV for sample #2, whereas a weak emission band (shoulder peak) resulting from BSFs was observed for sample #1, which was in agreement with the results of CL measurements. The emission arising from the donor-acceptor pair recombination (D 0 A 0 ) at 3.3 eV has also been identified for sample #2. 30) Yellow luminescence (YL) was not observed from 2.0 to 2.5 eV in either of the samples, which was a typical feature of carrier recombination via native point defects and impurities in GaN. SCAAT TM bulk seed itself showed a strong YL peak intensity due to a vacancy-impurity complex such as Ga vacancy coupled with impurities such as oxygen or hydrogen, which was identified using a positron annihilation measurement. 31) We previously performed planar growth with a thickness of ∼0.5 mm on the SCAAT TM ( ) 1011 substrate, and the BSF density was observed to be as small as 3.0 cm −1 , which was an order of magnitude smaller than that observed in this study. 22) We assume that there are two possibilities to explain this difference in BSF density. One is that BSF densities could be attributed to the difference between the substrate surface conditions prior to the growth process. Thus, chemically and mechanically polished treated substrates that were used in our previous study resulted in a relatively small BSF density. In contrast, the as-grown SCAAT TM bulk seed that was employed in this study led to a relatively large BSF density. The other is that BSF could be generated due to the difference in a substrate shape between the planar and the three-dimensional bulk seed comprising semipolar { } 1011 planes. Reference 32 reported that BSFs for types I 1 and I 2 were generated during the semipolar { } 1101 side facets of the triangle shape growth by using SiO 2 stripe mask on c-plane sapphire. Furthermore, Ref. 33 reported that stacking faults with high density were observed during GaN nano-rod growth along the c-direction. Reference 34 reported that the exposed c-plane facets presenting during growth accounts for   the consequent coalescence behavior of these facets and the formation of high-density BSFs. Thus, the emergence and consequent coalescence of facets along with some undulations toward á ñ 1120 in Fig. 2(b) would result in BSF generation. In future studies, we intend to achieve a reduced BSF density for the ( ) 1011 plane to optimize the treatment of the SCAAT TM seed surface and to elucidate the cause of occurrence of BSF.

Conclusions
To summarize, GaN with a film thickness of 200-600 μm was grown on the as-grown three-dimensional SCAAT TM bulk seed that comprised only semipolar { } 1011 planes at temperatures as high as 1390°C using THVPE; further, GaN films were evaluated using various characterization methods. XRC-FWHM values were as small as ∼40″, which was almost similar to the XRC-FWHM value of the seed that was used. The curvature radii were as large as 40-64 m. Additionally, carrier concentrations were as small as 5.1 × 10 17 -9.1 × 10 17 cm −3 , which can be attributed to the oxygen impurities in GaN. The BSF densities ranged between 3.4 × 10 1 and 5.4 × 10 1 cm −1 and increased during growth probably because of the as-grown SCAAT TM seed surface condition. Further investigation of the treatment of the SCAAT TM seed surface is required to suppress the occurrence of BSFs. In future, we intend to achieve a low BSF density for the ( ) 1011 plane to optimize the treatment of the SCAAT TM seed surface.