P-type doping in selected areas of gallium nitride (GaN) using magnesium (Mg)-ion implantation and subsequent ultra-high-pressure annealing (UHPA) are investigated to improve the performance of vertical GaN power devices. UHPA allows a high-temperature process without decomposition of the GaN surface and virtually complete activation of the implanted Mg ions in GaN. In the present paper, we provide an overview of recent challenges in making UHPA more realistic as an industrial process. Instead of UHPA at more than 1400 °C for a short duration, prolonged UHPA at 1300 °C demonstrates a comparable acceptor activation of Mg-ion-implanted GaN. This can reduce the annealing pressure to approximately 300 MPa and enlarge the processable wafer diameter. The second challenge is controlling the doping profiles in the lateral and vertical directions. We demonstrate fine patterning of the p-type regions, which indicates the limited lateral diffusion of Mg through UHPA. However, controlling the vertical doping profile is challenging. The nitrogen vacancies formed by ion implantation reduce the effective acceptor concentration near the surface, which can be compensated for by sequential nitrogen ion implantation. Defect-assisted Mg diffusion to the deeper region causes a redistribution of the Mg atoms and should be considered in the design of a device. Such anisotropic diffusion of Mg to the c-axis has potential applications in the fabrication of unique vertical device structures such as super junctions.

The great potential of vertical gallium nitride (GaN) power devices in high-power converters because of the high critical electric field of GaN,1–3 robustness of p–n junctions for reverse bias stress,4,5 high bulk electron mobility,6–8 and channel mobility over 100 cm2 V−1 s−1 for GaN metal–oxide–semiconductor field-effect transistors (MOSFETs) is abundantly clear.9–14 In fact, 600 V–50 A low-loss Schottky barrier diodes (SBDs),15 1–4 kV class p–n diodes (PNDs) with high Baliga's figure-of-merits (FOM) of 4.6–20 GW cm−2,4,6,16–18 and vertical FETs with blocking voltages greater than 1 kV and specific on-resistances less than 2 m Ω cm219–24 have been demonstrated. To further improve the performance and reduce cost, a breakthrough in the fabrication process is required. If impurity doping in selected areas using ion implantation is flexibly available, cost-effective device structures, which have contributed significantly to cost reduction for silicon (Si)-based power devices, as displayed in Fig. 1, can be applied to GaN.

FIG. 1.

Schematics of device applications using selective area Mg-ion implantation.

FIG. 1.

Schematics of device applications using selective area Mg-ion implantation.

Close modal

For a planar-type vertical MOSFET, called a double-implanted MOSFET (DMOSFET) in Fig. 1(a), p-type well regions and n+-type source contacts can be fabricated via ion implantation without mesa isolation. Such a simple fabrication process leads to reduction in fabrication costs and was recently demonstrated for GaN by the Fuji Electric group.23 The first challenge in a DMOSFET is a fine control of the acceptor concentration (Na) in the p-well because it determines the threshold voltage.25 Residual point defects formed through ion implantation and post-implantation annealing can influence the stability of the threshold voltage and, therefore, must be minimized. Edge termination [Fig. 1(b)] such as field-limiting rings (FLRs) or junction termination extensions (JTEs) is a common element in devices with high blocking voltages. Mesa termination with a field plate without the use of ion implantation can relax the electric field concentration.4,15 However, such complicated processes increase the fabrication cost. In addition, the doping concentration in the n-type drift layer for the field plate structure must be reduced compared to the ideal value expected from Baliga's FOM for GaN owing to residual electric field crowding, resulting in an increase in on-resistance. Therefore, edge termination using ion implantation is crucial for exploiting the superior physical properties of GaN. For both FLRs and JTEs, precise control of the Na value is required because a small gap from the optimal Na causes a significant reduction in the breakdown voltage.26,27 A junction barrier Schottky (JBS) diode, as displayed in Fig. 1(c), has significant potential for application. A JBS diode is often used as an antiparallel rectifier for the FET in a high-voltage power converter. Especially, a wide bandgap-based JBS diode can replace a conventional Si-based antiparallel p–n diode because of its low leakage current at a high operation temperature and its small reverse recovery.28 

Although a unipolar SBD is attractive because of its fast recovery in the switching operation,15 the reverse leakage current at a high operation temperature limits the applicable source voltage in a system. The extension of the depletion layer from the implanted p-type region can effectively reduce electric field crowding near the n-type GaN surface. For this structure, controlling the dopant diffusion in the lateral direction is crucial because the width of the n-type region is extremely sensitive to the on-state and off-state properties. The advanced structure is a MOSFET composed of a drift layer with a super junction (SJ) as illustrated in Fig. 1(d). This structure requires technology to distribute the dopants to a depth of a few micrometers. We discuss the potential of this method in Sec. V.

To realize the attractive device structures displayed in Fig. 1, common challenges remain in the ion-implantation technology. The first challenge is highly efficient acceptor activation via post-implantation annealing. N-type doping using Si or germanium ion implantation was demonstrated more than 20 years ago.29–32 Annealing at greater than 1100 °C for GaN specimens with high dosages commonly led to a donor activation greater than 50%. Conversely, p-type doping using magnesium (Mg)-ion implantation was more difficult owing to the doping asymmetry for III–V compound semiconductors; that is, donor-like point defects such as nitrogen (N) vacancies (VN) were easily formed in GaN33 and could compensate the acceptors. Therefore, a higher annealing temperature to remove implantation-induced defects is required for p-type activation. However, annealing temperatures greater than 1200 °C led to the degradation of the protective layers during post-implantation annealing, resulting in the decomposition of the GaN surface at the defects of the protective layer. An aluminum nitride (AlN) protective layer is relatively stable and allows annealing up to approximately 1300 °C for a relatively short duration. Although this condition cannot fully activate implanted Mg acceptors,25 MOSFETs functioned by optimizing the Mg dosage and VN compensation using N-ion sequential implantation. Multicycle rapid thermal annealing over 1300 °C34,35 and Mg-ion implantation for thermally stable N-polar GaN36 are alternative approaches to avoid decomposition during post-implantation annealing. However, the limitations of the annealing temperature and duration have inhibited the increase in the acceptor concentration for the Mg-ion-implanted region. To solve this problem, we suggest ultra-high-pressure annealing (UHPA).37 UHPA realized annealing up to 1480 °C under N2 pressure of 1 GPa, where the acceptor activation exceeded 70% of implanted Mg ions. The detailed methodology is described in Sec. II. However, a pressure of 1 GPa requires a thick wall of a high-pressure chamber and limits the processable wafer size, as in the high-pressure growth method,38 which could increase the manufacturing cost. If the annealing temperature is decreased, the equilibrium N2 pressure required for annealing can also be reduced. In Sec. III, we demonstrate that a prolonged UHPA duration at approximately 1300 °C provides an effective activation comparable to that of UHPA at 1400 °C for a short duration.

The second challenge is to control the p-type dopant profiles in the lateral and vertical directions. This is the key to selective area doping. Although devices using selective-area p-type doping have recently been reported,23,25–27 the spatial distributions of the Mg atoms in GaN were unclear. Furthermore, the acceptor activation ratio can also be varied in the depth direction owing to the distribution of vacancies after implantation and their migration. In Sec. IV, these points are discussed based on different cross-sectional analyses. In Sec. V, we demonstrate the development of vertical GaN devices fabricated by Mg-ion implantation and UHPA processes and discuss future perspectives using this technology. Section VI concludes our presentation.

The majority of the n-type GaN specimens implanted with Mg ions in this study were grown by metalorganic vapor phase epitaxy (MOVPE) on freestanding GaN substrates. The typical Si donor concentration was 2 × 1016 cm−3. Certain specimens described in Sec. III were grown by hydride vapor phase epitaxy.39 For the specimens discussed in Secs. III and IV, Mg ion implantation was performed at a tilt angle of 7° relative to the c-axis of the GaN with a rotation of the m-axis by 15° relative to the tilted plane40 to form a 300-nm-deep box-shaped Mg profile of 1019 cm−3. The multiple implantation energies were 300, 150, 70, and 30 keV with the corresponding dosages of 3.0 × 1014, 1.1 × 1014, 5 × 1013, and 2 × 1013 cm−3, respectively.37 N-ions were sequentially implanted for certain specimens discussed in Sec. IV. The implantation energies of the N-ions were 150, 70, and 20 keV with the respective dosages of 2.0 × 1014, 7 × 1013, and 5 × 1013 cm−3. Different implantation conditions were applied to the GaN devices, as discussed in Sec. V, based on the optimal device design predicted by device simulation.

GaN decomposes at a temperature as low as 900 °C under atmospheric pressure of N2.41 It is for this reason that annealing experiments using a protective layer are performed. If the N2 pressure in the annealing ambient increases, the temperature at which GaN decomposes also increases. Based on the thermodynamics of GaN and experimental studies, Karpinski et al.41,42 determined an accurate phase diagram of the GaN–Ga–N2 system. Using GaN specimens with low threading dislocation densities, we performed a replication study for Karpinski's diagram, as indicated in Fig. 2.43 The n-type GaN samples implanted with Mg ions were annealed for 5 min at the respective pressures and temperatures. Surfaces without decomposition were obtained above the solid curve, where GaN was a solid phase, whereas the surfaces were significantly decomposed on and below the equilibrium curve. The annealing condition in the solid GaN region is the basic concept of UHPA. In addition, Ga vapor pressure is also required to maintain a solid GaN phase. Hence, the Mg-ion-implanted GaN specimens were covered by GaN powder during annealing.44 The majority of the results in this study are based on this method.39 This approach is acceptable in terms of the solid GaN stability phase; however, it could possibly not be applicable to industrial semiconductor processes.

FIG. 2.

(a) Phase diagram of GaN–Ga–N2 and (b)–(k) surface morphologies after UHPA at each condition for 5 min. (b)–(k) are AFM images; (i) is a Nomarski microscope image. GaN can maintain a solid phase under N2 overpressure above the curve (white zone) and is decomposed under pressure on and below the curve (yellow zone). Compared with the conventional RTA with AlN capping layer in (k), UHPA under sufficient N2 pressure provided better morphologies. Modified with permission from Sakurai et al., IEEE 32th International Symposium on Power Semiconductor Devices & IC’s (ISPSD). Copyright 2020 IEEE.

FIG. 2.

(a) Phase diagram of GaN–Ga–N2 and (b)–(k) surface morphologies after UHPA at each condition for 5 min. (b)–(k) are AFM images; (i) is a Nomarski microscope image. GaN can maintain a solid phase under N2 overpressure above the curve (white zone) and is decomposed under pressure on and below the curve (yellow zone). Compared with the conventional RTA with AlN capping layer in (k), UHPA under sufficient N2 pressure provided better morphologies. Modified with permission from Sakurai et al., IEEE 32th International Symposium on Power Semiconductor Devices & IC’s (ISPSD). Copyright 2020 IEEE.

Close modal

To make UHPA more suitable for the manufacturing process, we improved the process. To realize the wafer-level process, we designed an annealing chamber using the hot isostatic pressure (HIP) method.45 In this approach, a sample was isostatically pressured by outside N2 gas up to medium pressure (typically 200 MPa); subsequently, the autoclave was compressed while elevating the temperature to the target pressure. Using the HIP equipment at the Japan Ultra-high Temperature Materials Research Institute (JUTEM), we annealed a number of 2-in. wafers up to 980 MPa and 1500 °C. If the maximum temperature could be reduced to 1300 °C, the pressure of the autoclave can be reduced to 500 MPa or less,40 resulting in a 6-in. wafer that could be processed in the future. To avoid contamination from foreign materials, we recently used a face-to-face process where the GaN surface was covered by a dummy GaN wafer as the Ga vapor source. Metal contamination was below the detection limit of secondary ion mass spectrometry (SIMS), as indicated in Fig. 3(a). Figures 3(b) and 3(c) display the Nomarski microscope images before and after UHPA at 1300 °C under an N2 pressure of 500 MPa for 15 min, respectively. A smooth surface comparable to that of the powder process was obtained. Moreover, the advanced face-to-face process could prevent particles and scratches originating from GaN powders. We also confirmed that the results of the Hall effect and Mg-diffusion profiles were consistent between the powder-covered process and face-to-face geometry. Thus, we believe that UHPA can be applied in industrial processes.

FIG. 3.

(a) Profiles of metal contamination in a 2-in. n-type GaN after UHPA at 1300 °C for 15 min under N2 pressure of 500 MPa. The 2-in. UHPA process was performed using the HIP equipment at JUTEM, Japan. Nomarski differential interference contrast microscope images for n-type GaN surfaces (b) before and (c) after UHPA in the same area.

FIG. 3.

(a) Profiles of metal contamination in a 2-in. n-type GaN after UHPA at 1300 °C for 15 min under N2 pressure of 500 MPa. The 2-in. UHPA process was performed using the HIP equipment at JUTEM, Japan. Nomarski differential interference contrast microscope images for n-type GaN surfaces (b) before and (c) after UHPA in the same area.

Close modal

The specimens used for the Hall-effect measurements and device fabrication were annealed in an N2 ambient at 850 °C for 60 min. This is because hydrogen atoms were introduced from the UHPA ambient and passivated Mg acceptors.46,47 Unintentional moisture may exist in the UHPA system because of the difficulty to achieve a perfect removal by purge. Then the residual moisture can become a hydrogen source under high pressure and temperature.46 The size of the Hall device was a 7.5 × 7.5 mm2 and 0.6-mm-diameter nickel/gold Ohmic electrodes were formed on its four corners by sintering in oxygen gas at 500 °C. Cathodoluminescence (CL) spectra and intensity mapping at the top face were employed at an acceleration voltage of 5 keV, which corresponded to an electron penetration depth of approximately 160 nm. At a temperature of 10 K, the electron current density for the CL spectra was typically 8.3 × 10−5 A cm−2. Cross-sectional CL spectra were acquired for the cleaved m-face at 30 K. Cross-sectional impurity and acceptor distributions were also analyzed for the selectively-doped areas by time-of-flight (TOF)-SIMS and scanning nonlinear dielectric microscopy (SNDM). Vertical devices were designed using technology computer-aided design (TCAD) simulations by Synopsys software, where the relative dielectric constant of GaN was assumed to be 10.4.48 The device structures and process are explained in Sec. V.

The annealing temperature is the most important process parameter in the UHPA process. The influence of UHPA temperature to efficiently activate the Mg acceptors was investigated. In Fig. 4, we revisit the results of the Hall-effect measurements for Mg-ion-implanted GaN after UHPA for a duration of 5 min at 1300 to 1480 °C under 1 GPa.39 For specimens after UHPA at 1300 °C, hole concentrations were one order of magnitude less than those of the specimens annealed at greater than 1375 °C. We estimated the residual donor concentration (Nd) of a 1300 °C-annealed sample to be approximately 90% of the Na value.39 The results suggest that point defects induced by the ion implantation primarily compensate acceptors for the 1300 °C-UHPA, whereas they were reduced by elevating temperature. Indeed, the Nd/Na ratios were less than 20% after UHPA at a temperature greater than 1375 °C.39 

FIG. 4.

Temperature dependence of (a) hole concentration and (b) hole mobility for Mg-implanted samples with UHPA at different temperatures. Modified with permission from Hirukawa et al., Appl. Phys. Express 14, 056501 (2021). Copyright 2021 The Japan Society of Applied Physics.

FIG. 4.

Temperature dependence of (a) hole concentration and (b) hole mobility for Mg-implanted samples with UHPA at different temperatures. Modified with permission from Hirukawa et al., Appl. Phys. Express 14, 056501 (2021). Copyright 2021 The Japan Society of Applied Physics.

Close modal

Based on the phase diagram in Fig. 2, the equilibrium N2 pressure at 1400 °C was approximately 600 MPa, which was overly high to obtain an adequate inner volume of the autoclave. Thus, we examined acceptor activation by prolonging the annealing duration at 1300 °C. The Mg diffusion and the Hall effect measurements were reported in Ref. 49. The CL spectrum is an acceptable indicator of the acceptor activation. In Fig. 5(a), the CL spectra of the sample set annealed for different durations at 1300 °C are compared with those of the reference samples annealed for 5 min at 1400 and 1480 °C. The CL spectra were composed of a near-band-edge (NBE) emission at approximately 3.47 eV, the donor–acceptor pair (DAP) band with a peak at approximately 3.28 eV, and green luminescence (GL) band at approximately 2.35 eV. For the spectrum of the sample annealed for 5 min at 1300 °C, the intensity of the DAP band was approximately a quarter of that of the reference samples, which corresponded to the low activation ratio indicated by the Hall effect result in Fig. 4(a). In addition, the intensity of the GL band originating from VN-related defects50–52 for the 5 min-annealed sample was virtually double than that of the references. By prolonging the annealing duration to 30 min, the intensity of the DAP band recovered to a level comparable to or exceeding that of the reference samples, as indicated in Fig. 5(b). These results were consistent with a previous report by Breckenridge et al.53 

FIG. 5.

(a) CL spectra at 10 K and (b) CL intensity integrated in energy range 2.85–3.44 eV corresponding DAP band as a function of annealing duration. Specimens annealed at 1300 °C for more than 30 min [solid curves in (a) demonstrated DAP emission equal to or greater than UHPA 1400 and 1480 °C for 5 min [dashed curves in (a)].

FIG. 5.

(a) CL spectra at 10 K and (b) CL intensity integrated in energy range 2.85–3.44 eV corresponding DAP band as a function of annealing duration. Specimens annealed at 1300 °C for more than 30 min [solid curves in (a) demonstrated DAP emission equal to or greater than UHPA 1400 and 1480 °C for 5 min [dashed curves in (a)].

Close modal

Using annular dark-field scanning-transmission electron spectroscopy (ADF-STEM), the structures of the extended defects in the sample sets annealed for different durations at 1300 °C were compared to those in the sample annealed for a duration of 0.5 min at 1480 °C. For the sample annealed for 0.5 min at 1480 °C [Fig. 6(e)], dislocation loops having diameters of a few hundred micrometers were observed.54 We previously assigned a loop to a collapsed vacancy disk with an intrinsic nature.54 Such vacancy disks plausibly resulted from the migration and agglomeration of vacancies during UHPA.55 That is, the majority of the vacancies in the clear region between loops were removed. Therefore, the formation of dislocation loops led to an increase in the effective Na. As indicated in Figs. 6(a) and 6(b), primarily oblong and dot-like defects were observed after annealing for durations less than 15 min, whose natures were Mg-segregated inversion domains and extrinsic stacking faults, respectively.55 Conversely, intrinsic dislocation loops were observed for the specimens annealed for more than 30 min at 1300 °C, as indicated in Figs. 6(c) and 6(d). The inversion domains and extrinsic stacking faults were virtually removed. The results suggest that prolonging the annealing duration at approximately 1300 °C has the same effect as short annealing at more than 1400 °C. This is an important finding for reducing UHPA pressure and increasing the processable wafer diameter.

FIG. 6.

(a)–(d) Cross-sectional ADF-STEM images of samples annealed at 1300 °C for 5–60 min. (e) Sample annealed at 1480 °C for 0.5 min. Modified with permission from Nakashima et al., Appl. Phys. Express 14, 011005 (2021). Copyright 2020 The Japan Society of Applied Physics.

FIG. 6.

(a)–(d) Cross-sectional ADF-STEM images of samples annealed at 1300 °C for 5–60 min. (e) Sample annealed at 1480 °C for 0.5 min. Modified with permission from Nakashima et al., Appl. Phys. Express 14, 011005 (2021). Copyright 2020 The Japan Society of Applied Physics.

Close modal

Although fine patterning of different conduction-type areas is required to improve the performance of power devices, annealing typically causes dopant diffusion. First, we investigated the Mg diffusion in the p-type areas formed by ion implantation in the lateral direction (the in-plane normal to the c-axis of the GaN) during UHPA. Three representative device patterns, PNDs, JBS, and FLRs, were observed using scanning electron microscopy (SEM). After Mg ion implantation with a 300-nm-deep box shape to n-type GaN in the patterned areas, UHPA was performed at 1400 °C for 5 min under 1 GPa. In Fig. 7, the bright regions are Mg-doped p-type areas; the dark regions are n-type areas formed by MOVPE growth. This contrast is related to the difference in efficiencies of secondary electron emissions between p-type and n-type regions.56 In a JBS pattern, the designed widths of the n-type and p-type areas were 1.5 and 2 μm, respectively. Therefore, patterning on the order of micrometers is possible using UHPA. Surprisingly, an alignment pattern with a depth of 0.2 μm deep maintained its original form, even after high-temperature UHPA (not displayed). The results indicate that the influence of mass transport57 is minimal for annealing under high pressure.

FIG. 7.

Surface SEM images after UHPA at 1400 °C for 5 min. Mg ions were implanted into selective area for (a) PNDs, (b) JBS, and (c) JBS with FLRs. Bright contrasts indicate formation of p-type areas; dark areas are n-type.

FIG. 7.

Surface SEM images after UHPA at 1400 °C for 5 min. Mg ions were implanted into selective area for (a) PNDs, (b) JBS, and (c) JBS with FLRs. Bright contrasts indicate formation of p-type areas; dark areas are n-type.

Close modal

More precise doping profiles were obtained using TOF-SIMS, SNDM, and monochromatic CL imaging. Figures 8(a) and 8(b) display (1–100) cross-sectional views of Mg-ion-implanted areas with 5-μm-wide stripes before and after UHPA at 1400 °C for 5 min. Even after UHPA, Mg diffusion in the lateral direction is limited, whereas Mg atoms are distributed in the vertical direction (parallel to the c-axis) over 1 μm. Because not all Mg atoms are always activated, the distributions of the effective acceptors were analyzed using SNDM. Before UHPA [Fig. 8(c)], the implanted area was shown in white, indicating high resistivity. Conversely, after UHPA, a pink region indicating p-type conduction could be observed in Fig. 8(d), where the white regions surrounding the p-type areas are depletion layers. The p-type area was laterally expanded by approximately 0.5 μm compared to the original implanted area, which indicates the lateral diffusion of Mg, although the distribution of the absolute acceptor concentration cannot be determined in this dC/dV image. Figure 8(e) displays the monochromatic CL mapping from the top view at a photon energy of approximately 3.275 eV, which corresponds to the DAP band for the specimen after UHPA. The gray region in the center corresponds to the original implanted region; the bright edges on both sides are the areas of lateral diffusion. Note that the p-type conduction near the surface in the implanted area was imperfect, as indicated in Fig. 8(d), because the implantation damage of the main box-shaped depth region was not perfectly recovered. Because the bright edges in the CL mapping were the regions not influenced by implantation damage, more intense DAP emission was observed. From the width of the bright edges in Fig. 8(e) that was estimated from the line profile to the horizontal direction, Mg atoms were possibly distributed in the lateral direction by approximately 0.5 μm from the implanted edge. This value might be overestimation for the actual diffusion length because a luminescence measurement is generally sensitive to a small concentration of impurities.

FIG. 8.

(a) and (b) (1–100) Cross sections of Mg distributions detected by TOF-SIMS and (c) and (d) polarity maps via SNDM analyses. (a) and (c) are images as-implanted; (b) and (d) are those after UHPA at 1400 °C for 5 min under 1 GPa. (e) Surface CL mapping for DAP emission of [1–100] Mg-implanted stripe around 3.275 eV after UHPA.

FIG. 8.

(a) and (b) (1–100) Cross sections of Mg distributions detected by TOF-SIMS and (c) and (d) polarity maps via SNDM analyses. (a) and (c) are images as-implanted; (b) and (d) are those after UHPA at 1400 °C for 5 min under 1 GPa. (e) Surface CL mapping for DAP emission of [1–100] Mg-implanted stripe around 3.275 eV after UHPA.

Close modal

SNDM analysis suggests that vertical Mg diffusion was more significant than lateral diffusion. Detailed depth profiles of the acceptors were investigated using cross-sectional CL analyses. Figures 9(a) and 9(b) display the CL spectra at depths of 0.16 and 0.5 μm. For comparison, the cross-sectional CL spectrum for MOVPE-grown GaN doped with a Mg concentration of 1019 cm−3 (labeled Epi-GaN: Mg) is indicated by a dashed line. At the 0.16 μm depth, the intensity of the DAP band for the Mg-ion-implanted GaN (labeled Mg I/I) was approximately one-seventh of that for Epi-GaN: Mg. At 0.5 μm depth, the intensity of DAP emission for the Mg I/I was increased to one-third of that of the Epi-GaN: Mg. For the Mg I/I sample, the GL band originating from VN-related defects at the 0.16 μm depth was higher than that at the 0.5 μm depth. The sequential N-ion implantation labeled as Mg/N I/I in Fig. 9(a), evidently suppressed the GL band near the surface and simultaneously enhanced DAP emission.

FIG. 9.

Cross-sectional CL spectra at 30 K for Mg and Mg/N-ions implanted GaN at depths of (a) 0.16 μm and (b) 0.50 μm from the surface. Dashed spectra indicate the Mg-doped GaN grown by the MOVPE method.

FIG. 9.

Cross-sectional CL spectra at 30 K for Mg and Mg/N-ions implanted GaN at depths of (a) 0.16 μm and (b) 0.50 μm from the surface. Dashed spectra indicate the Mg-doped GaN grown by the MOVPE method.

Close modal

Figure 10 displays the depth profiles of the Mg atoms before and after UHPA and CL intensities at energies of the DAP and 2.35 eV bands. Note that the CL intensity at 2.35 eV involves not only the GL band but also yellow luminescence (YL), having a peak at approximately 2.2 eV and originating from residual carbon atoms in the MOVPE-grown GaN specimens.58,59 The 2.35 eV emission within 0.5 μm depth from the surface was mainly due to the GL band; the main component over 1 μm depth was attributed to the high-energy component of the YL band. For the Mg I/I sample, the DAP band was significantly reduced within a 0.2 μm depth from the surface [Fig. 10(b)] and the GL band increased in the corresponding depth region [Fig. 10(c)]. These results suggest that the VN-related defects significantly compensated for the Mg acceptors near the surface. The N-ion implantation effectively compensated for the VN defects and increased the effective acceptor concentration near the surface. We previously found that the compensation effect of VN was optimally achieved under the condition of an Mg-ion dosage equal to an N-ion dosage.52 Tanaka et al.23 reported that sequential N-ion implantation enhanced the breakdown voltages of p–n junctions formed by Mg-ion implantation because of the reduction effect of VN. In addition, the Mg concentrations in the depth region of 0.1–0.3 μm increased for the sequential N-ion implanted specimen, as indicated in Fig. 10(a). These results suggest that VN compensation via N-ion implantation is key to improving the surface acceptor concentration. Conversely, the vertical diffusion of Mg atoms was approximately double that of the lateral diffusion. Two possible reasons for such a diffusion anisotropy are considered. One is the vacancy-assisted diffusion effect originating from the ion-implantation.47,60 The other is Mg diffusion along threading dislocations.61 For GaN specimens used in this study, the average distance between each dislocation is approximately 10 μm (the threading dislocation density of ∼106 cm−2), which is much longer than the lateral diffusion length (∼0.5 μm). Therefore, the impact of Mg diffusion along dislocations is negligibly small for the overall Mg profile in Fig. 10(a). On the other hand, Uedono et al.60 pointed out that the depth profile of vacancy-type defects in the as-implanted sample distributed to the depth much deeper than the box-shape implanted region, and Mg atoms were redistributed based on the vacancy profile after UHPA. Since vacancies via the ion-implantation are not introduced in the masked area, the lateral diffusion can be suppressed compared to the vertical direction. This anisotropic redistribution of Mg atoms should be considered in the design of vertical devices.

FIG. 10.

(a) Depth profiles of Mg after UHPA at 1400 °C under N2 pressure of 1 GPa. Dashed lines indicate the simulated Mg and N profiles for as-implanted sample. CL intensities as a function of analysis depth at (b) 3.275 eV (DAP band) and (c) 2.35 eV (GL band). Note that signals at 2.35 eV in the deeper region involve YL bands originating from carbon incorporated during MOVPE growth.

FIG. 10.

(a) Depth profiles of Mg after UHPA at 1400 °C under N2 pressure of 1 GPa. Dashed lines indicate the simulated Mg and N profiles for as-implanted sample. CL intensities as a function of analysis depth at (b) 3.275 eV (DAP band) and (c) 2.35 eV (GL band). Note that signals at 2.35 eV in the deeper region involve YL bands originating from carbon incorporated during MOVPE growth.

Close modal

As mentioned in Sec. IV, it was clarified that the lateral diffusion of Mg activated by UHPA was suppressed compared to the vertical diffusion. Therefore, it is possible to fabricate GaN structures using a selective-area doping technique, as indicated in Fig. 1. In this section, the application of Mg-ion implantation to GaN vertical devices is discussed. The edge termination structure displayed in Fig. 1(b) is a typical and effective application of ion implantation in a device. The breakdown voltage of a vertical device is limited by electric field crowding at the edge of the device. To maximize the potential of a material, it is necessary to form an edge termination structure that mitigates electric-field crowding of the device edge. Popular edge termination structures are shallow bevel mesa structures, FLRs, and JTE structures. For example, a shallow bevel mesa structure does not require ion implantation, and the process is relatively simple. Furthermore, because there is no steep step on the slope, it has an acceptable affinity for the MOSFET process. However, to obtain a high breakdown voltage and avalanche capability with only a shallow bevel mesa structure, there are restrictions on the doping concentrations and thicknesses of both the p-type and n-type GaN layers.62–64 Therefore, we investigated an edge termination structure that can be applied to a wider range of device structures by combining a beveled structure and a selective p-type GaN region formed by Mg-ion implantation.26,27 In Ref. 26, after forming a Mg-ion implantation region with a width of 3 μm to overlap the p–n interface on the surface of the bevel, UHPA was used to activate the Mg acceptors. Although we confirmed the improvement in the breakdown voltage, a single guard ring resulted in a limited improvement effect, resulting in a hard breakdown. To further improve the breakdown voltage, we fabricated multiple p-type GaN rings (i.e., FLRs) on the outside of the previous structure and verified the effect. Figure 11(a) displays the structure of the fabricated device.27 The layered structure formed via MOVPE was n-type GaN (Si:2 × 1016 cm−3, thickness 7 μm), p-type GaN (Mg: 2 × 1018 cm−3, thickness 0.45 μm), and p+-type GaN (Mg: >1 × 1019 cm−3, thickness 0.1 μm) from the bottom to the top. The first fabrication process involved the formation of the bevel mesa using the thermal reflow process of the photoresist and subsequent chlorine-based inductively coupled plasma reactive ion etching.62 This process was optimized to obtain a shallow bevel angle of approximately 6°. In the next step, Mg-ion implantation was conducted at room temperature with a 180 eV energy and a dose of 5 × 1013 cm−2 under ⟨0001⟩ channeling conditions. Subsequently, we performed UHPA under a N2 pressure of 1 GPa at 1400 °C for 15 min. The UHPA at 1400 °C for more than 5 min allows the almost complete activation of Mg acceptors as discussed in Fig. 4. Finally, the anode and cathode electrodes were formed.

FIG. 11.

(a) Schematic of PND with shallow bevel termination and FLRs. (b) Electric field distributions and breakdown voltages of several edge terminations. (c) Reverse bias characteristics for PNDs with and without FLRs. Modified with permission from Matys et al., Appl. Phys. Express 14, 074002 (2021). Copyright 2021 The Japan Society of Applied Physics.

FIG. 11.

(a) Schematic of PND with shallow bevel termination and FLRs. (b) Electric field distributions and breakdown voltages of several edge terminations. (c) Reverse bias characteristics for PNDs with and without FLRs. Modified with permission from Matys et al., Appl. Phys. Express 14, 074002 (2021). Copyright 2021 The Japan Society of Applied Physics.

Close modal

When designing the devices, we simulated the electric field distribution at the breakdown voltage using the impact ionization coefficients in Ref. 3. Figure 11(b) displays the simulated results for the bevel mesas of the edge terminations composed without a guard ring, with a single Mg-ion implanted ring, and twelve FLRs having different ring spaces Ln. The breakdown voltages were significantly enhanced by the FLRs. The breakdown voltage for devices with FLRs depended on Ln and demonstrated the highest value for an optimal Ln of 1.0 μm. Figure 11(c) indicates the reverse-bias characteristics of the fabricated devices displayed in Fig. 11(a). The experimental results were in acceptable agreement with the simulation, demonstrating effective edge termination using Mg-ion implantation and the subsequent UHPA process.

Trench MOSFETs can minimize the on-resistance by reducing the cell pitch and are widely used in Si MOSFETs. However, for a trench MOSFET with a high reverse bias, electric field crowding at the bottom of the trench is a significant issue. At the corners of the bottom of the trench, the electric field tends to crowd and a high electric field is applied to the gate oxide, leading to a lack of reliability such as insulator destruction and lifetime shortening. Therefore, an electric field relaxation structure would be desirable. If Mg ion implantation in the selective area is possible, a highly effective electric field relaxation structure becomes possible. We designed an electric field relaxation structure using TCAD simulations. Figure 12(a) displays the device structure with a cell pitch of 5 μm used in the simulation. Figure 12(b) indicates the equipotential line and electric field distribution in the device without any countermeasure for electric field crowding. The breakdown voltage VB, on-resistance RON, and electric field strength at the gate oxide EOX are also indicated. In this simulation, RON is composed of a channel, drift, and substrate resistances. Strong electric-field crowding can be observed at the bottom of the trench, as displayed in Fig. 12(a). The electric field strength applied to the gate oxide reached 7 MV cm−1 at the breakdown voltage of GaN. A high electric field strength causes catastrophic destruction of the gate oxide shortly before the breakdown of GaN. An effective approach to reduce the electric field in the oxide is to form a p-type region outside the trench bottom. An example of the simulation results is displayed in Fig. 12(c). In this structure, the structural parameters include the acceptor concentration, lateral distance from the trench LJFET, and depth from the bottom of the p-type body Tbody. The structure in Fig. 12(c) is LJFET = 0.4 μm, Tbody = 1.5 μm; the acceptor concentration is the same as that of the p-type body. The electric field strength applied to the gate oxide was sufficiently reduced to 1.26 MV/cm and the breakdown voltage increased. However, an increase in the on-resistance caused by junction-gate field-effect transistor (JFET) resistance was observed. An increase in Tbody or decrease in LJFET reduces Eox, whereas the JFET resistance increases. Increasing Tbody also decreases the breakdown voltage by reducing the effective thickness of the drift layer. Thus, there is a strong trade-off between the oxide electric field and JFET resistance. To alleviate this trade-off, we investigated a structure that continuously changes the acceptor distribution in the additional p-type region. Channel ion implantation can distribute Mg deeply with relatively low energy.65,66 The Mg distribution assumed is displayed in Fig. 13(a). The distribution consisted of a flat concentration region and a gradually decreasing region. The depth of the flat Mg concentration region is indicated by xb; the channeling ion implantation distribution was used in the deeper region. The optimal performance was obtained when xb = 1.5 μm and LJFET = 0.6 μm. The simulation results are displayed in Fig. 13(b). Because the Mg distribution was inclined, the electric field crowding around the p–n junction was relaxed and a breakdown voltage greater than 1 kV with low EOX was obtained. In addition, because the electric field spreads laterally, the carrier density of the drift layer could be increased to 4.0 × 1016 cm−3, which contributed to the reduction of the on-resistance and a low JFET resistance of LJFET = 0.6 μm. However, in terms of fabrication, xb = 1.5 μm is overly deep because Mg ion energy of greater than 1 MeV is required. Therefore, xb = 0.5 μm, which is possible using 400 keV Mg ion energy, was examined. To compensate for the weakness of the shallow xb, an additional p-type pillar of the same Mg distribution was inserted under the trench. The simulation results are presented in Fig. 13(c). In this structure, the carrier concentration in the drift layer was 5.5 × 1016 cm−3; nevertheless, the breakdown voltage increased. This is because the electric field laterally expanded more than in the device displayed in Fig. 13(b) by the p-type pillar under the trench, which was similar to the effect of the SJ device. The highest performance was obtained for devices of similar size. Therefore, deep p-type pillars are highly effective for realizing high-performance devices.

FIG. 12.

(a) Simulated trench MOSFET structure with 5 μm cell pitch, and equipotential lines and electric field distribution in device under 600 V (b) without and (c) with an electric field relaxation structure. The simulated breakdown voltage VB, electric field strength on the gate insulator EOX, and specific on-resistance RON are also indicated.

FIG. 12.

(a) Simulated trench MOSFET structure with 5 μm cell pitch, and equipotential lines and electric field distribution in device under 600 V (b) without and (c) with an electric field relaxation structure. The simulated breakdown voltage VB, electric field strength on the gate insulator EOX, and specific on-resistance RON are also indicated.

Close modal
FIG. 13.

(a) Acceptor concentration as a function of depth used for the simulation of the electric field relaxation structure. (b) Simulated result of equipotential lines and electric field distribution in device for xb = 1.5 μm. Calculated VB, EOX, and RON values are also indicated. (c) Simulated result of device with p-type pillars outside and under trench.

FIG. 13.

(a) Acceptor concentration as a function of depth used for the simulation of the electric field relaxation structure. (b) Simulated result of equipotential lines and electric field distribution in device for xb = 1.5 μm. Calculated VB, EOX, and RON values are also indicated. (c) Simulated result of device with p-type pillars outside and under trench.

Close modal

Such a deeply distributed doping profile can be realized not only by using channeling ion implantation but also by combining anisotropic Mg diffusion in GaN during UHPA, as discussed in Sec. IV, or using the vacancy-guided diffusion method.67 As indicated here, when using deep p-type pillars and inclined acceptor distribution, the trade-off relation was effectively relaxed, resulting in improved performance. These results were simulated; it is necessary to proceed with experimental verification.

Selective-area p-type doping by ion implantation is a useful technique for vertical GaN devices; its realization has been required for many years. We recently developed a method called UHPA for activating ion-implanted Mg, providing a method for selective-area p-type doping of GaN by Mg-ion implantation. The initial experiment achieved an Mg activation rate greater than 70% by annealing at a temperature of 1400 °C under N2 pressure of 1 GPa.

However, for the practical use of the process equipment, it is necessary to reduce the processing pressure and temperature. Therefore, we investigated reducing the annealing temperature and determined that annealing at 1300 °C for 30 min enabled Mg activation equivalent to that at 1400 °C for 5 min. This finding demonstrates the possibility of reducing the pressure, which can increase the processable wafer diameter. In addition, the lateral diffusion of Mg was evaluated for the application of the ion implantation process to device fabrication. We found that the lateral diffusion of implanted Mg ions in GaN was approximately 0.5 μm, which was less than half of the diffusion in the depth direction. Based on these results, a selective-area p-type doping technology using Mg-ion implantation was applied to device fabrication. At the edge termination of the p–n junction, an FLR structure was formed by the Mg-ion implantation and subsequent UHPA, and the breakdown voltage was improved. Furthermore, an electric-field relaxation structure in the trench MOSFET was designed using a TCAD simulation. By forming deep p-type GaN pillars outside and under the trench via ion implantation, the device performance was dramatically improved, similar to that of SJ devices. Using this approach, Mg ion implantation can significantly expand the degree of freedom in the device design. In the future, we will proceed with a basic analysis of Mg ion implantation and expand its application to vertical GaN devices.

This work was supported by MEXT “Program for research and development of next-generation semiconductor to realize energy-saving society” (Grant No. JPJ005357) and MEXT “Program for the Creation of innovative core technology for power electronics” (Grant No. JPJ009777). A portion of this research was also supported by the Polish National Science Centre through Project No. 2018/29/B/ST5/00338 and Polish National Centre for Research and Development through Project No. TECHMATSTRATEG-III/0003/2019-00.

The authors have no conflicts to disclose.

Tetsu Kachi: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Tetsuo Narita: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Hideki Sakurai: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). Maciej Matys: Data curation (equal); Formal analysis (equal); Investigation (equal); Visualization (equal); Writing – review & editing (equal). Keita Kataoka: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – review & editing (equal). Kazufumi Hirukawa: Data curation (equal); Formal analysis (equal); Investigation (equal); Visualization (equal); Writing – review & editing (equal). Kensuke Sumida: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Visualization (equal). Masahiro Horita: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). Nobuyuki Ikarashi: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Kacper Sierakowski: Data curation (equal); Formal analysis (equal); Validation (equal). Michal Bockowski: Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). Jun Suda: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Project administration (lead); Resources (lead); Validation (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
L.
Cao
,
J.
Wang
,
G.
Harden
,
H.
Ye
,
R.
Stillwell
,
A. J.
Hoffman
, and
P.
Fay
,
Appl. Phys. Lett.
112
,
262103
(
2018
).
2.
D.
Ji
,
B.
Ercan
, and
S.
Chowdhury
,
Appl. Phys. Lett.
115
,
073503
(
2019
).
3.
T.
Maeda
,
T.
Narita
,
S.
Yamada
,
T.
Kachi
,
T.
Kimoto
,
M.
Horita
, and
J.
Suda
,
J. Appl. Phys.
129
,
185702
(
2021
).
4.
H.
Ohta
,
N.
Asai
,
F.
Horikiri
,
Y.
Narita
,
T.
Yoshida
, and
T.
Mishima
,
IEEE Electron Device Lett.
41
,
123
(
2020
).
5.
T.
Narita
,
Y.
Nagasato
,
M.
Kanechika
,
T.
Kondo
,
T.
Uesugi
,
K.
Tomita
,
S.
Ikeda
,
S.
Yamaguchi
,
Y.
Kimoto
,
M.
Kosaki
,
T.
Oka
,
J.
Kojima
, and
J.
Suda
,
Appl. Phys. Lett.
118
,
253501
(
2021
).
6.
I. C.
Kizilyalli
,
A. P.
Edwards
,
O.
Aktas
,
T.
Prunty
, and
D.
Bour
,
IEEE Trans. Electron Devices
62
,
414
(
2015
).
7.
N.
Sawada
,
T.
Narita
,
M.
Kanechika
,
T.
Uesugi
,
T.
Kachi
,
M.
Horita
,
T.
Kimoto
, and
J.
Suda
,
Appl. Phys. Express
11
,
041001
(
2018
).
8.
H.
Fujikura
,
T.
Konno
,
T.
Kimura
,
Y.
Narita
, and
F.
Horikiri
,
Appl. Phys. Lett.
117
,
012103
(
2020
).
9.
W.
Huang
,
T.
Khan
, and
T. P.
Chow
,
IEEE Electron Device Lett.
27
,
796
(
2006
).
10.
T.
Nomura
,
H.
Kambayashi
,
Y.
Niiyama
,
S.
Otomo
, and
S.
Yoshida
,
Solid-State Electron.
52
,
150
(
2008
).
11.
H.
Otake
,
K.
Chikamatsu
,
A.
Yamaguchi
,
T.
Fujishima
, and
H.
Ohta
,
Appl. Phys. Express
1
,
011105
(
2008
).
12.
K.
Yamaji
,
M.
Noborio
,
J.
Suda
, and
T.
Kimoto
,
Jpn. J. Appl. Phys.
47
,
7784
(
2008
).
13.
S.
Takashima
,
K.
Ueno
,
H.
Matsuyama
,
T.
Inamoto
,
M.
Edo
,
T.
Takahashi
,
M.
Shimizu
, and
K.
Nakagawa
,
Appl. Phys. Express
10
,
121004
(
2017
).
14.
D.
Ji
,
C.
Gupta
,
S. H.
Chan
,
A.
Agarwal
,
W.
Li
,
S.
Keller
,
U. K.
Mishra
, and
S.
Chowdhury
, in
2017 IEEE International Electron Devices Meeting
(
IEEE
,
2017
), p.
223
.
15.
T.
Oka
,
Jpn. J. Appl. Phys.
58
,
SB0805
(
2019
).
16.
K.
Nomoto
,
B.
Song
,
Z.
Hu
,
M.
Zhu
,
M.
Qi
,
N.
Kaneda
,
T.
Mishima
,
T.
Nakamura
,
D.
Jena
, and
H. G.
Xing
,
IEEE Electron Device Lett.
37
,
161
(
2016
).
17.
H.
Fu
,
K.
Fu
,
X.
Huang
,
H.
Chen
,
I.
Baranowski
,
T.-H.
Yang
,
J.
Montes
, and
Y.
Zhao
,
IEEE Electron Device Lett.
39
,
1018
(
2018
).
18.
J.
Wang
,
L.
Cao
,
J.
Xie
,
E.
Beam
,
R.
McCarthy
,
C.
Youtsey
, and
P.
Fay
,
Appl. Phys. Lett.
113
,
023502
(
2018
).
19.
H.
Nie
,
Q.
Diduck
,
B.
Alvarez
,
A. P.
Edwards
,
B. M.
Kayes
,
M.
Zhang
,
G.
Ye
,
T.
Prunty
,
D.
Bour
, and
I. C.
Kizilyalli
,
IEEE Electron Device Lett.
35
,
939
(
2014
).
20.
T.
Oka
,
T.
Ina
,
Y.
Ueno
, and
J.
Nishii
,
Appl. Phys. Express
8
,
054101
(
2015
).
21.
D.
Shibata
,
R.
Kajitani
,
M.
Ogawa
,
K.
Tanaka
,
S.
Tamura
,
T.
Hatsuda
,
M.
Ishida
, and
T.
Ueda
, in
IEEE International Electron Devices Meeting
,
San Francisco
(
IEEE
,
2016
), p.
248
.
22.
Y.
Zhang
,
M.
Sun
,
D.
Piedra
,
J.
Hu
,
Z.
Liu
,
Y.
Lin
,
X.
Gao
,
K.
Shepard
, and
T.
Palacios
, in
IEEE International Electron Devices Meeting
,
San Francisco
(
IEEE
,
2017
), p.
215
.
23.
R.
Tanaka
,
S.
Takashima
,
K.
Ueno
,
H.
Matsuyama
, and
M.
Edo
,
Jpn. J. Appl. Phys.
59
,
SGGD02
(
2020
).
24.
J.
Liu
,
M.
Xiao
,
R.
Zhang
,
S.
Pidaparthi
,
H.
Cui
,
A.
Edwards
,
M.
Craven
,
L.
Baubutr
,
C.
Drowley
, and
Y.
Zhang
,
IEEE Trans. Electron Devices
68
,
2025
(
2021
).
25.
R.
Tanaka
,
S.
Takashima
,
K.
Ueno
,
H.
Matsuyama
,
M.
Edo
, and
K.
Nakagawa
,
Appl. Phys. Express
12
,
054001
(
2019
).
26.
M.
Matys
,
T.
Ishida
,
K. P.
Nam
,
H.
Sakurai
,
T.
Narita
,
T.
Uesugi
,
M.
Bockowski
,
J.
Suda
, and
T.
Kachi
,
Appl. Phys. Lett.
118
,
093502
(
2021
).
27.
M.
Matys
,
T.
Ishida
,
K. P.
Nam
,
H.
Sakurai
,
K.
Kataoka
,
T.
Narita
,
T.
Uesugi
,
M.
Bockowski
,
T.
Nishimura
,
J.
Suda
, and
T.
Kachi
,
Appl. Phys. Express
14
,
074002
(
2021
).
28.
T.
Duong
,
A.
Hefner
,
K.
Hobart
,
S.-H.
Ryu
,
D.
Grider
,
D.
Berning
,
J. M.
Ortiz-Rodriguez
,
E.
Imhoff
, and
J.
Sherbondy
, in
2011 Twenty-Sixth Annual IEEE Applied Power Electronics Conference and Exposition (APEC)
, (IEEE,
2011
), pp.
1057
1063
.
29.
S. J.
Pearton
,
C. B.
Vartuli
,
J. C.
Zolper
,
C.
Yuan
, and
R. A.
Stall
,
Appl. Phys. Lett.
67
,
1435
(
1995
).
30.
J. C.
Zolper
,
H. H.
Tan
,
J. S.
Williams
,
J.
Zou
,
D. J. H.
Cockayne
,
S. J.
Pearton
,
M. H.
Crawford
, and
R. F.
Karlicek
,
Appl. Phys. Lett.
70
,
2729
(
1997
).
31.
Y.
Nakano
and
T.
Kachi
,
Appl. Phys. Lett.
79
,
1468
(
2001
).
32.
Y.
Nakano
and
T.
Jimbo
,
J. Appl. Phys.
92
,
3815
(
2002
).
33.
J. L.
Lyons
and
C. G.
Van de Walle
,
Npj Comput. Mater.
3
,
12
(
2017
).
34.
B. N.
Feigelson
,
T. J.
Anderson
,
M.
Abraham
,
J. A.
Freitas
,
J. K.
Hite
,
C. R.
Eddy
, and
F. J.
Kub
,
J. Cryst. Growth
350
,
21
(
2012
).
35.
J. D.
Greenlee
,
B. N.
Feigelson
,
T. J.
Anderson
,
J. K.
Hite
,
K. D.
Hobart
, and
F. J.
Kub
,
ECS J. Solid State Sci. Technol.
4
,
P382
(
2015
).
36.
T.
Narita
,
T.
Kachi
,
K.
Kataoka
, and
T.
Uesugi
,
Appl. Phys. Express
10
,
016501
(
2017
).
37.
H.
Sakurai
,
M.
Omori
,
S.
Yamada
,
Y.
Furukawa
,
H.
Suzuki
,
T.
Narita
,
K.
Kataoka
,
M.
Horita
,
M.
Bockowski
,
J.
Suda
, and
T.
Kachi
,
Appl. Phys. Lett.
115
,
142104
(
2019
).
38.
D.
Tomida
,
Q.
Bao
,
M.
Saito
,
R.
Osanai
,
K.
Shima
,
K.
Kojima
,
T.
Ishiguro
, and
S. F.
Chichibu
,
Appl. Phys. Express
13
,
055505
(
2020
).
39.
K.
Hirukawa
,
K.
Sumida
,
H.
Sakurai
,
H.
Fujikura
,
M.
Horita
,
Y.
Otoki
,
K.
Sierakowski
,
M.
Bockowski
,
T.
Kachi
, and
J.
Suda
,
Appl. Phys. Express
14
,
056501
(
2021
).
40.
T.
Narita
,
A.
Uedono
, and
T.
Kachi
,
Phys. Status Solidi B
(published online, 2022).
41.
J.
Karpinski
,
J.
Jun
, and
S.
Porowski
,
J. Cryst. Growth
66
,
1
(
1984
).
42.
J.
Karpinski
and
S.
Porowski
,
J. Cryst. Growth
66
,
11
(
1984
).
43.
H.
Sakurai
,
T.
Narita
,
K.
Hirukawa
,
S.
Yamada
,
A.
Koura
,
K.
Kataoka
,
M.
Horita
,
N.
Ikarashi
,
M.
Bockowski
,
J.
Suda
, and
T.
Kachi
, “
Impacts of high temperature annealing above 1400 °C under N2 overpressure to activate acceptors in Mg-implanted GaN
,” in
IEEE 32th International Symposium on Power Semiconductor Devices & IC’s (ISPSD)
,
Vienna, Austria
(IEEE,
2020
).
44.
K.
Sierakowski
,
R.
Jakiela
,
B.
Lucznik
,
P.
Kwiatkowski
,
M.
Iwinska
,
M.
Turek
,
H.
Sakurai
,
T.
Kachi
, and
M.
Bockowski
,
Electronics
9
,
1380
(
2020
).
45.
N. L.
Loh
and
K. Y.
Sia
,
J. Mater. Process. Technol.
30
,
45
(
1992
).
46.
T.
Narita
,
H.
Sakurai
,
M.
Bockowski
,
K.
Kataoka
,
J.
Suda
, and
T.
Kachi
,
Appl. Phys. Express
12
,
111005
(
2019
).
47.
H.
Sakurai
,
T.
Narita
,
M.
Omori
,
S.
Yamada
,
A.
Koura
,
M.
Iwinska
,
K.
Kataoka
,
M.
Horita
,
N.
Ikarashi
,
M.
Bockowski
,
J.
Suda
, and
T.
Kachi
,
Appl. Phys. Express
13
,
086501
(
2020
).
48.
A. S.
Barker
, Jr
and
M.
Ilegems
,
Phys. Rev. B
7
,
743
(
1973
).
49.
K.
Sumida
,
K.
Hirukawa
,
H.
Sakurai
,
K.
Sierakowski
,
M.
Horita
,
M.
Bockowski
,
T.
Kachi
, and
J.
Suda
,
Appl. Phys. Express
14
,
121004
(
2021
).
50.
M. A.
Reshchikov
,
D. O.
Demchenko
,
J. D.
McNamara
,
S.
Fernández-Garrido
, and
R.
Calarco
,
Phys. Rev. B
90
,
035207
(
2014
).
51.
K.
Kojima
,
S.
Takashima
,
M.
Edo
,
K.
Ueno
,
M.
Shimizu
,
T.
Takahashi
,
S.
Ishibashi
,
A.
Uedono
, and
S. F.
Chichibu
,
Appl. Phys. Express
10
,
061002
(
2017
).
52.
H.
Sakurai
,
T.
Narita
,
K.
Kataoka
,
K.
Hirukawa
,
K.
Sumida
,
S.
Yamada
,
K.
Sierakowski
,
M.
Horita
,
N.
Ikarashi
,
M.
Bockowski
,
J.
Suda
, and
T.
Kachi
,
Appl. Phys. Express
14
,
111001
(
2021
).
53.
M. H.
Breckenridge
,
J.
Tweedie
,
P.
Reddy
,
Y.
Guan
,
P.
Bagheri
,
D.
Szymanski
,
S.
Mita
,
K.
Sierakowski
,
M.
Bockowski
,
R.
Collazo
, and
Z.
Sita
,
Appl. Phys. Lett.
118
,
022101
(
2021
).
54.
T.
Nakashima
,
E.
Kano
,
K.
Kataoka
,
S.
Arai
,
H.
Sakurai
,
T.
Narita
,
K.
Sierakowski
,
M.
Bockowski
,
M.
Nagao
,
J.
Suda
,
T.
Kachi
, and
N.
Ikarashi
,
Appl. Phys. Express
14
,
011005
(
2021
).
55.
K.
Iwata
,
H.
Sakurai
,
S.
Arai
,
T.
Nakashima
,
T.
Narita
,
K.
Kataoka
,
M.
Bockowski
,
M.
Nagao
,
J.
Suda
,
T.
Kachi
, and
N.
Ikarashi
,
J. Appl. Phys.
127
,
105106
(
2020
).
56.
S. R.
Alugubelli
,
H.
Fu
,
K.
Fu
,
H.
Liu
,
Y.
Zhao
, and
F. A.
Ponce
,
J. Appl. Phys.
126
,
015704
(
2019
).
57.
S.
Nitta
,
Y.
Yukawa
,
Y.
Watanabe
,
M.
Kosaki
,
M.
Iwaya
,
S.
Yamaguchi
,
H.
Amano
, and
I.
Akasaki
,
Mater. Sci. Eng., B
93
,
139
(
2002
).
58.
J. L.
Lyons
,
A.
Janotti
, and
C. G.
Van de Walle
,
Phys. Rev. B
89
,
035204
(
2014
).
59.
M. A.
Reshchikov
,
M.
Vorobiov
,
D. O.
Demchenko
,
Ü
Özgür
,
H.
Morkoç
,
A.
Lesnik
,
M. P.
Hoffmann
,
F.
Hörich
,
A.
Dadgar
, and
A.
Strittmatter
,
Phys. Rev. B
98
,
125207
(
2018
).
60.
A.
Uedono
,
H.
Sakurai
,
J.
Uzuhashi
,
T.
Narita
,
K.
Sierakowski
,
S.
Ishibashi
,
S. F.
Chichibu
,
M.
Bockowski
,
J.
Suda
,
T.
Ohkubo
,
N.
Ikarashi
,
K.
Hono
, and
T.
Kachi
,
Phys. Status Solidi B
(published online, 2022).
61.
J.
Chen
,
W.
Yi
,
T.
Kimura
,
S.
Takashima
,
M.
Edo
, and
T.
Sekiguchi
,
Appl. Phys. Express
12
,
051010
(
2019
).
62.
T.
Maeda
,
T.
Narita
,
H.
Ueda
,
M.
Kanechika
,
T.
Uesugi
,
T.
Kachi
,
T.
Kimoto
,
M.
Horita
, and
J.
Suda
,
IEEE Electron Device Lett.
40
,
941
(
2019
).
63.
K.
Zeng
and
S.
Chowdhury
,
IEEE Trans. Electron Devices
67
,
2457
(
2020
).
64.
T.
Shoji
,
T.
Narita
,
Y.
Nagasato
,
M.
Kanechika
,
T.
Kondo
,
T.
Uesugi
,
K.
Tomita
,
S.
Ikeda
,
T.
Mori
,
S.
Yamaguchi
,
Y.
Kimoto
,
J.
Kojima
, and
J.
Suda
,
Appl. Phys. Express
14
,
114001
(
2021
).
65.
T.
Nishimura
,
K.
Ikeda
, and
T.
Kachi
,
Appl. Phys. Express
14
,
066503
(
2021
).
66.
T.
Nishimura
and
T.
Kachi
,
Appl. Phys. Express
14
,
116502
(
2021
).
67.
K.
Shima
,
R.
Tanaka
,
S.
Takashima
,
K.
Ueno
,
M.
Edo
,
K.
Kojima
,
A.
Uedono
,
S.
Ishibashi
, and
S. F.
Chichibu
,
Appl. Phys. Lett.
119
,
182106
(
2021
).