The development of gallium nitride (GaN) vertical-type metal-oxide-semiconductor field-effect transistors and p–i–n diode devices has gathered increasing attention. These devices require an n-type drift layer with a low doping level of 1016 cm−3 or less, minimized point defects inhibiting electron conduction, and a layer approximately 10 μm thick. Therefore, a practical method with a growth rate of at least several tens of μm/h and impurity concentrations of less than 1015 cm−3, except for that of dopants, is necessary. Halogen-free vapor-phase epitaxy (HF-VPE) has a high growth rate suitable for fabricating thick drift layers and utilizes a simple reaction between Ga vapor and ammonia gas (without a corrosive halogen gas), resulting in lower impurity levels. Herein, we eliminated the quartz content from the high-temperature zone to reduce the excess unintentional Si doping and identified that the nitrile gloves used for the growth preparation are other impurity contamination sources. We obtained a lightly n-type ([Si]=∼1016 cm−3) GaN layer, in which C, O, B, Fe, Mg, Al, Ca, Cr, Zn, Ni, Mn, and Ti impurity contents were below the detection limits of secondary ion mass spectrometry. Deep-level transient spectroscopy revealed that electron traps at EC − 0.26 and at EC − 0.59 eV were 2.7 × 1013 and 5.2 × 1014 cm−3, respectively. Moreover, the Hall effect analysis showed the acceptor-type defect-compensating donor content as approximately 2.7 × 1015 cm−3, resulting in a high electron mobility of HF-VPE GaN in the 30–710 K temperature range. Furthermore, we identified the Ca impurity as a deep acceptor, another killer defect leading to mobility collapse.

Gallium nitride (GaN) is one of the most promising materials for fabricating electronic devices, such as high-frequency and power devices, owing to its wide bandgap (3.4 eV), high electron mobility (>1000 cm2/V s), large critical breakdown electric field (3.3–3.7 MV/cm), and chemical stability.1–7 Recently, the development of high-power vertical-type GaN power devices, such as metal-oxide-semiconductor field-effect transistors (MOSFETs) and fin-channel junction-gate field effect transistors (Fin-JFET), has generated considerable research interest because they can withstand kV-class breakdown voltage with current densities of over a few hundred A/cm2, as in automotive applications.1,2,7–9 The drift layer of vertical-type power devices is designed based on the breakdown voltage; a 1.2 kV-class device generally has a doping concentration and thickness of approximately ∼1016 cm−3 and 10 μm, respectively.1–4,6,7 The total ON resistance of vertical-type power devices is dominantly affected by the channel and drift resistance. Thus, the improved electron mobility and controllable net carrier concentration due to the removal of unintentional acceptor/donor impurity concentration and acceptor-like deep-level trap concentration significantly improve the characteristics of vertical-type GaN power devices. To meet the requirements of GaN layers, such as drift layers, a practical growth method for lightly n-type layers ([Si]: ∼1016 cm−3) with a growth rate of at least several tens of μm/h and unintentional acceptor/donor concentration of less than 1015 cm−3 for controlling the net carrier concentration should be investigated.

A possible approach is growing n-type low-doping layers with high V/III ratios to reduce carbon (C) incorporation in metal–organic chemical vapor deposition (MOCVD).1,7 Because C is the most common impurity, lowering the C impurity concentration in the GaN layer is necessary to remove an acceptor-like deep-level trap and control the precise doping concentration to approximately 1016 cm−3 in the drift layer.1,7 However, the metal–organic Ga source makes reducing the C impurity concentration to less than 1015 cm−3 difficult. In addition, the high-V/III-ratio growth condition of MOCVD results in a low growth rate, which does not yield a thick drift layer.7 

An alternative approach for growing a thick drift layer is to use high-purity hydride vapor-phase epitaxy (HVPE)-grown GaN.10,11 This HVPE method reduces the source zone temperature and adopts opaque quartz to reduce radiation heat transmission and prevent the heating of the stainless steel used in the growth apparatus. Further, the carbon components near the wafer were removed, achieving growth of high-purity ([C] ∼1 × 1014 cm−3) GaN. However, the halogen gas utilized in the HVPE method may increase the maintenance cost due to by-products and may shorten the lifetime of the growth apparatus due to the corrosiveness of the gas. Hence, there remains a need for an alternative cost-effective approach should be explored.

To overcome these challenges, we propose a halogen-free vapor-phase epitaxy (HF-VPE) method for drift layer growth, which employs the simplest reaction to form GaN (Ga + NH3 → GaN + 3/2H2).12–19 The HF-VPE growth method is cost-effective owing to Ga material yield (>20%), by-product-free long-duration growth, and high growth rate (>100 μm/h).16,17 The HF-VPE adopted the cold-wall furnace design using a radio frequency coil used to heat the growth setup, unlike a general HVPE method, which adopted the hot-wall furnace design (supplementary material). It prevents the unintentional Si and other impurities from doping from the growth apparatus because the high-temperature zone is limited to RF-heated apparatus such as high-purity chemical vapor reaction TaC-coated graphite crucibles. In addition, the halogen-gas-free reaction has an advantage in the growth apparatus without corrosion, which may provide a lower impurity concentration than that of other growth methods. Therefore, the realization of high-purity growth based on the HF-VPE method comparable to HVPE-GaN should be attractive. However, the HF-VPE-grown GaN contains high unintentional Si content (>1017 cm−3), possibly from the heating of quartz parts.14 Moreover, the previously reported electron mobility results of HF-VPE-grown GaN were lower than those of HVPE samples, which implied the possibility of large concentrations of other impurities;14 therefore, the usefulness of HF-VPE in drift layer growth in vertical-type power devices was questionable.

In this study, we identified the sources of Si and other impurities. Most unintentional impurities attached to the growth setup came from elution from the nitrile glove by the organic solvent used in maintenance (details shown in the supplementary material). We reduced the unintentionally doped Si concentrations to 1 × 1016 cm−3 using a polyethylene glove instead of a nitrile glove and removing the quartz components in the high-temperature zone of the HF-VPE growth furnace. The concentrations of other impurities in the HF-VPE layers were reduced to less than 1015 cm−3, resulting in HF-VPE GaN layers with high electron mobilities. The impurity concentrations in the HF-VPE GaN layers were determined using secondary ion mass spectrometry (SIMS). We also conducted deep-level transient spectroscopy (DLTS), cathodoluminescence (CL) measurements, and positron annihilation spectroscopy (PAS) to verify the deep-level trap concentrations, vacancy-type defects, and nonradiative recombination centers (NRCs). We also performed van der Pauw–Hall measurements to measure the electron mobility.

GaN samples were prepared using HF-VPE. A commercially available freestanding GaN substrate and undoped 2 μm-thick MOCVD-GaN on sapphire (MO templates) were employed as substrates. We compared the characteristics of two groups of low-purity (LP) and high-purity (HP) HF-VPE-grown layers. LP HF-VPE GaN layers were grown by referring to the method reported in a previous study.14 To decrease the concentration of unintentional Si and other impurities in the growth layers of HP HF-VPE GaN, we removed the quartz components in the high-temperature zone and carefully handled the growth furnace components using contamination-free polyethylene gloves. The growth conditions of HF-VPE and growth substrates of the HF-VPE GaN samples are summarized in the supplementary material. The growth rates of these samples were 13–45 μm/h, that is, almost ten times higher than that of MOCVD and sufficiently high to grow thick drift layers.

The impurity concentrations of all HF-VPE samples in this study are summarized in Table I. The concentrations of C (≤5 × 1013 cm−3), B (≤8 × 1014 cm−3), O (≤1 × 1014 cm−3), Fe (≤1 × 1015 cm−3), Mg (≤5 × 1013 cm−3), Al (≤5 × 1013 cm−3), Ca (≤3 × 1013 cm−3), Cr (≤7 × 1013 cm−3), Zn (≤3 × 1016 cm−3), Ni (≤2 × 1015 cm−3), Mn (≤7 × 1013 cm−3), and Ti (≤5 × 1013 cm−3) in HP-1 were equal to or less than the background or detection limit levels except for the unintentionally doped Si donor concentration. These results provide compelling evidence that HF-VPE is a high-purity-GaN growth method.

TABLE I.

Impurity concentrations in HF-VPE GaN, determined using SIMS. HF-VPE GaN layers were categorized as low-purity and high-purity HF-VPE samples. The impurity concentration of the MOCVD-GaN layer, measured in the CL analysis, was also determined.

Sample name Si (cm−3) C (cm−3) O (cm−3) Fe (cm−3) Ca (cm−3) Mg (cm−3) Al (cm−3) Cr (cm−3)
LP-1 (on semi-insulating HVPE)  4 × 1016  5 × 1013 (RC)  2 × 1015 (RC)  5 × 1015  1 × 1016  2 × 1015  2 × 1015  Not measured 
LP-2 (on n-type HVPE)  2 × 1017  ≤ 8 × 1015  ≤ 6 × 1015  1 × 1017  2 × 1014  1 × 1015  2 × 1016  9 × 1015 
LP-3 (on MO template)  3 × 1016  ≤8 × 1015  ≤3 × 1016  3 × 1015  2 × 1016  5 × 1014  Not measured  Not measured 
HP-1 (on semi-insulating HVPE)  2 × 1016  ≤ 5 × 1013 (RC)  ≤ 1 × 1014 (RC)  ≤ 1 × 1015  ≤ 3 × 1013  ≤ 5 × 1013  ≤ 5 × 1013  ≤ 7 × 1013 
HP-2 (on n-type HVPE)  1 × 1016  ≤ 8 × 1015  ≤ 6 × 1015  ≤ 3 × 1014  ≤ 1 × 1013  ≤ 1 × 1014  Not measured  ≤ 5 × 1013 
HP-3 (on MO template)  3 × 1016  ≤3 × 1015  ≤6 × 1015  1 × 1015  1 × 1015  5 × 1014  Not measured  Not measured 
MOCVD-GaN (on HVPE)  4 × 1016  6 × 1015  ≤ 6 × 1015  ≤ 3 × 1014  ≤ 2 × 1013  ≤ 8 × 1013  Not measured  Not measured 
Sample name Si (cm−3) C (cm−3) O (cm−3) Fe (cm−3) Ca (cm−3) Mg (cm−3) Al (cm−3) Cr (cm−3)
LP-1 (on semi-insulating HVPE)  4 × 1016  5 × 1013 (RC)  2 × 1015 (RC)  5 × 1015  1 × 1016  2 × 1015  2 × 1015  Not measured 
LP-2 (on n-type HVPE)  2 × 1017  ≤ 8 × 1015  ≤ 6 × 1015  1 × 1017  2 × 1014  1 × 1015  2 × 1016  9 × 1015 
LP-3 (on MO template)  3 × 1016  ≤8 × 1015  ≤3 × 1016  3 × 1015  2 × 1016  5 × 1014  Not measured  Not measured 
HP-1 (on semi-insulating HVPE)  2 × 1016  ≤ 5 × 1013 (RC)  ≤ 1 × 1014 (RC)  ≤ 1 × 1015  ≤ 3 × 1013  ≤ 5 × 1013  ≤ 5 × 1013  ≤ 7 × 1013 
HP-2 (on n-type HVPE)  1 × 1016  ≤ 8 × 1015  ≤ 6 × 1015  ≤ 3 × 1014  ≤ 1 × 1013  ≤ 1 × 1014  Not measured  ≤ 5 × 1013 
HP-3 (on MO template)  3 × 1016  ≤3 × 1015  ≤6 × 1015  1 × 1015  1 × 1015  5 × 1014  Not measured  Not measured 
MOCVD-GaN (on HVPE)  4 × 1016  6 × 1015  ≤ 6 × 1015  ≤ 3 × 1014  ≤ 2 × 1013  ≤ 8 × 1013  Not measured  Not measured 
a

The symbol “≤” means the impurity concentration in the sample is less than the SIMS detection limit or the background level.

b

The “RC” indicates that the concentration was determined using the SIMS raster change method.

To evaluate the electron and hole trap concentrations after the reduction of impurities, DLTS analysis was conducted on the LP-2 and HP-2 samples.20 DLTS revealed electron traps at an energy of 0.26 eV below the conduction band minimum (EC−0.26 eV) and at EC−0.59 eV (labeled E1 and E3) in both the LP-2 and HP-2 samples (Fig. 1). The calculated trap concentrations of E1 and E3 were 3.2 × 1016 and 3.0 × 1016 cm−3 for LP-2 and 2.7 × 1013 and 5.2 × 1014 cm−3 for HP-2, respectively (details shown in the supplementary material). Furthermore, minority carrier transient spectroscopy did not reveal the hole trap H1 identified as CN21 at the energy of 0.88 eV above the valence band maximum. This result confirms that the HF-VPE GaN growth method can achieve C-free GaN growth.

FIG. 1.

E1 and E3 electron trap densities in LP-2 and HP-2.

FIG. 1.

E1 and E3 electron trap densities in LP-2 and HP-2.

Close modal

To clarify the relationship between the impurity concentrations, NRCs, and deep-level defects in HF-VPE samples with different impurity concentrations grown on HVPE freestanding substrates, CL measurements were conducted at 10 K. The CL spectra of the band edge and entire region of these samples are shown in Figs. 2(a) and 2(b), respectively. The CL spectra of Si-doped MOCVD-grown GaN ([Si] = 4 × 1016 cm−3) on an HVPE freestanding substrate are also presented for comparison. Figure 2(a) shows that the CL spectra of most of the samples have peaks at the same energies (3.48, 3.47, 3.39, and 3.30 eV), which are attributable to the free exciton (FX), donor-bound exciton (DBE), and longitudinal optical (LO) phonon replicas of FX.22 In addition, the LP-1 and HP-1 samples have other peaks at 3.28 and 3.19 eV, which may be attributed to the donor–acceptor pair (DAP).22 Although the unintentionally doped impurity concentration in LP-1 was at most 1 × 1016 cm−3, the near-band edge (NBE) integrated intensity was only approximately 1/2, and the DAP peak intensity doubled when compared with that of HP-1.

FIG. 2.

CL spectra of band edge and entire region of (a) HF-VPE GaN layers (LP-1, HP-1, and LP-2) and (b) MOCVD-GaN measured at 10 K. (c) Dependence of the integrated value of near-band edge (2.85 − 4.0 eV) CL intensities of each sample on the total impurity concentrations, except for that of the Si donor impurity.

FIG. 2.

CL spectra of band edge and entire region of (a) HF-VPE GaN layers (LP-1, HP-1, and LP-2) and (b) MOCVD-GaN measured at 10 K. (c) Dependence of the integrated value of near-band edge (2.85 − 4.0 eV) CL intensities of each sample on the total impurity concentrations, except for that of the Si donor impurity.

Close modal

The CL spectrum of HP-1 had the highest CL intensity among the samples at the band edge region and no deep-level defect-related emission peaks. The impurity concentration in the HP-1 layer did not exceed the detection limit of SIMS, except for the Si concentration of 2.0 × 1016 cm−3. The origin of DAP peaks in HP-1 is not clear. Figure 2(c) shows the dependence of the integrated intensities of the NBE region on the total impurity (except for Si donor) concentrations of those samples. The integrated intensity of the NBE region of HP-1 was approximately twice that of the other samples with similar Si impurity concentrations, as shown in Fig. 2(c), indicating the lowest concentrations of NRCs and deep-level defects. This is compelling evidence that reducing unintentional impurities minimizes the concentrations of NRCs and deep-level defects.

For the MOCVD-GaN (HVPE substrate for the MOCVD-GaN layer has almost the same dislocation density as that of the HVPE substrate for the HP-1 GaN layer, see the supplementary material), the NBE integrated intensity is approximately 1/2 that of HP-1, and broad deep-level defect-related emission peaks are observable around 2.20 eV, which are attributable to yellow luminescence (YL).22 The MOCVD-GaN layer contains a higher C impurity concentration (6 × 1015 cm−3) than that of HP-1 (<5 × 1013 cm−3). YL has been commonly observed in MOCVD-grown samples; many researchers have reported that YL originates from the incorporation of C at N sites (CN) in GaN.22–24 This suggests that the C impurity noticeably reduces the NBE CL intensity via CN deep levels. Evidently, low-concentration contamination (∼1 × 1016 cm−3) of unintentionally doped impurities in GaN considerably reduces the NBE intensity of the CL spectra.

Next, PAS analyses were conducted to evaluate the number of point defects and NRCs (NNRC) in the LP-1, HP-1, and LP-2 layers using different methods.25–27 Chichibu et al. reported that the dominant intrinsic-vacancy-type defect is divacancy VGaVN comprising a Ga vacancy (VGa) and an N vacancy (VN) in the n-type GaN; moreover, VGaVN is a major intrinsic NRC in n-type GaN.28 The S values for LP-1, HP-1, and LP-2 in the bulk region are equal to or lower than the defect-free GaN S value of 0.442 (see the supplementary material). These results suggest that the electroneutral or negatively charged vacancy-type defect densities of all samples were lower than the detection limit of PAS for vacancy-type defects (<1015 cm−3).25–27 Chichibu et al. also reported that the diffusion lengths of positrons (Ld) corresponded to the inverse third root of NNRC.28, Ld derived from the SE curves of each sample is shown in Fig. 3. The Ld value of HP-1 was 115 ± 2 nm, which is comparable to that of the reported high-quality Na-flux and HVPE freestanding samples.26,28,29 The calculated NNRC value in HP-1 was 6.5 × 1014 cm−3. By contrast, the Ld values of the LP-1 and LP-2 layers were 76 ± 2 and 19 ± 1 nm, respectively. Thus, the calculated NNRC values in the LP-1 and LP-2 layers were 2.3 × 1015 and 1.4 × 1017 cm−3, respectively. The obtained variation in the NNRC values corresponded to that of the impurity concentrations of these samples, suggesting that the change in the diffusion length of the positrons was mainly owing to the scattering of positrons by impurities in the HV-VPE GaN layers. In addition, the vacancy-type defect concentrations and NNRC of HP-1 were less than 1015 cm−3, which is consistent with the SIMS, DLTS, and CL measurements.

FIG. 3.

Diffusion lengths of positrons derived from the SE curves for LP-1, HP-1, and LP-2 samples. The NNRC values represent the calculated concentrations of NRCs.28 

FIG. 3.

Diffusion lengths of positrons derived from the SE curves for LP-1, HP-1, and LP-2 samples. The NNRC values represent the calculated concentrations of NRCs.28 

Close modal

Finally, to elucidate the effects of impurity reduction and the resulting decrease in electron and hole traps, vacancy-type defects, and NRCs on the Hall mobility of the GaN layer, van der Pauw–Hall measurements were conducted in the temperature range of 30–710 K. The dependence of the Hall mobility μH on the carrier concentration n at 300 K for various unintentional impurity concentrations in the HF-VPE-GaN layers is shown in Fig. 4(a).

FIG. 4.

(a) Dependence of Hall mobility μH on carrier concentration for GaN layers at 300 K. The red- and blue-shaded rectangles and circles represent the Hall mobilities of the HF-VPE GaN layers grown on HVPE freestanding substrates and an MO template, respectively. The green rectangle, purple rhombus, and black rectangles denote the values reported by Fujikura et al.,11,30 Sawada et al.,7 and Ohshima et al.,31 respectively, and the black open circles represent those reported by Nakamura et al.32,33 (b) Carrier concentration as a function of temperature in the range of 30–710 K for LP-1 and HP-1. (c) and (d) Dependence of Hall mobility μH on temperature for LP-1 and HP-1. The dashed lines represent the total fitting curve for Hall mobility.

FIG. 4.

(a) Dependence of Hall mobility μH on carrier concentration for GaN layers at 300 K. The red- and blue-shaded rectangles and circles represent the Hall mobilities of the HF-VPE GaN layers grown on HVPE freestanding substrates and an MO template, respectively. The green rectangle, purple rhombus, and black rectangles denote the values reported by Fujikura et al.,11,30 Sawada et al.,7 and Ohshima et al.,31 respectively, and the black open circles represent those reported by Nakamura et al.32,33 (b) Carrier concentration as a function of temperature in the range of 30–710 K for LP-1 and HP-1. (c) and (d) Dependence of Hall mobility μH on temperature for LP-1 and HP-1. The dashed lines represent the total fitting curve for Hall mobility.

Close modal

μH of HP-1 (grown on freestanding HVPE substrate) and HP-3 (grown on MO template) [closed red rectangle and circle, respectively, in Fig. 4(a)] were 1158 and 793 cm2/V s at 300 K, respectively, at the Si concentration of approximately 2 × 1016 cm−3; these values are higher than other reported GaN mobilities for similar carrier concentrations and nearly match the values reported by Sawada et al.7 Furthermore, these samples were obtained with a growth rate of over 30 μm/h, which is suitable for growing a thick drift layer, using the practical HF-VPE growth method.

Conversely, most of the HF-VPE GaN layers on the MO template had considerably lower μH [blue closed triangles within the dashed black oval in Fig. 4(a)] than that of HP-3 despite similar Si concentrations. The μH values of LP-1 and LP-3 (grown on the MO template) were 741 and 213 cm2/V s at Si concentrations of 6 × 1016 and 3 × 1016 cm−3, respectively, which were approximately 1/2–1/5 those of HP-1 and HP-3, respectively. These results are qualitatively similar to those of GaN layers with low donor concentrations, as reported by Kaess et al.34 and Fujikura et al.,11,30 where mobility collapse may have been caused by acceptor (carbon)-impurity-related compensation in the lightly n-doped GaN layers. These reports explained that C-impurity-related compensation reduces the net free electron concentration and increases ionized impurity scattering, considerably deteriorating the electron mobility of the samples. However, as shown in Table I, LP-1 and LP-3 did not contain C impurities. Shen et al. reported that Ca impurities act as acceptors in GaN;35 further, these samples had a Ca impurity concentration of ∼2 × 1016 cm−3, which was almost equivalent to the donor impurity concentration.

The dependence of the carrier concentrations of LP-1 and HP-1 on temperature is shown in Fig. 4(b). The temperature dependencies of the carrier concentrations, donor concentration (Nd), and compensated acceptor concentration (Na) were obtained from Hall measurement results and by fitting with the charge neutrality equation based on semiconductor statics.36, Nd and Na for LP-1 were 8.9 × 1016 and 2.6 × 1016 cm−3, respectively; the Na value was comparable to the Ca impurity concentration in LP-1 (1 × 1016 cm−3). According to Shen et al., Ca potentially acts as a deep acceptor in n-GaN as per their first principal calculation.35 Our results provide compelling experimental evidence corroborating that Ca impurities function as deep acceptors and represent one of the sources leading to mobility collapse. Conversely, the carrier concentration n of HP-1 decreased exponentially as temperature decreased, which was in good agreement with the fitting curves. (Future studies should clarify the absence of constant carrier concentration above 470 K region of LP-1 and HP-1.) The calculated Nd for HP-1 was 2.2 × 1016 cm−3, which matched the Si concentration in HP-1, and the calculated Na of HP-1 was significantly low at 2.7 × 1015 cm−3, which was 1/10 that of LP-1. The calculated Na of the HP-1 sample was slightly larger than the impurity concentrations determined by SIMS ([C]< 5 × 1013 cm−3, [Fe]< 1 × 1015 cm−3, [Ca]< 3 × 1013 cm−3) and other defect densities derived from DLTS (E3: 5 × 1014 cm−3) and PAS (NNRC < 6.5 × 1014 cm−3). The cause of this discrepancy between the Na value and the measurement results is not yet entirely clear. The cause might be attributed to intrinsic crystal defects that act as acceptors within the HP-1 layer.

LP-1 had local minimum values of n at 60 K, which increased as the temperature decreased, eventually saturating at low temperatures. These results are qualitatively similar to the reported hopping conductivity observed in low-temperature regions.1,37–39 Hopping conduction is related to the tunnel effect of localized electrons from nearby impurity elements to other impurity elements. This localization of electrons must originate from impurities; therefore, the behavior of n of LP-1 could be attributed to the high impurity concentrations in these samples.

Figures 4(c) and 4(d) show the temperature dependence of μH for LP-1 and HP-1, respectively. μH behavior significantly differs in the low-temperature region. The μH value of LP-1 peaks at 1916 cm2/V s at 119 K and declines with a further increase in the temperature. The μH value of HP-1 peaks at 4809 cm2/V s at 94 K; the peak value is more than twice that of LP-1. Furthermore, in contrast to the μH of LP-1, that of HP-1 is almost constant in the high range of 4625–4809 at 50−120 K. This μH value of HP-1 was slightly low compared with the reported HVPE-grown μH value (7386 cm2/V s at 48 K) at having similar Nd and Na concentrations.40 However, the consistently high μH of HP-1 may be attributed to the low acceptor compensation and absence of hopping conduction even in the low-temperature regions.

Various scattering mechanisms (ionized impurity, neutral impurity, piezoelectric, optical phonon, acoustic-deformation-potential, and dislocation scattering) and theoretical mobilities were considered by fitting the parameters of Nd, Na, and donor ionization energy.36 The physical constants and formulas used for the theoretical calculation were referred from the data reported by Kyle et al.36 and Sawada et al.7 The effect of dislocation scattering on the electron mobilities of LP-1 and HP-1 was weak owing to low dislocation densities (∼107 cm−2) being negligible. The μH value of HP-1 showed good agreement with the theoretical fitting results over the entire temperature range (30–710 K) at Nd: 2.2 × 1016 and Na: 2.7 × 1015 cm−3 of HP-1. The theoretical maximum electron mobility Nd: 2.2 × 1016 and Na: 0 cm−3 is approximately 15 000 cm2/V s; therefore, the electron mobility of HP-1 still has room for improvement by reducing Na value. These results demonstrate that reducing the impurity concentration, which decreases the number of electron/hole traps and compensated acceptors (Na) in the GaN layer, is essential for obtaining the theoretical high Hall mobility.

We demonstrated that low-impurity lightly doped n-type layers grown by the HF-VPE GaN method exhibit high μH in the entire temperature range of Hall mobility owing to the low concentrations of electron/hole traps, NRCs, vacancy-type defects, and compensated acceptors. The results prove that the HF-VPE growth method provides a straightforward, cost-effective, and high-purity GaN layer suitable for fabricating devices.

See the supplementary material for the growth conditions of the HF-VPE GaN layers, deep-level transient spectroscopy (DLTS) analysis data, Cr and Ni impurity concentrations in LP-2, x-ray rocking curves and estimated dislocation densities, PAS analysis, calculated Nd and Na, and Hall mobility measurement results of the HF-VPE-GaN layers grown on MO templates.

The study was partially supported by MEXT “Program for research and development of next-generation semiconductor to realize energy-saving society” (Program Grant No. JPJ005357). We thank H. Ueda and S. Ito for their support in conducting the DLTS experiments.

The authors have no conflicts to disclose.

Taishi Kimura: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (lead); Validation (equal); Writing – original draft (lead); Writing – review & editing (equal). Hiroshi Amano: Investigation (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). Daisuke Nakamura: Investigation (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). Hiroki Shimazu: Data curation (equal); Investigation (equal); Writing – review & editing (equal). Keita Kataoka: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Kenji Itoh: Data curation (equal); Investigation (equal); Writing – review & editing (equal). Tetsuo Narita: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Akira Uedono: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Yutaka Tokuda: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Daiki Tanaka: Data curation (equal); Investigation (equal); Writing – review & editing (equal). Shugo Nitta: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (equal).

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

1.
T.
Narita
,
K.
Tomita
,
K.
Kataoka
,
Y.
Tokuda
,
T.
Kogiso
,
H.
Yoshida
,
N.
Ikarashi
,
K.
Iwata
,
M.
Nagao
,
N.
Sawada
,
M.
Horita
,
J.
Suda
, and
T.
Kachi
,
Jpn. J. Appl. Phys., Part 1
59
,
SA0804
(
2020
).
2.
T.
Kachi
,
Jpn. J. Appl. Phys., Part 1
53
,
100210
(
2014
).
3.
M.
Kanechika
,
M.
Sugimoto
,
N.
Soejima
,
H.
Ueda
,
O.
Ishiguro
,
M.
Kodama
,
E.
Hnyashi
,
K.
Itoh
,
T.
Uesugi
, and
T.
Kachi
,
Jpn. J. Appl. Phys., Part 2
46
,
L503
(
2007
).
4.
M.
Kodama
,
M.
Sugimoto
,
E.
Hayashi
,
N.
Soejima
,
O.
Ishiguro
,
M.
Kanechika
,
K.
Itoh
,
H.
Ueda
,
T.
Uesugi
, and
T.
Kachi
,
Appl. Phys. Express
1
,
021104
(
2008
).
5.
B. J.
Baliga
,
Semicond. Sci. Technol.
28
,
074011
(
2013
).
6.
T.
Oka
,
T.
Ina
,
Y.
Ueno
, and
J.
Nishii
,
Appl. Phys. Express
8
,
054101
(
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.
J. C.
Liu
,
M.
Xiao
,
R. Z.
Zhang
,
S.
Pidaparthi
,
H.
Cui
,
A.
Edwards
,
M.
Craven
,
L.
Baubutr
,
C.
Drowley
, and
Y. H.
Zhang
,
IEEE Trans. Electron Devices
68
(
4
),
2025
(
2021
).
9.
J.
Liu
,
M.
Xiao
,
Y.
Zhang
,
S.
Pidaparthi
,
H.
Cui
,
A.
Edwards
,
L.
Baubutr
,
W.
Meier
,
C.
Coles
, and
C.
Drowley
, in
IEEE International Electron Devices Meeting (IEDM)
,
2020
.
10.
T.
Kimura
,
T.
Konno
, and
H.
Fujikura
,
Appl. Phys. Lett.
118
,
182104
(
2021
).
11.
H.
Fujikura
,
T.
Konno
,
T.
Kimura
,
Y.
Narita
, and
F.
Horikiri
,
Appl. Phys. Lett.
117
,
012103
(
2020
).
12.
T.
Kimura
,
Y.
Aoki
,
K.
Horibuchi
, and
D.
Nakamura
,
J. Appl. Phys.
120
,
245703
(
2016
).
13.
T.
Kimura
,
K.
Horibuchi
,
K.
Kataoka
, and
D.
Nakamura
,
J. Cryst. Growth
494
,
17
(
2018
).
14.
T.
Kimura
,
K.
Kataoka
,
A.
Uedono
,
H.
Amano
, and
D.
Nakamura
,
Appl. Phys. Express
13
,
085509
(
2020
).
15.
T.
Kimura
,
S.
Sato
,
K.
Kataoka
,
T.
Morikawa
, and
D.
Nakamura
,
ACS Appl. Mater. Interfaces
11
,
4233
(
2019
).
16.
D.
Nakamura
and
T.
Kimura
,
Appl. Phys. Express
10
,
095503
(
2017
).
17.
D.
Nakamura
and
T.
Kimura
,
Appl. Phys. Express
11
,
065502
(
2018
).
18.
D.
Nakamura
,
T.
Kimura
, and
K.
Horibuchi
,
Appl. Phys. Express
10
,
045504
(
2017
).
19.
T.
Kimura
,
M.
Murase
,
Y.
Yamada
,
N.
Mizoshita
, and
D.
Nakamura
,
Nanoscale Adv.
4
,
3718
(
2022
).
20.
Y.
Tokuda
, in
CS ManTech Conference Digest
,
2014
, Vol.
19
.
21.
T.
Narita
,
Y.
Tokuda
,
T.
Kogiso
,
K.
Tomita
, and
T.
Kachi
,
J. Appl. Phys.
123
,
161405
(
2018
).
22.
M. A.
Reshchikov
and
H.
Morkoc
,
J. Appl. Phys.
97
,
061301
(
2005
).
23.
M. A.
Reshchikov
,
Phys. Status Solidi B
2022
,
2200488
.
24.
J. L.
Lyons
,
E. R.
Glaser
,
M. E.
Zvanut
,
S.
Paudel
,
M.
Iwinska
,
T.
Sochacki
, and
M.
Bockowski
,
Phys. Rev. B
104
,
075201
(
2021
).
25.
A.
Uedono
,
S. F.
Chichibu
,
Z. Q.
Chen
,
M.
Sumiya
,
R.
Suzuki
,
T.
Ohdaira
,
T.
Mikado
,
T.
Mukai
, and
S.
Nakamura
,
J. Appl. Phys.
90
,
181
(
2001
).
26.
A.
Uedono
,
M.
Imanishi
,
M.
Imade
,
M.
Yoshimura
,
S.
Ishibashi
,
M.
Sumiya
, and
Y.
Mori
,
J. Cryst. Growth
475
,
261
(
2017
).
27.
S. F.
Chichibu
,
A.
Uedono
,
T.
Onuma
,
T.
Sota
,
B. A.
Haskell
,
S. P.
DenBaars
,
J. S.
Speck
, and
S.
Nakamura
,
Appl. Phys. Lett.
86
,
021914
(
2005
).
28.
S. F.
Chichibu
,
A.
Uedono
,
K.
Kojima
,
H.
Ikeda
,
K.
Fujito
,
S.
Takashima
,
M.
Edo
,
K.
Ueno
, and
S.
Ishibashi
,
J. Appl. Phys.
123
,
161413
(
2018
).
29.
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
259
,
2200183
(
2022
).
30.
H.
Fujikura
,
T.
Konno
,
T.
Yoshida
, and
F.
Horikiri
,
Jpn. J. Appl. Phys., Part 1
56
,
085503
(
2017
).
31.
Y.
Oshima
,
T.
Yoshida
,
T.
Eri
,
M.
Shibata
, and
T.
Mishima
,
Jpn. J. Appl. Phys., Part 1
45
,
7685
(
2006
).
32.
S.
Nakamura
,
Jpn. J. Appl. Phys., Part 2
30
,
L1705
(
1991
).
33.
S.
Nakamura
,
T.
Mukai
, and
M.
Senoh
,
Jpn. J. Appl. Phys., Part 1
31
,
2883
(
1992
).
34.
F.
Kaess
,
S.
Mita
,
J. Q.
Xie
,
P.
Reddy
,
A.
Klump
,
L. H.
Hernandez-Balderrama
,
S.
Washiyama
,
A.
Franke
,
R.
Kirste
,
A.
Hoffmann
,
R.
Collazo
, and
Z.
Sitar
,
J. Appl. Phys.
120
,
105701
(
2016
).
35.
J. X.
Shen
,
D.
Wickramaratne
,
C. E.
Dreyer
,
A.
Alkauskas
,
E.
Young
,
J. S.
Speck
, and
C. G.
Van de Walle
,
Appl. Phys. Express
10
,
021001
(
2017
).
36.
E. C. H.
Kyle
,
S. W.
Kaun
,
P. G.
Burke
,
F.
Wu
,
Y. R.
Wu
, and
J. S.
Speck
,
J. Appl. Phys.
115
,
193702
(
2014
).
37.
S.
Ji
,
K.
Eto
,
S.
Yoshida
,
K.
Kojima
,
Y.
Ishida
,
S.
Saito
,
H.
Tsuchida
, and
H.
Okumura
,
Appl. Phys. Express
8
,
121302
(
2015
).
38.
Y.
Kajikawa
,
Phys. Status Solidi C
14
,
1600129
(
2017
).
39.
A.
Wolos
,
Z.
Wilamowski
,
M.
Piersa
,
W.
Strupinski
,
B.
Lucznik
,
I.
Grzegory
, and
S.
Porowski
,
Phys. Rev. B
83
,
165206
(
2011
).
40.
D.
Huang
,
F.
Yun
,
M. A.
Reshchikov
,
D.
Wang
,
H.
Morkoç
,
D. L.
Rode
,
L. A.
Farina
,
Ç.
Kurdak
,
K. T.
Tsen
,
S. S.
Park
, and
K. Y.
Lee
,
Solid·State Electron.
45
(
5
),
711
(
2001
).

Supplementary Material