Deep centers and dislocation densities in undoped n GaN, grown by hydride vapor phase epitaxy (HVPE), were characterized as a function of the layer thickness by deep level transient spectroscopy and transmission electron microscopy, respectively. As the layer thickness decreases, the variety and concentration of deep centers increase, in conjunction with the increase of dislocation density. Based on comparison with electron-irradiation induced centers, some dominant centers in HVPE GaN are identified as possible point defects.

Electron capture behaviours for major traps in thin epitaxial and thick freestanding GaN samples have been experimentally and theoretically studied by using deep-level transient spectroscopy (DLTS). According to the logarithmic dependence of the DLTS signal on the filling pulse width, most of the traps in thin epitaxial GaN layers with high dislocation density behave as line defects. In sharp contrast, the same traps in thick free-standing GaN layers with low dislocation density behave as point defects. The most likely explanation for these phenomena is that the electron traps in question tend to segregate around dislocations, when present in large numbers.

Deep centers in Si-doped n-GaN layers grown by reactive molecular beam epitaxy have been studied by deep-level transient spectroscopy as a function of growth conditions. Si-doped GaN samples grown on a Si-doped n+-GaN contact layer at 800 °C show a dominant trap C1 with activation energy ET=0.44 eV and capture cross-section σT=1.3×10−15 cm−2, while samples grown at 750 °C on an undoped semi-insulating GaN buffer show prominent traps D1 and E1, with ET=0.20 eV and σT=8.4×10−17 cm2, and ET=0.21 eV and σT=1.6×10−14 cm2, respectively. Trap E1 is believed to be related to a N-vacancy defect, since the Arrhenius signature for E1 is very similar to the previously reported trap E, which is produced by 1-MeV electron irradiation in GaN materials grown by both metalorganic chemical-vapor deposition and hydride vapor-phase epitaxy.

AlGaN ∕ GaN ∕ SiC Schottky barrier diodes (SBDs), with and without Si3N4 passivation, have been characterized by temperature-dependent current-voltage and capacitance-voltage measurements, and deep level transient spectroscopy (DLTS). A dominant trap A1, with activation energy of 1.0 eV and apparent capture cross section of 2×10−12cm2, has been observed in both unpassivated and passivated SBDs. Based on the well-known logarithmic dependence of DLTS peak height with filling pulse width for a line-defect related trap, A1, which is commonly observed in thin GaN layers grown by various techniques, is believed to be associated with threading dislocations. At high temperatures, the DLTS signal sometimes becomes negative, likely due to an artificial surface-state effect.

Electrical properties, including leakage currents, threshold voltages, and deep traps, of AlGaN/GaN heterostructure wafers with different concentrations of carbon in the GaN buffer layer, have been investigated by temperature dependent current-voltage and capacitance-voltage measurements and deep level transient spectroscopy ͑DLTS͒, using Schottky barrier diodes ͑SBDs͒. It is found that ͑i͒ SBDs fabricated on the wafers with GaN buffer layers containing a low concentration of carbon ͑low-͓C͔ SBD͒ or a high concentration of carbon ͑high-͓C͔ SBD͒ have similar low leakage currents even at 500 K; and ͑ii͒ the low-͓C͔ SBD exhibits a larger ͑negative͒ threshold voltage than the high-͓C͔ SBD. Detailed DLTS measurements on the two SBDs show that ͑i͒ different trap species are seen in the two SBDs: electron traps A x ͑0.9 eV͒, A 1 ͑0.99 eV͒, and A 2 ͑1.2 eV͒, and a holelike trap H 1 ͑1.24 eV͒ in the low-͓C͔ SBD; and electron traps A 1 , A 2 , and A 3 ͑ϳ1.3 eV͒, and a holelike trap H 2 ͑Ͼ1.3 eV͒ in the high-͓C͔ SBD; ͑ii͒ for both SDBs, in the region close to GaN buffer layer, only electron traps can be detected, while in the AlGaN/GaN interface region, significant holelike traps appear; and iii͒ all of the deep traps show a strong dependence of the DLTS signal on filling pulse width, which indicates they are associated with extended defects, such as threading dislocations. However, the overall density of electron traps is lower in the low-͓C͔ SBD than in the high-͓C͔ SBD. The different traps observed in the two SBDs are thought to be mainly related to differences in microstructure ͑grain size and threading dislocation density͒ of GaN buffer layers grown at different pressures.

We show that the usual Arrhenius analysis of the main electron-irradiation-induced defect trap in n-type GaN, observed by deep-level transient spectroscopy (DLTS), is not sufficiently accurate. Instead, an exact fitting of the DLTS spectrum for this trap reveals two components, each of which has a thermal energy near 60 meV, not the apparent 140–200 meV, as given in other DLTS studies. This result resolves the discrepancy between Hall-effect and DLTS determinations of the thermal energy of this defect center.

Deep-level transient spectroscopy measurements of n-type GaN epitaxial layers irradiated with 1-MeV electrons reveal an irradiation-induced electron trap at EC−0.18 eV. The production rate is approximately 0.2 cm−1, lower than the rate of 1 cm−1 found for the N vacancy by Hall-effect studies. The defect trap cannot be firmly identified at this time.

Schottky barrier diodes, on both Ga and N faces of a ∼300-μm-thick free-standing GaN layer, grown by hydride vapor phase epitaxy (HVPE) on Al2O3 followed by laser separation, were studied by capacitance–voltage and deep level transient spectroscopy (DLTS) measurements. From a 1/C2 vs V analysis, the barrier heights of Ni/Au Schottky contacts were determined to be different for the two polar faces: 1.27 eV for the Ga face, and 0.75 eV for the N face. In addition to the four common DLTS traps observed previously in other epitaxial GaN including HVPE-grown GaN a new trap B′ with activation energy ET=0.53 eV was found in the Ga-face sample. Also, trap E1 (ET=0.18 eV), believed to be related to the N vacancy, was found in the N-face sample, and trap C (ET=0.35 eV) was in the Ga-face sample. Trap C may have arisen from reactive-ion-etching damage.