Annealing Mg-implanted homoepitaxial GaN at temperatures above 1400 °C eliminates the formation of inversion domains and leads to improved dopant activation efficiency. Extended defects, in the form of inversion domains, contain electrically inactive Mg after post-implantation annealing at temperatures as high as 1300 °C (one GPa N2 overpressure), which results in a low dopant activation efficiency. Triple-axis x-ray data reveal that implant-induced strain is fully relieved after annealing at 1300 °C for 10 min, indicating that strain-inducing point defects formed during implantation have reconfigured and inversion domains are formed. However, annealing at temperatures of 1400–1500 °C (one GPa N2 overpressure) eliminates the presence of the inversion domains. While residual defects, such as dislocation loops, still exist after annealing at and above 1400 °C, chemical analysis at multiple dislocation loops shows no sign of Mg segregation. Meanwhile, an overall decreasing trend in the dislocation loop density is observed after annealing at the higher temperatures and longer times. Additionally, once inversion domains are formed and the samples are cooled to room temperature, they are shown to dissolve with subsequent annealing above 1400 °C. While such defects have been observed before, the important finding that such defects can be dissolved with a short, higher temperature step is key. Earlier work [Breckenridge et al., J. Appl. Phys. Lett. 118, 022101 (2021)] addressing electrical measurements of these types of samples showed that annealing at 1400 °C leads to a dopant activation efficiency that is an order of magnitude higher than that observed at 1300 °C. This work complements earlier work by identifying the inversion domains, which incorporate Mg, and points to the benefits, in terms of defect density and p-type dopant activation, of using higher temperature (>1400 °C) annealing cycles to activate Mg in GaN, even if the Mg-containing inversion domains had been formed during lower temperature annealing.

Vertical GaN power devices have emerged as promising candidates for next-generation high power applications due to superior material properties such as high breakdown voltage, low on-resistance, and high mobility compared to devices based on Si and SiC.1,2 GaN-based power switches offer higher voltage power with significantly higher efficiencies at a much smaller form factor. A technological limitation of GaN-based vertical devices (as well as other GaN devices that require high p-type doping concentrations) has been the inability to achieve planar high p-type doping. Ion implantation has been very effective in selective area p-type doping for Si- and SiC-based technology.3,4 However, p-type selective area doping with ion implantation in GaN remains a challenging task.5–7 

The activation of the implanted p-type dopants requires high temperature annealing. The targeted annealing temperature for GaN, based on the experience from other semiconductors, is ∼two third the melting point, corresponding to a temperature of ∼1500 °C. However, GaN is thermodynamically unstable at T > 850 °C under atmospheric pressure. Various efforts have aimed to avoid GaN decomposition during high temperature processing, such as using dielectric capping materials (e.g., AlN) with a rapid heating source8–10 or applying high N2 pressure without using any caps.11–13 After annealing, residual defects, such as dislocation loops, inversion domains, and stacking faults, have been observed in the implanted region.14–16 It is believed that these defects form from the point defects that are introduced through implantation and that different defects are stable at different temperatures. For example, we have shown earlier that stacking faults exist after a short time (10 min) annealing at 1300 °C but are not present after extended annealing (100 min). Dislocation loops are stable at the same temperature.17 After implantation and annealing at 1300 °C (10 min), cubic domains (∼10 nm regions embedded in the wurtzite lattice) are also present,16 which indicates a significant amount of lattice reconfiguration/rearrangement occurs at 1300 °C.

A high quality homoepitaxial structure provides important benefits to study the evolution of the extended defects. Due to the lack of inexpensive high quality and large size native substrate, GaN epitaxial layers for prior p-type ion implantation studies were typically grown on highly mismatched sapphire substrates. Our earlier work showed a large number of defects, partially due to the high intrinsic defect density (>108 cm−2) in such heteroepitaxial structures, remained after the activation anneal, which limits subsequent device performance.18 The recent developments in bulk growth techniques such as hydride vapor phase epitaxy and ammonothermal growth have led to bulk GaN substrates with substantially lower defect density (<106 cm−2).19,20 Using homoepitaxial GaN layers grown on ammonothermal substrates, we have developed x-ray topography techniques and combined these with electron microscopy and high-resolution x-ray diffraction to understand how implant-induced strain and defects evolve during subsequent high temperature processing steps.16 The next step, discussed here, is to understand how these extended defects affect dopant activation. For example, in the current literature, Mg segregation at inversion domains has been observed in highly Mg-doped epitaxial layers with Mg concentration above the 1019 cm−3 level.21–23 It was found that inversion domains play a role in the free carrier reduction in highly Mg-doped GaN. More recently, a few studies have focused on Mg-rich defects in Mg-implanted GaN after annealing. Kumar et al. showed that Mg-rich clusters started to form in Mg-implanted GaN after annealing at 1100 °C and remained after annealing at ∼1350 °C.24,25 Iwata et al. further discovered Mg segregation at the boundary of the inversion domains in Mg-implanted GaN after annealing at ∼1300 °C.16 Interestingly, these studies showed evidence of Mg segregation in extended defects after annealing at approximately the same temperature. However, the question remains how these inversion domains affect the dopant activation efficiency and if they remain stable after annealing at higher temperatures and/or for longer times. In our earlier study, post-implantation annealing at 1300 °C for 10 and 100 min results in changes in both defect concentration and defect structures.16 Therefore, when the annealing temperature approaches ∼two third the GaN melting point, it is expected that there would be further lattice rearrangement and reconfiguration of the defect structures. Uzuhashi et al. showed that conventional annealing (∼1300 °C) leads to the formation of numerous Mg-enriched defects, while the formation of the Mg-enriched defects is greatly suppressed by ultra-high-pressure annealing (UHPA) at 1400 °C.26 The same study also suggested that while the formation of Mg-enriched defects is suppressed, there are still signs of Mg segregation at extended defects, such as dislocation loops, that remained after annealing at 1400 °C. Work by Kano et al. further examined these defects that form during post-implant annealing and characterized Burger's vectors of the dislocations that bound these defects.27 Uzuhashi and Kano used cathodoluminescence and observed an order of magnitude increase in the emission intensities due to donor–acceptor pairs (DAPs) between their samples annealed at 1400 °C vs 1300 °C. However, the change in DAP intensities only allows for relative comparisons and only qualitative improvements in Mg activation efficiency are extracted. Additionally, none of the earlier studies assessed the stability of the Mg-enriched defects that formed at 1300 °C after subsequent higher temperature annealing.

Here, we leverage the current understanding of Mg-implanted GaN by providing a fundamental understanding that directly relates defects and electrical performance. Based on the defect characterization method with x-ray scattering and electron microscopy techniques that we developed previously,28–30 this study aims to assess (i) the key residual defects after annealing that have a significant impact on the dopant activation efficiency and (ii) the kinetics of the defect formation and dissolution at different temperatures. In this work, the stability of the Mg-enriched defects formed at 1300 °C is investigated by subsequent high temperature (>1400 °C) annealing. These results are anticipated to guide the key processing steps and requirements to achieve high activation efficiency p-type doping for vertical GaN devices.

GaN epitaxy films with a thickness of 2.5 μm were grown via metal organic chemical vapor deposition (MOCVD) on high quality ammonothermal GaN (0001) substrates.16,30 The Mg ions with a box profile were implanted at room temperature using six different accelerating voltages (350, 200, 150, 100, 60, and 25 keV) at a total dose level of 1 × 1015 cm−2. Using the SRIM simulation software,31 the maximum concentration was calculated to be 1.8 × 1019 cm−3 from the surface to a depth of ∼500 nm. Post-implantation annealing was performed on the implanted wafers at temperatures of 1300, 1400, and 1500 °C for a duration of either 10 min or 100 min and so will be referred to as, for example, the 1300 °C; 10 min, sample. An ultra-high N2 pressure of one GPa was applied during the annealing sequence to prevent decomposition of GaN without using a cap.17,30

Structural characterization was performed by a combination of x-ray scattering and electron scattering techniques. The lattice distortions and crystalline quality were assessed by triple-axis x-ray diffraction (TAXRD) and x-ray topography (XRT). TAXRD measurements were performed using a Jordan Valley (Bruker) D1 diffractometer, with an incident beam mirror to produce a parallel beam, followed by a Si (220) channel cut collimator (Cu Kα1 radiation). The scattered beam optics included a Si (220) channel cut crystal.18,19,32 Synchrotron double crystal x-ray topography measurements were performed at the 1-BM Beamline of the Advanced Photon Source, Argonne National Laboratory with a photon energy of 8.05 keV. The first crystal was a highly asymmetric Si (333) beam expander and the sample was oriented for diffraction of the ( 11 2 ¯ 4 ) reflection in the glancing incidence geometry.29,30 Topography images were taken by exposing at different positions on the rocking curve and recorded separately on different films (hereinafter referred to as single exposure images). Information of the specific post-annealing defects was obtained using a transmission electron microscope (TEM). TEM samples roughly 100 nm thick were made using an FEI Nova 600 dual beam focused ion beam (FIB) and further thinned by a gentle Ar + ion beam with 0.3 kV incident energy to remove any FIB induced damage layer. A scanning transmission electron microscope (STEM) images using bright-field (BF) or high-angle annular dark-field (HAADF) detectors, energy-dispersive x-ray spectroscopy (EDX) (∼1 at. % detection limit), and high-resolution TEM images were taken using the FEI Talos and Argonne Chromatic and Aberration-Corrected (ACAT) TEM, respectively, at the Center for Nanoscale Materials, Argonne National Laboratory.33 

The triple-axis x-ray ω:2θ line scans near the GaN (0004) peak of all the samples are shown in Fig. 1(a). The presence of the intensity with fringes to the left of the main peak in the as-implanted sample is due to lattice distortion as a result of the implant-induced point defect formation.34–36 After annealing, in all cases, the implantation-induced strain is completely relieved. Previously, we showed that annealing under a similar condition at 1300 °C for 10 min removed the implantation-induced strain for a single implant (Mg: 100 keV) at a dose level of 2 × 1014 cm−2 (Ref. 17). In this work, the dose level is five times of that previous work, and the same annealing condition is still sufficient to relieve the strain. Meanwhile, annealing at higher temperatures and longer time does not have any further impact on the strain state. Thus, in all cases, the point defects, which were responsible for the strain, have reformed into more stable configurations that include extended defects. The residual defect structures after annealing were first screened using x-ray topography. Figures 1(b)1(e) show the single exposure curves x-ray topography images, taken at the peak of the (1124) rocking curves, from the as-implanted sample and samples annealed at 1300, 1400, and 1500 °C. In all cases, dot-shaped defects are observed, which correspond to pre-existing, individual threading dislocations. The dislocation density is on the order of 104 cm−2, which is consistent with the dislocation density level in ammonothermal GaN reported in the current literature.20 The as-implanted and the 1500 °C annealed samples show similar characteristics: only individual threading dislocations with no extended defects present. The 1400 °C sample consists of isolated loop-shaped defects that do not diffract at the same angle as the rest of the material. Separate images capture the loops diffracting at an angular difference of ∼8 arc sec away from the bulk of the GaN, quantifying the small but measurable local lattice distortion at the loop defects. The amount of the local lattice distortion here is comparable to an earlier work from our group, using laboratory XRT equipments, to determine the localized tilt (a few arc secs) around micropipe defects in SiC substrates.36,37 The 1300 °C sample, on the other hand, exhibits large non-diffracting features that are not observed in the other samples [arrow in Fig. 1(c)]. The feature size in XRT is associated with extended strain fields around the defects and points to a different type of defect formation than is present in the as-implanted wafers or in the wafers that were annealed at higher temperatures. To further assess the differences in defect formation under these different annealing conditions, TEM and STEM measurements were employed. Figure 2 shows the STEM two-beam condition bright-field images taken under different diffraction conditions. Figures 2(a)2(d) are taken under g = 0002 to show, for example, defects with a screw component and Figs. 2(e)2(h) are taken under g = 11 2 ¯ 0 to show defects with an edge component. The four samples can be divided into two groups based on the extended defects observed. The samples annealed at 1400 and 1500 °C showed very similar defect structures. The prominent residual defects after annealing at and above 1400 °C are dislocation loops. These dislocation loops are prismatic loops, dominated by edge characteristics and therefore show strong contrast only under under g = 0002 [Figs. 2(f)2(h)]. Similar prismatic loops have been observed in our earlier study focusing on Mg-implanted GaN (single implantation at 100 keV) after UPHA at 1300 °C.17 Most of the loops are not completely visible. Because those loops lie on crystallographic planes that are not perfectly parallel to the sample orientation, i.e., {1010}, only a portion of the loop is oriented to show contrast, an example is given with a blue arrow in Fig. 2(f). In the extreme case, when the loop is oriented perpendicular to the sample orientation, for example {1120}, it will show up as a bright line surrounded by sharp contrast on both sides, an example is given with a red arrow in Fig. 2(f). The contrast is caused by lattice distortion near the dislocation loop. On some of the dislocation loops, there are some patches with darker contrast (examples are given with yellow arrows). They were observed in all three samples annealed at and above 1400 °C. Details of these patch features will be discussed later.

FIG. 1.

(a) Triple-axis x-ray ω:2θ line scans near the GaN (0004) peak for the samples showing the implant-induced strain was fully relieved after annealing at 1300 °C for 10 min and annealing at temperatures ≥1400 °C has no further impact on the strain state; x-ray topography images exposed at a single point along a rocking curve for: (b) as-implanted; (c) annealed at 1300 °C for 10 min; (d) annealed at 1400 °C for 100 min, and (e) annealed at 1500 °C for 10 min. The white dots that appeared in all four images [(b)–(e)] are individual dislocations, and the densities are on the order of 104 cm−2.

FIG. 1.

(a) Triple-axis x-ray ω:2θ line scans near the GaN (0004) peak for the samples showing the implant-induced strain was fully relieved after annealing at 1300 °C for 10 min and annealing at temperatures ≥1400 °C has no further impact on the strain state; x-ray topography images exposed at a single point along a rocking curve for: (b) as-implanted; (c) annealed at 1300 °C for 10 min; (d) annealed at 1400 °C for 100 min, and (e) annealed at 1500 °C for 10 min. The white dots that appeared in all four images [(b)–(e)] are individual dislocations, and the densities are on the order of 104 cm−2.

Close modal
FIG. 2.

Cross section STEM two-beam condition bright-field images with diffraction vector g = 0002 , (a) 1300 °C annealed 10 min sample; (b) 1400 °C annealed 10 min sample; (c) 1400 °C annealed 100 min sample; (d) 1500 °C annealed 10 min sample; and diffraction vector g = 11 2 ¯ 0 ; (e) 1300 °C annealed 10 min sample; (f) 1400 °C annealed 10 min sample; (g) 1400 °C annealed 100 min sample; (h) 1500 °C annealed 10 min sample. In (f), the arrows show examples of dislocation loops that are not parallel to the FIB cut. All images share the same scale bar shown in (a).

FIG. 2.

Cross section STEM two-beam condition bright-field images with diffraction vector g = 0002 , (a) 1300 °C annealed 10 min sample; (b) 1400 °C annealed 10 min sample; (c) 1400 °C annealed 100 min sample; (d) 1500 °C annealed 10 min sample; and diffraction vector g = 11 2 ¯ 0 ; (e) 1300 °C annealed 10 min sample; (f) 1400 °C annealed 10 min sample; (g) 1400 °C annealed 100 min sample; (h) 1500 °C annealed 10 min sample. In (f), the arrows show examples of dislocation loops that are not parallel to the FIB cut. All images share the same scale bar shown in (a).

Close modal

The 1300 °C 10 min sample exhibits an additional, different defect structure than the others. Triangular and trapezoidal shape defects were observed under g = 0002 , these defects did not exist in the three samples annealed at higher temperatures. A magnified STEM bright-field image taken under g = 0002 for the 1300 °C 10 min sample is shown in Fig. 3(a). Examples of the triangular and trapezoidal shape defects are circled in red and highlighted in the blue box. The triangular defects showed structural characteristics similar to those of pyramidal inversion domains (PIDs). The base of the PIDs is along the {0001} plane and the sidewalls (highlighted with dotted lines) are along the {1123} planes, inclined at ∼47°, as shown by the high-resolution TEM image in Fig. 3(b). To verify that the fringes observed in the HRTEM correspond to the atomic positions, a focal series of images were taken such that the defocus was adjusted by 2 nm per step. In post-processing, the images were aligned and the series of images were compared to atomic models reported in the literature of PIDs that describe that extra plane of atoms along the basal plane.22,23,27 The lattice is distorted inside the PID when compared to the surrounding GaN matrix. An additional atomic layer (highlighted with a green box) was observed near the base of the pyramid (basal plane of GaN), causing lattice bending of the following ten layers. For an individual PID [the STEM BF image is shown in Fig. 3(c)], the energy-dispersive x-ray analysis (EDX) map in Fig. 3(d) shows the Mg signal at the PID. A line profile is generated by integrating the intensity in the yellow box in Fig. 3(d) and shown in Fig. 3(e). From the EDX map and the line profile, an increased Mg signal is observed at the base of the PID (∼9 at.  % or ∼8 × 1021 cm−3), where the extra layer of atoms is located, with a decrease in the Ga signal at the same position. This indicates that a compound including Mg and N forms in the PIDs.

FIG. 3.

(a) STEM image of the 1300 °C 10 min sample showing examples of pyramidal inversion domains (circled in red) and a trapezoidal inversion domain (boxed in blue); (b) HRTEM image of a pyramidal inversion domain showing the {1123} facets (highlighted with dotted white lines) and the extra layer of atoms near the base (boxed in green); (c) BF image of a pyramidal inversion domain; (d) EDX map showing the Mg segregation in the pyramidal inversion domain shown in (c). (e) EDX line profile generated by integrating the intensity in the yellow box in (d) (the arrow shows the direction of the line profile, i.e., the start is at the bottom).

FIG. 3.

(a) STEM image of the 1300 °C 10 min sample showing examples of pyramidal inversion domains (circled in red) and a trapezoidal inversion domain (boxed in blue); (b) HRTEM image of a pyramidal inversion domain showing the {1123} facets (highlighted with dotted white lines) and the extra layer of atoms near the base (boxed in green); (c) BF image of a pyramidal inversion domain; (d) EDX map showing the Mg segregation in the pyramidal inversion domain shown in (c). (e) EDX line profile generated by integrating the intensity in the yellow box in (d) (the arrow shows the direction of the line profile, i.e., the start is at the bottom).

Close modal

For the trapezoidal shape defects, we also observed facets along the {1123} planes on the edge, shown in Fig. 4(a) (highlighted with orange lines). The EDX map and line profile from the trapezoidal shape defects also show increased Mg signal and reduced Ga signal along the basal plane, as shown in Figs. 4(b) and 4(c), which is similar to what was observed in the PIDs. Therefore, we consider these trapezoidal shape defects as trapezoidal inversion domains (TIDs). It has been previously speculated that a singular defect type was the cause for the triangular defects (red circles) and elongated defects (blue box) observed in Fig. 3(a). The proposed defect character was a triangular prism shape that appeared as a triangle when parallel to the zone axis of the micrograph or as a line when perpendicular to the zone axis.27 However, this is not what is observed in this study. Instead, we speculate that the defects are pyramidal (hence the name PID) and are the initial stages of development and will eventually evolve into the TIDs. In fact, the center part of the largest TID in Fig. 3(a) (boxed in blue) consists of a series of small, aligned pyramids that form a sawtooth pattern, which is the basis of our claim. Plan view TEM of these defects observed in epitaxially grown GaN also supports the claim that these triangular defects have a pyramidal base and are not triangular prisms.38 Thus, it appears that both the PID and TID defects contain electrically inactive Mg atoms and could be a cause of the reduction in Mg activation efficiency. In the current literature, PIDs with signs of Mg segregation have been observed in both highly Mg-doped epitaxial layers with Mg concentration above the 1019 cm−3 level21–23 as well as in Mg-implanted GaN after annealing at 1350 °C.26,27 In MOVPE epitaxial Mg-doped p-GaN, Vennéguès et al. hypothesized that introducing a higher level of Mg in GaN results in the formation of Mg3N2 where inactive Mg atoms are accommodated into other sites or phases.22 However, our results do not support the formation of Mg3N2, which is characterized by two monolayers of Mg at the inversion domain boundary.39,40

FIG. 4.

(a) BF image of a trapezoidal inversion domain with the edge also showing the {1123} facets (highlighted with a diagonal orange line); (b) EDX map showing the Mg segregation in the trapezoidal inversion domain shown in (a). (c) EDX line profile generated by integrating the intensity in the yellow box in (b) (the arrow shows the direction of the line profile, i.e., the start is at the bottom).

FIG. 4.

(a) BF image of a trapezoidal inversion domain with the edge also showing the {1123} facets (highlighted with a diagonal orange line); (b) EDX map showing the Mg segregation in the trapezoidal inversion domain shown in (a). (c) EDX line profile generated by integrating the intensity in the yellow box in (b) (the arrow shows the direction of the line profile, i.e., the start is at the bottom).

Close modal

Another study from Remmele et al. further characterized the PIDs produced during epitaxy with TEM and modeled the domain boundary along the base to follow GaNMgNGa layers in an abcab stacking,39 which is further supported by first principles calculations.41 Our HRTEM results are in good agreement with this model and show that a similar interaction between Mg and N occurred in Mg-implanted GaN after annealing at 1300 °C. While Mg3N2 is the only reported compound in the Mg–N system,22 we suggest that Mg substitutes Ga sites on the Inversion Domain Boundary (IDB) but does not form Mg3N2 compound based on experimental observations. The base of PIDs contains a row of Mg atoms and thus prevents the Mg from becoming electrically active. However, in those epitaxially grown structures, it is difficult to remove the PIDs once they are formed. To suppress the formation of the PIDs in epitaxy, a low-temperature growth method is needed, such as metal-modulated epitaxy.42 Unlike the case for annealing the ion implantation damage, the MOCVD growth temperature (∼1150 °C) of highly Mg-doped epitaxial GaN is well below the temperature needed to deter inversion domain formation. In the case of Mg-implanted GaN, the current literature has not shown how to remove PIDs once they are formed. The PIDs and TIDs are not observed in any of the samples annealed at and above 1400 °C in this work. This suggests that the PIDs and TIDs either do not form or are dissolved at high temperatures (≥1400 °C) and are formed only at temperatures <1300 °C after implantation as presented here and also described in works by Kumar et al.,23,25 for implantation (and in works by Vennéguès et al.,21,22 for epitaxial structures in which high concentrations of Mg are present after epitaxial growth at ∼1150 °C). This provides insight for improving dopant activation efficiency by annealing Mg-implanted GaN at temperatures above 1400 °C. In our previous study of single implanted GaN, PIDs were not observed when annealed at 1300 °C for 10 min.17 While the peak concentration of Mg from the single implant is 1.6 × 1019 cm−3 in the as-implanted case, after annealing, diffusion causes the peak concentration to be reduced. Simulations of the Mg concentration profile after annealing at 1300 °C for 10 min (diffusivity of 2.9 × 10−12 cm2/s) reveal a reduction to ∼2 × 1018 cm−3. This agrees well with studies of PIDs reported in the literature for both epitaxial grown and implanted GaN where the Mg concentration threshold to form PIDs is above 1 × 1019 cm−3.14,21–23 Recently, Breckenridge et al.,30 reported the electrical data from a sister set of the samples annealed up to 1400 °C under similar conditions as those reported in this study. Their results showed that annealing at 1400 °C leads to higher dopant activation (∼10% at room temperature, ∼100% at 500 °C, because the Mg level is not that shallow – at ∼Ev + 0.2 eV), compared to annealing at 1300 °C (<1% at room temperature, ∼40% at 500 °C). Based on the model from Remmele et al., we estimate the amount of Mg bound in the PIDs and TIDs is 1.6 × 1018 cm−3, which is ∼10% of the total Mg content.39 This represents the lower bound of the Mg in PIDs and TIDs because the EDX clearly shows higher Mg concentrations within the entire defects [for example, Figs. 3(d) and 4(b)]. Our study complements the earlier study from Breckenridge et al.,30 by providing the structural analysis of the residual defects and showing the Mg segregation in PIDs and TIDs is a key mechanism for the reduction of dopant activation efficiency and can be avoided if the annealing temperature is ≥1400 °C.

As shown before, the prominent residual defects after annealing at and above 1400 °C are prismatic dislocation loops. A high-resolution TEM image from one of the small loops in the 1400 °C 10 min sample is shown in Fig. 5(a). Fast Fourier transform (FFT) patterns were obtained from three regions on the HRTEM image, outside of the loop (blue), along the perimeter of the loop (orange), and the bottom edge of the loop (red), as shown in Figs. 5(b)5(d). FFT taken at regions outside of the loop (blue) and on the bottom edge of the loop (red) show a pure hexagonal GaN pattern, highlighted with red circles. FFT near the perimeter of the loop shows additional patterns, highlighted with yellow circles that correspond to the atomic rearrangement along the perimeter of the dislocation loop. Iwata et al., have proposed that the atomic rearrangement of the dislocation loops is a 1 611–20 translation from both sides of a vacancy disk that collapses. Similar dislocation loops have been observed in irradiated hexagonal zirconium, which reveal the same electron diffraction pattern we observe here.43 There is an overall decreasing trend in the defect density after annealing at a higher temperature or for a longer time. The loop defect density was measured to be 3.8 × 109 cm−2 for 1400 °C 10 min case, 1.5 × 109 cm−2 for the 1400 °C 100 min case, and 1.1 × 109 cm−2 for 1500 °C 10 min case, which is consistent with loop defect density trends observed when annealing at 2 and 0.3 GPa.44 Annealing at high temperatures (≥1400 °C) showed significant reduction in the segregated Mg concentration (below the EDX detection limit) and a decreasing trend in defect density, both of which are expected to improve device performance.

FIG. 5.

(a) HRTEM image of a small dislocation loop, fast Fourier transform pattern from three regions (three boxes) are shown in (b)–(d); (b) FFT outside of the loop shows a pure hexagonal patternl; (c) FFT along the perimeter of the loop shows a hexagonal GaN plus additional diffraction spots associated with the atomic rearrangement along dislocation loops; (d) FFT on the bottom edge of the loop also shows a pure hexagonal pattern; (e) BF STEM image of a small dislocation loop and the (f) the EDX map showing no signs of Mg segregation in the loop.

FIG. 5.

(a) HRTEM image of a small dislocation loop, fast Fourier transform pattern from three regions (three boxes) are shown in (b)–(d); (b) FFT outside of the loop shows a pure hexagonal patternl; (c) FFT along the perimeter of the loop shows a hexagonal GaN plus additional diffraction spots associated with the atomic rearrangement along dislocation loops; (d) FFT on the bottom edge of the loop also shows a pure hexagonal pattern; (e) BF STEM image of a small dislocation loop and the (f) the EDX map showing no signs of Mg segregation in the loop.

Close modal

As mentioned earlier, there are some patches on dislocation loops in all the samples annealed at and above 1400 °C. Three examples are shown in Fig. 6. The size of the patches is ∼15 nm. In a recent study, Uzuhashi et al. observed a similar feature in Mg-implanted GaN with similar annealing conditions (UHPA at T = 1400 °C) but with a shorter annealing time and slower ramp rates (5 min and 0.45 °C/s) than the samples shown in this study (10 min and 0.66 °C/s). Their results show Mg accumulation at a single patch along a dislocation loop. However, the patches in our study show quite different characteristics. None of the patches in our samples show evidence of Mg segregation (detection limit ∼1 at. %), based on the EDX maps of ten patches throughout the TEM sample. Instead, there is a noticeable decrease in the N content at the patches, suggesting these features are associated with N vacancy related defects. This is clearly different than the case for the formation of PIDs in which the presence of Mg was accompanied by a reduction in the gallium concentration and not a reduction in nitrogen. This is supported by cathodoluminescence measurements done by Kano et al., that show the emission of green luminescence (GL) (correlated to the presence of N vacancies) that is observed at the higher annealing temperatures even after a significant reduction in the GL intensity from samples annealed at 1300 °C. We suggest that the Mg segregation seen in the patches formed above 1400 °C in Uzuhashi et al., and not observed in this study at the same annealing temperature, is a consequence of shorter time annealed at 1400 °C (5 min compared to 10 min in this study) in addition to ∼37 min spent between 1250 and 1350 °C as a consequence of the 0.45 °C/s ramp rate compared to ∼21 min in this experiment. Presuming these PIDs form around 1300 °C and are dissolved around 1400 °C, the experiment done by Uzuhashi experienced more time at the temperatures PIDs are formed and shorter time at 1400 °C, where we observe they dissolve. We acknowledge that the Mg content at the loops (>1400 °C) is likely below the EDX detection limit when comparing our results to the APT done in Uzuhashi and Kano's works. However, the amount of Mg observed at the loops (below the detection limit) must be very low because ∼100% activation efficiency is observed at 500 °C published by our collaborators at Breckenridge et al. In terms of the relative amounts of Ga, N, and Mg observed; where Mg is observed, we see a drop in the Ga content, not in the N content.

FIG. 6.

BF STEM and EDX of three representative patches along dislocation loops are shown. No Mg segregation was observed in any of the 10 total patches that were measured. However, a noticeable decrease in N content suggests that defects associated with N vacancies are present along these patches, which are further investigated with HRTEM.

FIG. 6.

BF STEM and EDX of three representative patches along dislocation loops are shown. No Mg segregation was observed in any of the 10 total patches that were measured. However, a noticeable decrease in N content suggests that defects associated with N vacancies are present along these patches, which are further investigated with HRTEM.

Close modal

Structural details of the patch feature are further investigated using HRTEM, as shown in Fig. 7. Figure 7(a) shows a lower magnification image capturing two patches on a dislocation loop. Figure 7(b) shows the HRTEM image of the bottom patch in Fig. 7(a). Within the patch, there are two interesting observations. First, is the presence of stacking faults boxed in red and evident in the FFT in Fig. 7(c). Next along the perimeter of the patch, there are additional diffraction spots that exhibit the same pattern that was observed along the perimeter of the dislocation loop seen in Fig. 5(c). This corresponds well with the speculation from Iwata et al. that the small patches along these loops are nanoscale versions of the dislocation loops themselves.

FIG. 7.

(a) Low magnification TEM image showing two patches on a dislocation loop. (b) HRTEM of the lower right patch from (a). (c) The FFT of the red square in (b) reveals stacking faults within these patches. (d) The FFT of the blue square in (b) shows hexagonal GaN with extra diffraction spots that correlate to atomic rearrangement due to the dislocation loop that was also observed in Fig. 5.

FIG. 7.

(a) Low magnification TEM image showing two patches on a dislocation loop. (b) HRTEM of the lower right patch from (a). (c) The FFT of the red square in (b) reveals stacking faults within these patches. (d) The FFT of the blue square in (b) shows hexagonal GaN with extra diffraction spots that correlate to atomic rearrangement due to the dislocation loop that was also observed in Fig. 5.

Close modal

We addressed the stability of the PIDs through a combination of annealing steps. First, we assessed the role of lower temperature annealing steps on the formation of the PID defects. As noted in our earlier study,16 annealing at 1000 °C for 10 min relieves the implant-induced strain. However, the impact of such a lower temperature anneal (or equivalently, a slow ramp rate) on the formation of the Mg-containing PIDs had not been previously addressed. Figure 8 shows the TEM images of a sample annealed at 1000 °C for 30 min followed in the chamber by annealing at 1300 °C for 10 min. The PID and related defects are present after the two-step anneal with a density comparable to a 1300 °C step alone (Fig. 2). This indicates that the atomic rearrangements after the lower temperatures do not have a significant impact on the PID formation, which, in turn, suggests that the ramp rate (∼ 0.5 °C/s here) does not influence the defect formation at 1300 °C, although the formation of such defects after rapid thermal annealing7 has not been widely studied. On the other hand, Fig. 9 shows the TEM images from a sample that had been annealed at 1300 °C for 10 min and brought back to room temperature [same sample as Figs. 2(a) and 2(e)] and then re-annealed at 1450 °C for 10 min. The PIDs are absent after the higher temperature annealing step, confirming that defects that may form during an anneal at 1300 °C can be dissolved by subsequent high temperature annealing, supporting our earlier assessment that PID defects formed at lower temperatures can be removed during higher temperature annealing, and can also explain why others26 observed an Mg-containing defect after a shorter (5 min) anneal at 1400 °C as well as a slower ramp rate to 1400 °C (to include more time in the temperature range of 1300 °C to form PIDs and related defects during the ramp-up).

FIG. 8.

Cross-sectional bright-field STEM images for the sample annealed at 1000 °C for 30 min followed by annealing at 1300 °C for 10 min with diffraction vector (a) g = 0002 and (b) g = 11 2 ¯ 0 . PID formation and density are comparable to the sample only annealed at 1300 °C as in Fig. 2(a).

FIG. 8.

Cross-sectional bright-field STEM images for the sample annealed at 1000 °C for 30 min followed by annealing at 1300 °C for 10 min with diffraction vector (a) g = 0002 and (b) g = 11 2 ¯ 0 . PID formation and density are comparable to the sample only annealed at 1300 °C as in Fig. 2(a).

Close modal
FIG. 9.

Cross-sectional bright-field STEM images for the sample initially annealed at 1300 °C for 10 min followed by annealing at 1450 °C for 10 min with diffraction vector (a) g = 0002 and (b) g = 11 2 ¯ 0 . The PIDs that form during an anneal at 1300 °C [Figs. 2(a) and 2(e)] can be dissolved by subsequent high temperature annealing.

FIG. 9.

Cross-sectional bright-field STEM images for the sample initially annealed at 1300 °C for 10 min followed by annealing at 1450 °C for 10 min with diffraction vector (a) g = 0002 and (b) g = 11 2 ¯ 0 . The PIDs that form during an anneal at 1300 °C [Figs. 2(a) and 2(e)] can be dissolved by subsequent high temperature annealing.

Close modal

The defect characteristics in Mg-implanted homoepitaxial GaN with ultra-high N2 pressure annealing were investigated. Annealing of Mg-implanted GaN with one GPa N2 pressure at 1300 °C for 10 min completely removed the implant-induced strain. Extended defects remain after annealing with both pyramidal and trapezoidal inversion domains, as well as prismatic dislocation loops. The inversion domains show clear Mg segregation and produce electrically inactive Mg that limits dopant activation efficiency. Increasing the annealing temperature ≥1400 °C under the same condition results in a decrease in residual loop defect density with no sign of Mg segregation at these defects. Pyramidal and trapezoidal inversion domains also do not exist with annealing temperatures ≥1400 °C. This study provides the structural analysis of the residual defects, showing Mg segregation in PIDs and TIDs is one of the mechanisms for inactive Mg and can be avoided if the post-implant annealing temperature is ≥1400 °C, even if there are lower temperature annealing steps prior to the ≥1400 °C step. This work complements an earlier study with dopant activation efficiency data showing that annealing at 1400 °C leads to higher dopant activation compared to annealing at 1300 °C.30 Results from this work are expected to help achieve higher activation efficiency of p-type doping for devices including vertical GaN device structures.

This research was supported through the ARPA-E PNDIODES program under Contract No. DE-AR0001116 at UCLA. This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. The synchrotron x-ray topography measurements were carried out at 1-BM beamline of the Advanced Photon Source, Argonne National Laboratory. TEM sample preparation and measurements were performed at the Center for Nanoscale Materials, Argonne National Laboratory. A part of this research was supported by the Polish National Science Centre through Project No. 2018/29/B/ST5/00338 as well as by the Polish National Centre for Research and Development through Project No. TECHMATSTRATEG-III/0003/2019-00.

The authors have no conflicts to disclose.

K. Huynh: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Y. Wang: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – review & editing (equal). M. E. Liao: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (equal); Writing – review & editing (equal). J. Tweedie: Investigation (equal); Methodology (equal); Resources (equal). P. Reddy: Investigation (equal); Methodology (equal); Resources (equal). M. H. Breckenridge: Investigation (equal); Methodology (equal); Resources (equal). R. Collazo: Investigation (equal); Methodology (equal); Resources (equal). Z. Sitar: Investigation (equal); Methodology (equal); Resources (equal). K. Sierakowski: Investigation (equal); Methodology (equal); Resources (equal). M. Bockowski: Investigation (equal); Methodology (equal); Resources (equal). X. Huang: Investigation (equal); Methodology (equal); Resources (equal). M. Wojcik: Investigation (equal); Methodology (equal); Resources (equal). M. S. Goorsky: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Supervision (equal); Visualization (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.

1.
I. C.
Kizilyalli
,
A. P.
Edwards
,
O.
Aktas
,
T.
Prunty
, and
D.
Bour
,
IEEE Trans. Electron Devices.
62
,
414
422
(
2015
).
2.
Y.
Saitoh
,
K.
Sumiyoshi
,
M.
Okada
,
T.
Horii
,
T.
Miyazaki
,
H.
Shiomi
,
M.
Ueno
,
K.
Katayama
,
M.
Kiyama
, and
T.
Nakamura
,
Appl. Phys. Express
3
,
081001
(
2010
).
3.
L.
Rubin
and
J.
Poate
, “Ion implantation in silicon technology,”
Ind. Phys.
9
(3),
12
15
(
2003
).
4.
V.
Heera
,
D.
Panknin
, and
W.
Skorupa
,
Appl. Surf. Sci.
184
,
307
(
2001
).
5.
T.
Narita
,
H.
Yoshida
,
K.
Tomita
,
K.
Kataoka
,
H.
Sakurai
,
M.
Horita
,
M.
Bockowski
,
N.
Ikarashi
,
J.
Suda
,
T.
Kachi
, and
Y.
Tokuda
,
J. Appl. Phys.
128
(
9
), 090901 (
2020
).
6.
H.
Fu
,
K.
Fu
,
S.
Chowdhury
,
T.
Palacios
, and
Y.
Zhao
,
IEEE Trans. Electron Devices
68
(
7
),
3200
3211
(
2021
).
7.
S. J.
Pearton
,
J. C.
Zolper
,
R. J.
Shul
, and
F.
Ren
,
J. Appl. Phys.
86
(
1
),
1
78
(
1999
).
8.
T. J.
Anderson
,
B. N.
Feigelson
,
F. J.
Kub
,
M. J.
Tadjer
,
K. D.
Hobart
,
M. A.
Mastro
,
J. K.
Hite
, and
C. R.
Eddy
, Jr.
,
Electron. Lett.
50
(
3
),
197
198
(
2014
).
9.
B. N.
Feigelson
,
T. J.
Anderson
,
M.
Abraham
,
J. A.
Freitas
, Jr.
,
J. K.
Hite
,
C. R.
Eddy
, Jr.
, and
F. J.
Kub
,
J. Cryst. Growth
350
(
1
),
21
26
(
2012
).
10.
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
(
9
),
P382
(
2015
).
11.
M.
Kuball
,
J. M.
Hayes
,
T.
Suski
,
J.
Jun
,
M.
Leszczynski
,
J.
Domagala
,
H. H.
Tan
,
J. S.
Williams
, and
C.
Jagadish
,
J. Appl. Phys.
87
,
2736
2741
(
2000
).
12.
K.
Sierakowski
,
R.
Jakiela
,
B.
Lucznik
,
P.
Kwiatkowski
,
M.
Iwinska
,
M.
Turek
,
H.
Sakurai
,
T.
Kachi
, and
M.
Bockowski
,
Electronics
9
,
1380
(
2020
).
13.
S.
Porowski
,
I.
Grzegory
,
D.
Kolesnikov
,
W.
Lojkowski
,
V.
Jager
,
W.
Jager
,
V.
Bogdanov
,
T.
Suski
, and
S.
Krukowski
,
J. Phys. Condens. Matter
14
,
11097
(
2002
).
14.
A.
Kumar
,
W.
Yi
,
J.
Uzuhashi
,
T.
Ohkubo
,
J.
Chen
,
T.
Sekiguchi
,
R.
Tanaka
,
S.
Takashima
,
M.
Edo
, and
K.
Hono
,
J. Appl. Phys.
128
,
065701
(
2020
).
15.
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
).
16.
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
).
17.
Y.
Wang
,
K.
Huynh
,
M. E.
Liao
,
H.-M.
Yu
,
T.
Bai
,
J.
Tweedie
,
M. H.
Breckenridge
,
R.
Collazo
,
Z.
Sitar
,
M.
Bockowski
,
Y.
Liu
, and
M. S.
Goorsky
,
Phys. Status Solidi B
257
,
1900705
(
2020
).
18.
Y.
Wang
,
T.
Bai
,
C.
Li
,
M.
Tadjer
,
T.
Anderson
,
J.
Hite
,
M.
Mastro
,
C.
Eddy
,
K.
Hobart
,
B.
Feigelson
, and
M.
Goorsky
,
ECS J. Solid State Sci. Technol.
8
(
2
),
P70
P76
(
2019
).
19.
R.
Kucharski
,
T.
Sochacki
,
B.
Lucznik
, and
M.
Bockowski
,
J. Appl. Phys.
128
,
050902
(
2020
).
20.
S.
Suihkonen
,
S.
Pimputkar
,
S.
Sintonen
, and
F.
Tuomisto
,
Adv. Electron. Mater.
3
,
1600496
(
2017
).
21.
P.
Vennéguès
,
M.
Benaissa
,
B.
Beaumont
,
E.
Feltin
,
P.
De Mierry
,
S.
Dalmasso
,
M.
Leroux
, and
P.
Gibart
,
Appl. Phys. Lett.
77
,
880
(
2000
).
22.
P.
Vennéguès
,
M.
Leroux
,
S.
Dalmasso
,
M.
Benaissa
,
P.
De Mierry
,
P.
Lorenzini
,
B.
Damilano
,
B.
Beaumont
,
J.
Massies
, and
P.
Gibart
,
Phys. Rev. B
68
,
235214
(
2003
).
23.
K.
Iwata
,
T.
Narita
,
M.
Nagao
,
K.
Tomita
,
K.
Kataoka
,
T.
Kachi
, and
N.
Ikarashi
,
Appl. Phys. Express
12
,
031004
(
2019
).
24.
A.
Kumar
,
J.
Uzuhashi
,
T.
Ohkubo
,
R.
Tanaka
,
S.
Takashima
,
M.
Edo
, and
K.
Hono
,
J. Appl. Phys.
126
,
235704
(
2019
).
25.
A.
Kumar
,
K.
Mitsuishi
,
T.
Hara
,
K.
Kimoto
,
Y.
Irokawa
,
T.
Nabatame
,
S.
Takashima
,
K.
Ueno
,
M.
Edo
, and
Y.
Koide
,
Nanosc. Res. Lett.
13
,
403
(
2018
).
26.
J.
Uzuhashi
,
J.
Chen
,
A.
Kumar
,
W.
Yi
,
T.
Ohkubo
,
R.
Tanaka
,
S.
Takashima
,
M.
Edo
,
K.
Sierakowski
,
M.
Bockowski
,
H.
Sakurai
,
T.
Kachi
,
T.
Sekiguchi
, and
K.
Hono
,
J. Appl. Phys.
131
,
185701
(
2022
).
27.
E.
Kano
,
K.
Kataoka
,
J.
Uzuhashi
,
K.
Chokawa
,
H.
Sakurai
,
A.
Uedono
,
T.
Narita
,
K.
Sierakowski
,
M.
Bockowski
,
R.
Otsuki
, and
K.
Kobayashi
,
J. Appl. Phys.
132
,
065703
(
2022
).
28.
Y.
Wang
,
M. E.
Liao
,
K.
Huynh
,
A. A.
Allerman
, and
M. S.
Goorsky
,
ECS J. Solid State Sci. Technol.
10
(
4
),
045010
(
2021
).
29.
M. E.
Liao
,
K.
Huynh
,
Y.
Wang
,
A. A.
Allerman
, and
M. S.
Goorsky
, “The structural evolution of dot-core GaN substrates with annealing under growth-like conditions,”
ECS Trans.
98
(6),
15
(
2020
).
30.
M. H.
Breckenridge
,
J.
Tweedie
,
P.
Reddy
,
Y.
Guan
,
P.
Bagheri
,
D.
Szymanski
,
S.
Mita
,
K.
Sierakowski
,
M.
Bockowski
,
R.
Collazo
, and
Z.
Sitar
,
Appl. Phys. Lett.
118
,
022101
(
2021
).
31.
F.
Ziegler
,
J. P.
Biersacke
, and
U.
Littmark
,
The Stopping and Range of Ions in Solids
(
Oxford
,
Pergamon
,
1985
), Vol. 1.
32.
M. E.
Liao
,
Y.
Wang
,
T.
Bai
, and
M. S.
Goorsky
,
ECS J. Solid State Sci. Technol.
8
,
P673
(
2019
).
33.
J.
Wen
,
D. J.
Miller
,
W.
Chen
,
T.
Xu
,
L.
Yu
,
S. B.
Darling
, and
N. J.
Zaluzec
,
Microsc. Microanal.
20
,
1507
(
2014
).
34.
S.
Hayashi
,
D.
Bruno
, and
M. S.
Goorsky
,
Appl. Phys. Lett.
85
,
236
(
2004
).
35.
S.
Hayashi
,
M. S.
Goorsky
,
A.
Noori
, and
D.
Bruno
,
J. Electrochem. Soc.
153
(
12
),
G1011
G1014
(
2006
).
36.
M. E.
Liao
,
Y.
Wang
,
T.
Bai
, and
M. S.
Goorsky
, “xfoliation of β-Ga2O3 along a non-cleavage plane using helium ion implantation,”
ECS J Solid State Sci. Technol.
8
(11),
P673
(
2019
).
37.
B. D.
Poust
,
T. S.
Koga
,
R.
Sandhu
,
B.
Heying
,
R.
Hsing
,
M.
Wojtowicz
,
A.
Khan
, and
M. S.
Goorsky
,
J. Phys. D
36
,
A102
(
2003
).
38.
Z.
Liliental-Weber
,
T.
Tomaszewicz
,
D.
Zakharov
, and
M. A.
O’Keefe
,
J. Cryst. Growth
281
(
1
),
125
134
(
2005
).
39.
T.
Remmele
,
M.
Albrecht
,
K.
Irmscher
,
R.
Fornari
, and
M.
Straßburg
,
Appl. Phys. Lett.
99
,
141913
(
2011
).
40.
M.
Leroux
,
P.
Vennéguès
,
S.
Dalmasso
,
M.
Benaissa
,
E.
Feltin
,
P.
De Mierry
,
B.
Beaumont
,
B.
Damilano
,
N.
Grandjean
, and
P.
Gibartm
,
Phys. Status Solidi A
192
(
2
),
394
400
(
2002
).
41.
J. E.
Northrup
,
Appl. Phys. Lett.
82
(
14
),
2278
2280
(
2003
).
42.
B. P.
Gunning
,
C. A. M.
Fabien
,
J. J.
Merola
,
E. A.
Clinton
,
W. A.
Doolittle
,
S.
Wang
,
A. M.
Fischer
, and
F. A.
Ponce
,
J. Appl. Phys.
117
,
045710
(
2015
).
43.
S. M.
Liu
and
W. Z.
Han
,
Tungsten
3
(
4
),
470
481
(
2021
).
44.
E.
Kano
,
R.
Otsuki
,
K.
Kobayashi
,
K.
Kataoka
,
K.
Sierakowski
,
M.
Bockowski
,
M.
Nagao
,
T.
Narita
,
T.
Kachi
, and
N.
Ikarashi
,
Phys. Status Solidi Rapid Res. Lett.
17
,
2300035
(
2023
).