Introduction

A critical issue in the fabrication of gallium nitride (GaN) devices is the achievement of significant and controllable p-type doping. It still remains a challenge because of the high n-type autodoping background present in as-grown materials and the large ionization energy of acceptors, such as Mg, Zn and Cd.Reference Strite and Morkoc ¹ The principal p-type dopant used for GaN is Mg with an ionization energy of 150-165meV.Reference Akasaki, Amano, Kito and Hiramatsu ² Such a large acceptor ionization energy is problematic and two to three orders of magnitude higher atomic doping level of Mg must be incorporated into GaN in order to achieve the desired hole concentration at room temperature.Reference Orton ³ Although Zn and Cd are conventionally used as p-type dopants in the growth of other III-V compounds, StriteReference Strite ⁴ suggested that the d-electron core relaxation in these elements is partially responsible for the enhanced depth, making efficient doping at room temperature impossible.

Beryllium (Be) was thought to be a shallower acceptor in GaN due to its large electronegativity and the absence of d-electrons. Ab initio calculationsReference Bernardini, Fiorentini and Bosin ⁵ predicted that Be behaves as a rather shallow acceptor in GaN, with a thermal ionization energy of 60 meV in wurtzite GaN. More evidence from photoluminescence (PL) spectra revealed that Be acts as an acceptor with an optical ionization energy ranging from 90-100 meV,Reference Dewsnip, Andrianov, Harrison, Orton, Lacklison, Hopper, Ren, Cheng and Foxon ^{6 ,} Reference Sanchez, Calle, Sanchez-Garcia, Calleja, Munoz, Molly, Somerford, Koschnick, Michael and Spaeth ⁷ 150 meV,Reference Ronning, Carlson, Thomson and Davis ⁸ to 250 meV.Reference Salvador, Kim, Aktas, Botchkarev, Fan and Morkoc ⁹ However, the size of Be atoms is so small that it seems more probable for them to stay at interstitial sites (Be_int) rather than at substitutional sites (Be_Ga) in GaN. Theoretical calculationsReference Neugebauer and Van De Walle ¹⁰ also pointed out that the formation energy of Be_int is much less than that of Be_Ga. Interstitial Be behaves like a double donor so that self-compensation is a significant drawback for the use of Be as an acceptor. That is why there is almost no achievement of electrical activiation of Be-doped GaN, except for Brandt et al. who obtained high mobility p-type materials from Be-O codoped cubic GaN by molecular beam epitaxy.Reference Brandt, Yang, Kostial and Ploog ¹¹

In this paper, we introduce a new annealing process in which the Be-implanted GaN wafers were first annealed in forming gas, followed by annealing in a pure N₂ atmosphere. Electrical activation of Be-implanted GaN was observed for the first time. The results confirmed the low activation energy of Be as an acceptor in GaN and also showed the possibility of p-type doping by Be implantation.

Experiment

The undoped GaN layers used in the experiments were 2µm thick and grown on c-plane sapphire substrates by MOCVD in a multiwafer rotating disk reactor at 1040 °C, with a ∼20 nm GaN buffer layer grown at 530 °C in advance. The background n-type carrier concentrations were around 9×10¹⁶ cm⁻³. The as-grown layers had featureless surfaces and were transparent with a strong near band-edge luminescence at 3.44 eV at room temperature. Be was implantedFootnote ⁺ into two pieces of the undoped GaN wafer at 40 keV. The doses were 3×10¹⁴ cm⁻² and 1×10¹⁵ cm⁻², respectively. Samples of dimensions 6×6 mmReference Akasaki, Amano, Kito and Hiramatsu ² were cut from the wafers. Annealing was performed in a RTP system equipped with halogen-tungsten lamps. One set of the implanted samples was sequentially annealed at 900 -1100 °C for 45 s in flowing N₂ according to a face-to-face geometry. The other set of implanted samples was annealed in forming gas (12% H₂ and 88% N₂) first at temperatures ranging from 500 to 1100 °C and then in flowing N2. Hall effect measurements were conducted at room and variable temperatures with a magnetic field of 0.32 Tesla. Indium dots were alloyed at the corners of each sample according to the Van der Pauw geometry. Secondary ion mass spectrometry (SIMS) analyses were carried out in a CAMECA ims-6f microscope. An O₂ Footnote ⁺ beam of 200 nA and 8 keV impact energy was used to sputter the samples. The quantification of the H concentration was not available due to the lack of reference. However, the HFootnote ⁺ intensities in all the samples have been normalized with respect to NFootnote ⁺ , and therefore were comparable.

Results and Discussion

From the variable temperature Hall measurements, a plot of the sheet carrier concentration/temperature product (psT^−3/2 ) vs.reciprocal temperature can be constructed. Figure 1 shows such a plot for a Be-implanted GaN sample annealed at 600 °C and 10 s in forming gas first, followed at 1100 °C and 45 s in N₂. As shown by Gotz et al., Reference Gotz, Johnson, Chen, Liu, Kuo and Imler ¹² the normal slope analysis in a linear region of the carrier concentration alone versus reciprocal temperature data could yield around 1/3 more than the real activation energy. We think the Arrhenius plot of the sheet carrier concentration/temperature product vs. the inverse temperature is more appropriate to estimate the ionization level of Be. Under nondegenerate conditions, the hole concentration can be expressed as the following relation: p ∝ p_oT^3/2exp(−E_a/kT), where p_o is the acceptor concentration and E_a is the activation energy of the acceptors. So, ln(pT^−3/2) should be proportional to the inverse T. Due to non-uniformity of the implanted atoms in the substrate layer, here we use the sheet carrier concentration p_s to plot, instead of p. From Figure 1, the ionization energy of Be in GaN is calculated to be 127 meV, which is lower than that of Mg and Ca reported for implanted GaN.^2, Reference Zolper, Wilson, Pearton and Stall ¹³ Based on the ionization energy estimated above, only 0.73% of Be acceptors would be ionized at room temperature. The activation efficiency can thus be estimated to be around 100% for this sample (considering p_s = 2.14×10¹³ cm ⁻² at room temperature), if the activation efficiency is defined as sheet carrier concentration /(dose×ionization rate). The activation efficiency of Be is similar to that of Ca.Reference Zolper, Wilson, Pearton and Stall ¹³

FIG.1 Arrhenius plot of the sheet carrier concentration/temperature product of Be-implanted GaN annealed in forming gas first, followed in N₂, the dose is 3×10Reference Eckey, von Gfug, Holst, Hoffmann, Kaschner, siegle, Thomsen, Schineller, Heime, Heuken, Schon and Beccard ¹⁴ cm⁻²

The results of room temperature Hall measurement are summarized in Table I. Several remarkable things should be pointed out. (i) The implanted samples without post-annealing process still showed n-type conductive characteristics with a bit of reduction in electron concentrations, unlike most implanted GaN samples which had high resistivities due to defect states within the band gap that act as traps for the carriers. This may be the results of less damage produced by the smaller mass of Be, low implantation energy (40 KeV) and the resistance of GaN materials to damage by implantation, with considerable dynamic recovery of implantation-induced disorder. High-resolution x-ray diffractometry (HRXRD) was performed to detect the implantation damage. There was no difference in the rocking curves between the as-grown sample and the as-implanted sample. However, it should be noted that HRXRD results could not demonstrate conclusively that the implantation damage was not significant because HRXRD is not very sensitive to implant damage for the lower dose sample. (ii) The new annealing process for the high dose (1×10¹⁵ cm⁻²) implanted samples produced high-resistivity material. This is probably because, besides the native donors and interstitial Be, more defects were introduced by the higher dose implantation, which compensated the effects of the activated acceptors. This is similar to the case of high Mg doping in GaN.Reference Eckey, von Gfug, Holst, Hoffmann, Kaschner, siegle, Thomsen, Schineller, Heime, Heuken, Schon and Beccard ¹⁴ (iii) p-type doping was achieved for the lower dose (3×10¹⁴ cm⁻²) implanted samples, with a sheet hole concentration around 10¹³ cm⁻². To confirm the realization of p-type doping, thermal probe measurements were performed. We used Mg-doped p-type GaN and undoped n-type GaN samples as references. The results were consistent with the Hall measurements, thus confirming that the Be-implanted and annealed samples were indeed p-type.

Figure 2 depicts the SIMS profiles of Be and H for different samples. Fig.2(a) shows the depth profiles of Be and H in the as-grown sample. The measured Be concentration was due to a background level noise governed by the intrinsic impurity and the detection limit of our SIMS measurement conditions. The depth profile of Be in the as-implanted sample exhibited a peak at a depth of 120 nm as shown in Fig.2(b). The H profile in this sample was almost the same as that in the as-grown sample. There was no significant channeling tail of Be, unlike that in the Be as-implanted samples (45 keV, 5 × 10¹⁴ cm⁻²) in REF 15. We think the differences are from the implantation process, such as beam divergence or other random factors, but probably not from some factors like energy and dose, which are similar to ours. After annealing at 1100 °C for 45 s, the peak was nearer the surface than that in the as-implanted sample. The projected range was 73 nm (Fig.2(c)). The H profile remained similar. The Be redistribution in this case can be attributed to defect-assisted diffusion.Reference Wilson, Zavada, Cao, Singh, Pearton, Guo, Pennycook, Fu, Sekhar, Scarvepalli, Shul, Han, Rieger, Zolper and Abernathy ¹⁶ It is noted that some groupsReference Zolper and Pearton ^15, Reference Ronning, Linthicum, Carlson, Hartlieb, Thomson, Gehrke and Davis ¹⁷ did not see measurable redistribution of Be in GaN. In Fig. 2(d), the Be and H profiles of the sample annealed with the new process were substantially different from the others. In the top 230 nm layer, the Be profile showed a plateau-like shape, at a concentration of ∼1.2×10¹⁹ cm⁻³.

Fig.2 (a)

Fig.2 (b)

Fig.2 (c)

Fig.2 (d)

At depths beyond 0.5 µm, the Be level was much higher than that in the as-implanted sample and N₂-annealed sample, indicating enhanced in-diffusion of Be. In addition, accumulation of H was observed in the near-surface region, where the H concentration was one to two orders of magnitude higher than the background level. From the study of thermal stability of hydrogen in Reference Akasaki, Amano, Kito and Hiramatsu ² H-implanted p-type GaN,Reference Pearton, Wilson, Zavada, Han and Shul ¹⁸ the diffusion of the deuterium started at a lower temperature of about 500-700 °C and Mg-H complexes were formed, but these were completely gone by 1000 °C. However, for our samples, even an 1100 °C annealing could not entirely break the bonds of potential H complexes and remove the H. From the analysis of Hall and thermal probe measurements, it seems that there exists some relationship between the enhanced diffusion of Be and the activation of the acceptors, similar to the case of Zn diffusion and the optical activation of acceptors in Zn-implanted GaN.Reference Suski, Jun, Leszczynski, Teisseyre, Strite, Rockett, Pelzmann, Kamp and Ebeling ¹⁹

The effect of the new annealing process on Be activation can be tentatively explained as follows. After implantation, there are many Be atoms at interstitial sites, which could easily form a large concentration of substitutional-interstitial (SI) Be-related complexes. Annealing in an atmosphere containing hydrogen might cause the formation of Be_int-H complexes first. The Be_int-H complexes can move into Ga vacancies (V_Ga) more easily and are then converted to Be_Ga-H-N complexes. The energy required to break the H-bonds in the Be_Ga-H-N complexes might be lower than that needed to move the Be_int directly to the substitutional sites during conventional annealing in N₂ only. The subsequent annealing in flowing N₂ will then depassivate the complexes and activate the implanted Be. An independent inference from a first-principles calculationReference Reboredo and Pantelides ²⁰ also pointed out that an annealing stage in an ambient including hydrogen before conventional annealing in N₂ would improve activation of p-type dopants in GaN.