Shallow acceptors in GaNT. A. G. Eberlein, R. Jones, S. Öberg, and P. R. Briddon Citation: Applied Physics Letters 91, 132105 (2007); doi: 10.1063/1.2776852 View online: http://dx.doi.org/10.1063/1.2776852 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/91/13?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Properties of the main Mg-related acceptors in GaN from optical and structural studies J. Appl. Phys. 115, 053507 (2014); 10.1063/1.4862928 Passivation and activation of Mg acceptors in heavily doped GaN J. Appl. Phys. 110, 044508 (2011); 10.1063/1.3626461 Dual nature of acceptors in GaN and ZnO: The curious case of the shallow Mg Ga deep state Appl. Phys. Lett. 96, 142114 (2010); 10.1063/1.3383236 Effect of p-type activation ambient on acceptor levels in Mg-doped GaN J. Appl. Phys. 96, 415 (2004); 10.1063/1.1755856 p -type conduction in as-grown Mg-doped GaN grown by metalorganic chemical vapor deposition Appl. Phys. Lett. 72, 1748 (1998); 10.1063/1.121172

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

128.235.251.160 On: Sat, 20 Dec 2014 08:43:43

Shallow acceptors in GaNT. A. G. Eberleina� and R. JonesSchool of Physics, University of Exeter, Stocker Road, Exeter EX4 4QL, United Kingdom

S. ÖbergDepartment of Mathematics, Luleå University of Technology, Luleå S-97187, Sweden

P. R. BriddonSchool of Natural Sciences, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU,United Kingdom

�Received 3 July 2007; accepted 6 August 2007; published online 25 September 2007�

Recent high resolution photoluminescence studies of high quality Mg doped GaN show the presenceof two acceptors. One is due to Mg and the other labeled A1 has a shallower acceptor defect. Theauthors investigate likely candidates for this shallow acceptor and conclude that CN is the mostlikely possibility. The authors also show that the CN is passivated by H and the passivated complexis more stable than MgGa–H. © 2007 American Institute of Physics. �DOI: 10.1063/1.2776852�

After Si, GaN is one of the most important semiconduct-ing material and is unrivaled for bright light emission. As aresult of technological advances in GaN epitaxy, nitrides arewidely used to produce blue-green to ultraviolet emitters,ultraviolet photodetectors, and high electron mobility transis-tors. Notwithstanding remarkable success in fabricating thesedevices, the atomic processes that govern the incorporationof dopants intentional and otherwise and their interactionswith native point defects are not yet fully understood. This isparticularly true for p-type doping.

Whereas n-type doping is achieved with O and Si, pdoping is problematic. Oxygen and Si donors have levels 33and 30 meV, respectively, and these donors are ionized justabove room temperature whereas the only useful shallow ac-ceptor in GaN is Mg, which has an acceptor level about220 meV. Growth of GaN doped with Mg leads to high re-sistivity material due, in part, to the formation of Mg–H pairswhich are electrically inactive.1,2 These, however, can beeliminated either by thermal treatments at temperaturesabove �350 °C �Ref. 3� or using electron irradiation4,5 lead-ing to a very broad electron paramagnetic resonance signalattributed to substitutional Mg.2 The doping limit achievedwith Mg is about 2�1018 cm−3 carriers with Mg concentra-tions about ten times larger. Increasing Mg leads to poorermaterial quality due to formation of VN donors and com-plexes with Mg. Thus, Mg is far from the ideal acceptor andit has generally been thought that there are no other choices.However, recent photoluminescence �PL� studies6 haveshown that this picture is incomplete and the true situation isfar more complex and intriguing.

The recent PL studies on high quality homoepitaxial Mgdoped GaN layers grown on thick hydride vapor phase epi-taxy GaN substrates by metal-organic chemical vapor depo-sition �MOCVD� in a Thomas Swann reactor reveal twoemission lines due to acceptor bound excitons �ABEs� at3.466 and 3.455 eV labeled ABE1 and ABE2 and due to twoshallow acceptors labeled here A1 and A2.7

Correlated with each acceptor is a family of donor-acceptor pair �DAP� transitions. A1 is correlated with a DAPtransition with a zero phonon line at 3.27 eV, while A2 is

correlated with a broad peak with a maximum at 3.1 eV.Since the ABE1 binding energy is 0.01 eV less than ABE2the A1 acceptor has, according to Hayne’s rule, a level about0.1 eV shallower than A2. The ultraviolet light �UVL� or3.27 eV DAP emission correlated with A1 increases insamples lightly doped with Mg but has been detected in ma-terial in which Mg is not present implying that A1 has noth-ing to do with Mg.6

It is known that the 3.27 eV DAP emission along withthe ABE1 emission at 3.466 eV with which it is correlatedare strongly reduced by low energy ionizing radiation even atcryogenic temperatures implying a radiation enhanced an-nealing of A1. Moreover, in p-GaN, the 3.27 eV UVL emis-sion anneals around 500 °C �Ref. 8� but is thermally stableto higher temperatures in n-GaN. The effect of radiation andannealing suggest that the shallower acceptor A1 is a weaklybonded defect, but is certainly not Mg. Previous work7 hasassigned A2 to MgGa but left open the identity of A1. Thus,there are two shallow acceptors in MOCVD GaN doped withMg. The presence of a second shallower acceptor might ex-plain why electrical studies report an acceptor level at0.15 eV rather than the optical level at 0.22 eV.3

Although A1 is not related to Mg, studies of the excita-tion power dependence of the intensity of ABE1 surprisinglyshow that both A1 and A2 increase with Mg concentration.This could be explained by the incorporation of Mg on Gasites driving the A1 acceptor defect onto the N sublattice. Weshall return to this below.

It is one of the principal aims of this letter to identify thiscenter. We have investigated a number of candidates but theonly defect which comes close to A1 is carbon at a nitrogensite �CN�.

We carry out spin-density-functional calculations usingthe AIMPRO code. The explicit treatment of electronic corestates is avoided by using the pseudopotentials of Hartwig-sen et al.9 The calculations are done in 216 atom supercellsfor c-GaN and in 128 atoms for 2H-GaN. For the C and Mgacceptors in 2H-GaN convergence with respect to supercellsize is verified with 300 atom supercells. The increase insupercell size leads to a difference of the acceptor levels ofless than 0.01 eV for both acceptors. The basis set consists ofCartesian s-, p-, and d-type Gaussian functions centered ona�Electronic mail: eberlein@excc.ex.ac.uk

APPLIED PHYSICS LETTERS 91, 132105 �2007�

0003-6951/2007/91�13�/132105/2/$23.00 © 2007 American Institute of Physics91, 132105-1 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

128.235.251.160 On: Sat, 20 Dec 2014 08:43:43

each atom. Ga and N atoms contribute with a contractedbasis consisting of �4, 4, 5� �s , p ,d� Gaussian orbitals opti-mized for bulk GaN.10 For C, Mg, and O we used a basis of�4, 12, 24� �s , p ,d� Gaussian orbitals.

We first investigate c-GaN since the acceptor levels ofboth C and Mg are known and lie at Ev+0.215 and Ev+0.230 eV.11,12 We find that the Mg acceptor level is shal-lower than CN but by only 0.17 eV. Thus, both defects areacceptors and the disagreement with the ordering of the ex-perimental levels probably arises from the limitations of thetheory.

For w-GaN, we found the same difference in acceptorlevel of CN and MgGa. Hence, if we correct for the error wefind in c-GaN, we conclude that the acceptor level of CN inw-GaN would be 15 meV lower than that of MgGa and lie at0.210 eV. This is in good agreement with other theoreticalresults13 and experimental findings.14 We also looked at CGaand find its donor level to be deeper than SiGa by 0.2 eV.

The structure of hydrogen bound to CN contrasts withthat of MgGa. When bound to CN, H lies at an antibondingsite to C with a C–H bond length of 1.11 Å. In contrast,when bound to MgGa, H lies at an antibonding site of a Nneighbor of MgGa. The binding energy of H with C is1.66 eV and larger than the binding energy of H with Mgwhich is 1.59 eV. In both cases, H passivates the acceptoraction.

The VGa–OH defect15 has been suggested as a candidatefor A1. We find its acceptor level to lie about 0.8 eV aboveMgGa and hence place it at Ev+1.0 eV. This rules this defectout as a candidate for A1. Hence, we conclude that the mostlikely candidate for A1 is CN.

We can suggest a mechanism for the annealing of A1around 500 °C in p-GaN �Ref. 8� although it is stable atthese temperatures in n-GaN. It has been found in positronannihilation �PAS� studies in GaN:Mg, that neutral VN–Mgpairs are present in MOCVD grown material and these an-neal around 500–800 °C with activation energy around3 eV.16,17 Hence, if the annealing occurs by dissociation fol-lowed by diffusion of VN to the surface, as suggested by thePAS studies, then we can anticipate that defects on the Nsublattice are also unstable at these temperatures. In otherwords, a migrating VN defect will promote the diffusion ofCN which would then anneal around 500 °C. The samplesstudied by PAS were either highly resistive or p type. Thesituation in n-GaN counterdoped with Mg is different as thenthe formation energy of VN is so large that such defects willnot be present. Thus CN defects will be stable in n-GaN tohigher temperatures than in p type. In summary, there is thepossibility that CN is lost by instability of the N sublattice inp-GaN. It is less likely in n-GaN as the formation energy ofVN is then so large that such defects would be rare.

The second and more challenging property of the A1acceptor to explain is its instability in the presence of e-hpairs. Low energy �10 keV� irradiation, producing e-h pairs,apparently removes A1 centers even at cryogenictemperatures.4,5 It seems unlikely that CN is unstable by itselfin the presence of e-h pairs and so any instability must be

due to the complexing with another mobile defect. Now, it isknown that Mg–H defects are unstable in the presence of e-hpairs and dissociate producing H0 or H+ which are mobile.The dissociation energy of Mg–N–H being 1.59 eV is pro-vided by the recombination energy of e-h pairs. Now, mobileH could be subsequently trapped by CN or CN

− defects. Thiswould occur if the C–H bonded defect is more stable than theMg–H defect or if it is not susceptible to the same radiationenhanced dissociation that Mg–H defects suffer. Accordingto our calculations, CN–H defects are passive. Hence, theeffect of radiation is to activate Mg acceptors but passivate Cacceptors. Clearly, this should increase the 3.1 eV DAP bandand ABE2 intensity attributed to Mg and this is supported byexperiments on lightly Mg doped metal-organic vapor-phaseepitaxy GaN whereas the 3.27 eV band decreases, the bandat 3.1 eV increases in intensity.4 Finally, we point out thatthere is evidence for C–H defects in undoped MOCVDGaN.18 Fourier transform infrared studies revealed modes at2851, 2922, and 2956 cm−1. We find the stretch mode of theC–H defect to be 2855.5 cm−1 and very close to the firstdefect. The identity of the others is unclear at this stage.

In summary, we have found that C at a N site is a shal-low acceptor and a strong candidate for the shallow acceptorobserved in high quality grown Mg doped MOCVD GaN.We place its level 0.015 eV below Mg.

The authors thank Bo Monemar for helpful discussions.

1S. Nakamura, N. Iwasa, M. Senoh, and T. Mukai, Jpn. J. Appl. Phys., Part1 31, 1258 �1992�.

2M. E. Zvanut, D. M. Matlock, R. L. Henry, Daniel Koleske, and AlmaWickenden, J. Appl. Phys. 95, 1884 �2004�.

3E. Litwin-Staszewska, R. Piotrzkowski, L. Dmowski, P. Prystawko, and R.Czernecki, L. Konczewicz, J. Appl. Phys. 99, 033703 �2006�.

4O. Gelhausen, H. N. Klein, M. R. Phillips, and E. M. Goldys, Appl. Phys.Lett. 83, 3293 �2003�.

5O. Gelhausen, M. R. Phillips, E. M. Goldys, T. Paskova, B. Monemar, M.Strassburg, and A. Hoffmann, Phys. Rev. B 69, 125210 �2004�.

6M. A. Reshchikov, Appl. Phys. Lett. 88, 202104 �2006�.7B. Monemar, P. P. Paskov, J. P. Bergman, T. Paskova, S. Figge,J. Dennemarck, and D. Hommel, Phys. Status Solidi B 243, 1604 �2006�.

8W. Gotz, N. M. Johnson, J. Walker, D. P. Bour, and R. A. Street, Appl.Phys. Lett. 68, 667 �1996�.

9C. Hartwigsen, S. Goedecker, and J. Hutter, Phys. Rev. B 58, 3641�1998�.

10J. P. Goss, M. J. Shaw, and P. R. Briddon, Top. Appl. Phys. 104, 69�2007�.

11D. As, Mater. Res. Soc. Symp. Proc. 693, 17 �2002�.12D. J. As, F. Schmilgus, C. Wang, B. Schöttker, D. Schikora, and

K. Lischka, Appl. Phys. Lett. 70, 1311 �1997�.13A. F. Wright, J. Appl. Phys. 92, 2575 �2002�.14A. Armstrong, A. R. Arehart, D. Green, U. K. Mishra, J. S. Speck, and S.

A. Ringel, J. Appl. Phys. 98, 053704 �2005�.15B. Monemar, P. P. Paskov, F. Tuomisto, K. Saarinen, M. Iwaya,

S. Kamiyama, H. Amano, I. Akasaki, and S. Kimura, Physica B 376-377,440 �2006�.

16S. Hautakangas, J. Oila, M. Alatalo, K. Saarinen, L. Liszkay, D. Seghier,and H. P. Gislason, Phys. Rev. Lett. 90, 137402 �2003�.

17S. Hautakangas, K. Saarinen, L. Liszkay, J. A. Freitas,Jr., and R. L. Henry,Phys. Rev. B 72, 165303 �2005�.

18M. O. Manasreh, J. M. Baranowski, K. Pakula, H. X. Jiang, and JingyuLin, Appl. Phys. Lett. 75, 659 �2006�.

132105-2 Eberlein et al. Appl. Phys. Lett. 91, 132105 �2007�

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

128.235.251.160 On: Sat, 20 Dec 2014 08:43:43